diff --git a/common/arg.cpp b/common/arg.cpp index 8c119292d..dac587a07 100644 --- a/common/arg.cpp +++ b/common/arg.cpp @@ -447,6 +447,12 @@ bool common_params_handle_models(common_params & params, llama_example curr_ex) opts.download_mtp = spec_type_draft_mtp; opts.download_mmproj = !params.no_mmproj; + // sub-models (draft, mmproj, vocoder) are explicitly specified by the user, + // so we should not auto-discover mtp/mmproj siblings for them + common_download_opts sub_opts = opts; + sub_opts.download_mtp = false; + sub_opts.download_mmproj = false; + try { auto res = common_params_handle_model(params.model, opts); if (params.no_mmproj) { @@ -458,7 +464,7 @@ bool common_params_handle_models(common_params & params, llama_example curr_ex) // only download mmproj if the current example is using it for (const auto & ex : mmproj_examples) { if (curr_ex == ex) { - common_params_handle_model(params.mmproj, opts); + common_params_handle_model(params.mmproj, sub_opts); break; } } @@ -471,8 +477,8 @@ bool common_params_handle_models(common_params & params, llama_example curr_ex) params.speculative.draft.mparams.url.empty()) { params.speculative.draft.mparams.path = res.mtp.path; } - common_params_handle_model(params.speculative.draft.mparams, opts); - common_params_handle_model(params.vocoder.model, opts); + common_params_handle_model(params.speculative.draft.mparams, sub_opts); + common_params_handle_model(params.vocoder.model, sub_opts); return true; } catch (const common_skip_download_exception &) { return false; diff --git a/common/imatrix-loader.cpp b/common/imatrix-loader.cpp new file mode 100644 index 000000000..efe9aecee --- /dev/null +++ b/common/imatrix-loader.cpp @@ -0,0 +1,165 @@ +#include "imatrix-loader.h" +#include "common.h" +#include "log.h" +#include "gguf.h" + +#include +#include +#include + +static bool common_imatrix_load_legacy(const std::string & fname, common_imatrix & imatrix) { + std::ifstream in(fname, std::ios::binary); + if (!in) { + LOG_ERR("%s: failed to open %s\n", __func__, fname.c_str()); + return false; + } + + int n_entries; + in.read((char *) &n_entries, sizeof(n_entries)); + if (in.fail() || n_entries < 1) { + LOG_ERR("%s: no data in file %s\n", __func__, fname.c_str()); + return false; + } + + for (int i = 0; i < n_entries; ++i) { + int32_t len = 0; + in.read((char *) &len, sizeof(len)); + std::vector name_as_vec(len + 1); + in.read((char *) name_as_vec.data(), len); + if (in.fail()) { + LOG_ERR("%s: failed reading name for entry %d from %s\n", __func__, i + 1, fname.c_str()); + return false; + } + name_as_vec[len] = 0; + std::string name{ name_as_vec.data() }; + + int32_t ncall = 0; + in.read((char *) &ncall, sizeof(ncall)); + int32_t nval = 0; + in.read((char *) &nval, sizeof(nval)); + if (in.fail() || nval < 1) { + LOG_ERR("%s: failed reading number of values for entry %d\n", __func__, i); + return false; + } + + auto & e = imatrix.entries[std::move(name)]; + e.sums.resize(nval); + in.read((char *) e.sums.data(), nval * sizeof(float)); + if (in.fail()) { + LOG_ERR("%s: failed reading data for entry %d\n", __func__, i); + return false; + } + + e.counts.resize(1); + e.counts[0] = ncall; + } + + // the trailing data (chunk count + dataset name) is optional + if (in.peek() != EOF) { + int32_t n_calls = 0; + in.read((char *) &n_calls, sizeof(n_calls)); + imatrix.chunk_count = n_calls; + + if (!in.fail()) { + int32_t len = 0; + in.read((char *) &len, sizeof(len)); + if (!in.fail() && len > 0) { + std::vector dataset(len + 1, 0); + in.read(dataset.data(), len); + if (!in.fail()) { + imatrix.datasets.push_back(dataset.data()); + } + } + } + } + + imatrix.chunk_size = 0; + imatrix.is_legacy = true; + + return true; +} + +bool common_imatrix_load(const std::string & fname, common_imatrix & imatrix) { + struct ggml_context * ctx = nullptr; + struct gguf_init_params meta_gguf_params = { + /* .no_alloc = */ false, + /* .ctx = */ &ctx, + }; + struct gguf_context * ctx_gguf = gguf_init_from_file(fname.c_str(), meta_gguf_params); + if (!ctx_gguf) { + return common_imatrix_load_legacy(fname, imatrix); + } + + const int32_t n_entries = gguf_get_n_tensors(ctx_gguf); + if (n_entries < 1) { + LOG_ERR("%s: no data in file %s\n", __func__, fname.c_str()); + gguf_free(ctx_gguf); + ggml_free(ctx); + return false; + } + + const int64_t datasets_key = gguf_find_key(ctx_gguf, LLM_KV_IMATRIX_DATASETS); + const int64_t chunk_count_key = gguf_find_key(ctx_gguf, LLM_KV_IMATRIX_CHUNK_COUNT); + const int64_t chunk_size_key = gguf_find_key(ctx_gguf, LLM_KV_IMATRIX_CHUNK_SIZE); + + if (datasets_key != -1 && gguf_get_arr_type(ctx_gguf, datasets_key) == GGUF_TYPE_STRING) { + const int64_t n = gguf_get_arr_n(ctx_gguf, datasets_key); + imatrix.datasets.reserve(imatrix.datasets.size() + n); + for (int64_t i = 0; i < n; ++i) { + imatrix.datasets.push_back(gguf_get_arr_str(ctx_gguf, datasets_key, i)); + } + } + + imatrix.has_metadata = (datasets_key != -1 && chunk_count_key != -1 && chunk_size_key != -1); + imatrix.chunk_count = (chunk_count_key != -1) ? gguf_get_val_u32(ctx_gguf, chunk_count_key) : 0; + imatrix.chunk_size = (chunk_size_key != -1) ? gguf_get_val_u32(ctx_gguf, chunk_size_key) : 0; + + const std::string in_sum2_suffix{ ".in_sum2" }; + const std::string counts_suffix{ ".counts" }; + + std::map> sums_counts_for; + + for (struct ggml_tensor * cur = ggml_get_first_tensor(ctx); cur; cur = ggml_get_next_tensor(ctx, cur)) { + std::string name = cur->name; + + if (name.empty()) { continue; } + + if (string_remove_suffix(name, in_sum2_suffix)) { + sums_counts_for[std::move(name)].first = cur; + } else if (string_remove_suffix(name, counts_suffix)) { + sums_counts_for[std::move(name)].second = cur; + } + } + + for (const auto & sc : sums_counts_for) { + const std::string & name = sc.first; + const struct ggml_tensor * in_sum2 = sc.second.first; + const struct ggml_tensor * counts = sc.second.second; + + if (!in_sum2 || !counts) { + LOG_ERR("%s: mismatched sums and counts for %s\n", __func__, name.c_str()); + gguf_free(ctx_gguf); + ggml_free(ctx); + return false; + } + + auto & e = imatrix.entries[name]; + + const int64_t nval = ggml_nelements(in_sum2); + const int64_t ncounts = ggml_nelements(counts); + + e.sums.resize(nval); + for (int64_t j = 0; j < nval; ++j) { + e.sums[j] = ((const float *) in_sum2->data)[j]; + } + + e.counts.resize(ncounts); + for (int64_t j = 0; j < ncounts; ++j) { + e.counts[j] = std::lround(((const float *) counts->data)[j]); + } + } + + gguf_free(ctx_gguf); + ggml_free(ctx); + return true; +} diff --git a/common/imatrix-loader.h b/common/imatrix-loader.h new file mode 100644 index 000000000..ed00d724a --- /dev/null +++ b/common/imatrix-loader.h @@ -0,0 +1,26 @@ +#pragma once + +#include +#include +#include +#include + +inline constexpr const char * LLM_KV_IMATRIX_DATASETS = "imatrix.datasets"; +inline constexpr const char * LLM_KV_IMATRIX_CHUNK_COUNT = "imatrix.chunk_count"; +inline constexpr const char * LLM_KV_IMATRIX_CHUNK_SIZE = "imatrix.chunk_size"; + +struct common_imatrix_entry { + std::vector sums; + std::vector counts; +}; + +struct common_imatrix { + std::map entries; + std::vector datasets; + int32_t chunk_count = 0; + int32_t chunk_size = 0; + bool is_legacy = false; + bool has_metadata = false; +}; + +bool common_imatrix_load(const std::string & fname, common_imatrix & imatrix); diff --git a/ggml/src/ggml-cuda/mmvq.cu b/ggml/src/ggml-cuda/mmvq.cu index 4b0426590..bdfbfd2d3 100644 --- a/ggml/src/ggml-cuda/mmvq.cu +++ b/ggml/src/ggml-cuda/mmvq.cu @@ -682,12 +682,16 @@ static __global__ void mul_mat_vec_q( template __launch_bounds__(get_mmvq_mmid_max_batch_for_device()*ggml_cuda_get_physical_warp_size(), 1) static __global__ void mul_mat_vec_q_moe( - const void * __restrict__ vx, const void * __restrict__ vy, const int32_t * __restrict__ ids, - float * __restrict__ dst, + const void * vx_ptr, const void * vy_ptr, const int32_t * ids_ptr, + float * dst_ptr, const uint32_t ncols_x, const uint3 nchannels_y, const uint32_t nrows_x, const uint32_t stride_row_x, const uint32_t stride_col_y, const uint32_t stride_col_dst, const uint32_t stride_channel_x, const uint32_t stride_channel_y, const uint32_t stride_channel_dst, const uint32_t ncols_dst, const uint32_t ids_stride) { + const void * GGML_CUDA_RESTRICT vx = vx_ptr; + const void * GGML_CUDA_RESTRICT vy = vy_ptr; + const int32_t * GGML_CUDA_RESTRICT ids = ids_ptr; + float * GGML_CUDA_RESTRICT dst = dst_ptr; constexpr int qk = ggml_cuda_type_traits::qk; constexpr int qi = ggml_cuda_type_traits::qi; @@ -707,6 +711,7 @@ static __global__ void mul_mat_vec_q_moe( return; } + ggml_cuda_pdl_sync(); const uint32_t channel_x = ids[channel_dst + token_idx * ids_stride]; const uint32_t channel_y = fastmodulo(channel_dst, nchannels_y); @@ -726,6 +731,8 @@ static __global__ void mul_mat_vec_q_moe( } } + ggml_cuda_pdl_lc(); + // Warp-level reduction only - no shared memory needed #pragma unroll for (int i = 0; i < c_rows_per_block; ++i) { @@ -794,8 +801,9 @@ static void mul_mat_vec_q_moe_launch( const int64_t nblocks_rows = (nrows_x + rows_per_block - 1) / rows_per_block; const dim3 block_nums(nblocks_rows, nchannels_dst); const dim3 block_dims(warp_size, ncols_dst); + const ggml_cuda_kernel_launch_params launch_params = ggml_cuda_kernel_launch_params(block_nums, block_dims, 0, stream); - mul_mat_vec_q_moe<<>>( + ggml_cuda_kernel_launch(mul_mat_vec_q_moe, launch_params, vx, vy, ids, dst, ncols_x, nchannels_y, nrows_x, stride_row_x, stride_col_y, stride_col_dst, stride_channel_x, stride_channel_y, stride_channel_dst, diff --git a/src/llama-adapter.cpp b/src/llama-adapter.cpp index 4a1aaa955..3e0fe66af 100644 --- a/src/llama-adapter.cpp +++ b/src/llama-adapter.cpp @@ -41,7 +41,7 @@ bool llama_adapter_cvec::init(const llama_model & model) { auto it = ctx_map.find(buft); if (it == ctx_map.end()) { ggml_init_params params = { - /*.mem_size =*/ hparams.n_layer*ggml_tensor_overhead(), + /*.mem_size =*/ hparams.n_layer()*ggml_tensor_overhead(), /*.mem_buffer =*/ NULL, /*.no_alloc =*/ true, }; @@ -61,9 +61,9 @@ bool llama_adapter_cvec::init(const llama_model & model) { }; // make tensors - tensors.reserve(hparams.n_layer); + tensors.reserve(hparams.n_layer()); tensors.push_back(nullptr); // there's never a tensor for layer 0 - for (size_t il = 1; il < hparams.n_layer; il++) { + for (size_t il = 1; il < hparams.n_layer(); il++) { ggml_backend_buffer_type_t buft = model.select_buft(il); ggml_context * ctx = ctx_for_buft(buft); if (!ctx) { @@ -121,7 +121,7 @@ bool llama_adapter_cvec::apply( layer_start = il_start; layer_end = il_end; - for (size_t il = 1; il < hparams.n_layer; il++) { + for (size_t il = 1; il < hparams.n_layer(); il++) { assert(tensors[il] != nullptr); const size_t off = n_embd * (il - 1); // buffer doesn't have data for layer 0, since it's never present diff --git a/src/llama-context.cpp b/src/llama-context.cpp index 3c02fdb40..ff4544bdc 100644 --- a/src/llama-context.cpp +++ b/src/llama-context.cpp @@ -344,7 +344,7 @@ llama_context::llama_context( // enabling pipeline parallelism in the scheduler increases memory usage, so it is only done when necessary bool pipeline_parallel = model.n_devices() > 1 && - model.n_gpu_layers() > model.hparams.n_layer && + model.n_gpu_layers() > model.hparams.n_layer() && model.split_mode() == LLAMA_SPLIT_MODE_LAYER && cparams.offload_kqv && !model.has_tensor_overrides(); @@ -2361,7 +2361,7 @@ llm_graph_cb llama_context::graph_get_cb() const { // norm may be automatically assigned to the backend of the previous layer, increasing data transfer between backends // FIXME: fix in ggml_backend_sched - const bool full_offload = model.n_gpu_layers() > model.hparams.n_layer; + const bool full_offload = model.n_gpu_layers() > model.hparams.n_layer(); if (ubatch.n_tokens < 32 || full_offload) { if (il != -1 && strcmp(name, "norm") == 0) { const auto & dev_layer = model.dev_layer(il); @@ -3426,7 +3426,7 @@ llama_context * llama_init_from_model( if (params.flash_attn_type != LLAMA_FLASH_ATTN_TYPE_DISABLED && ggml_is_quantized(params.type_k)) { const uint32_t blck_size = ggml_blck_size(params.type_k); - for (uint32_t il = 0; il < model->hparams.n_layer; ++il) { + for (uint32_t il = 0; il < model->hparams.n_layer(); ++il) { if (model->hparams.n_embd_head_k(il) % blck_size != 0) { LLAMA_LOG_ERROR("%s: K cache type %s with block size %u does not divide n_embd_head_k=%u\n", __func__, ggml_type_name(params.type_k), blck_size, model->hparams.n_embd_head_k(il)); @@ -3437,7 +3437,7 @@ llama_context * llama_init_from_model( if (params.flash_attn_type != LLAMA_FLASH_ATTN_TYPE_DISABLED && ggml_is_quantized(params.type_v)) { const uint32_t blck_size = ggml_blck_size(params.type_v); - for (uint32_t il = 0; il < model->hparams.n_layer; ++il) { + for (uint32_t il = 0; il < model->hparams.n_layer(); ++il) { if (model->hparams.n_embd_head_v(il) % blck_size != 0) { LLAMA_LOG_ERROR("%s: V cache type %s with block size %u does not divide n_embd_head_v=%u\n", __func__, ggml_type_name(params.type_v), blck_size, model->hparams.n_embd_head_v(il)); @@ -3459,7 +3459,7 @@ llama_context * llama_init_from_model( } if (params.ctx_type == LLAMA_CONTEXT_TYPE_MTP && - model->hparams.nextn_predict_layers == 0) { + model->hparams.n_layer_nextn == 0) { LLAMA_LOG_WARN("%s: context type MTP requested but model doesn't contain MTP layers\n", __func__); return nullptr; } diff --git a/src/llama-graph.cpp b/src/llama-graph.cpp index 42cc5c23f..bcf04fd5b 100644 --- a/src/llama-graph.cpp +++ b/src/llama-graph.cpp @@ -1005,7 +1005,7 @@ llm_graph_context::llm_graph_context(const llm_graph_params & params) : cparams (params.cparams), ubatch (params.ubatch), n_embd (hparams.n_embd), - n_layer (hparams.n_layer), + n_layer (hparams.n_layer()), n_rot (hparams.n_rot()), n_ctx (cparams.n_ctx), n_head (hparams.n_head()), diff --git a/src/llama-hparams.cpp b/src/llama-hparams.cpp index 087afec55..e1e49d1cc 100644 --- a/src/llama-hparams.cpp +++ b/src/llama-hparams.cpp @@ -7,31 +7,38 @@ void llama_hparams::set_swa_pattern(uint32_t n_pattern, bool dense_first) { if (dense_first) { - for (uint32_t il = 0; il < n_layer; ++il) { + for (uint32_t il = 0; il < n_layer(); ++il) { is_swa_impl[il] = n_pattern == 0 || (il % n_pattern != 0); } } else { - for (uint32_t il = 0; il < n_layer; ++il) { + for (uint32_t il = 0; il < n_layer(); ++il) { is_swa_impl[il] = n_pattern == 0 || (il % n_pattern < (n_pattern - 1)); } } + + for (uint32_t il = n_layer(); il < n_layer_all; ++il) { + is_swa_impl[il] = false; + } } -// TODO: implement -//void llama_hparams::set_recr_pattern(uint32_t n_pattern, bool dense_first) { -// if (dense_first) { -// for (uint32_t il = 0; il < n_layer; ++il) { -// is_recr_impl[il] = n_pattern == 0 || (il % n_pattern != 0); -// } -// } else { -// for (uint32_t il = 0; il < n_layer; ++il) { -// is_recr_impl[il] = n_pattern == 0 || (il % n_pattern < (n_pattern - 1)); -// } -// } -//} +void llama_hparams::set_recr_pattern(uint32_t n_pattern, bool dense_first) { + if (dense_first) { + for (uint32_t il = 0; il < n_layer(); ++il) { + is_recr_impl[il] = n_pattern == 0 || (il % n_pattern != 0); + } + } else { + for (uint32_t il = 0; il < n_layer(); ++il) { + is_recr_impl[il] = n_pattern == 0 || (il % n_pattern < (n_pattern - 1)); + } + } + + for (uint32_t il = n_layer(); il < n_layer_all; ++il) { + is_recr_impl[il] = false; + } +} bool llama_hparams::is_swa_any() const { - for (uint32_t il = 0; il < n_layer; ++il) { + for (uint32_t il = 0; il < n_layer_all; ++il) { if (is_swa_impl[il]) { return true; } @@ -41,7 +48,7 @@ bool llama_hparams::is_swa_any() const { } uint32_t llama_hparams::n_head(uint32_t il) const { - if (il < n_layer) { + if (il < n_layer_all) { return n_head_arr[il]; } @@ -49,7 +56,7 @@ uint32_t llama_hparams::n_head(uint32_t il) const { } uint32_t llama_hparams::n_head_kv(uint32_t il) const { - if (il < n_layer) { + if (il < n_layer_all) { return n_head_kv_arr[il]; } @@ -57,7 +64,7 @@ uint32_t llama_hparams::n_head_kv(uint32_t il) const { } uint32_t llama_hparams::n_ff(uint32_t il) const { - if (il < n_layer) { + if (il < n_layer_all) { return n_ff_arr[il]; } @@ -76,7 +83,7 @@ uint32_t llama_hparams::n_gqa(uint32_t il) const { } uint32_t llama_hparams::n_rot(uint32_t il) const { - if (il < n_layer) { + if (il < n_layer_all) { return is_swa(il) ? n_rot_swa : n_rot_full; } @@ -98,7 +105,7 @@ uint32_t llama_hparams::n_embd_out() const { } uint32_t llama_hparams::n_embd_head_k(uint32_t il) const { - if (il < n_layer) { + if (il < n_layer_all) { return is_swa(il) ? n_embd_head_k_swa : n_embd_head_k_full; } @@ -106,7 +113,7 @@ uint32_t llama_hparams::n_embd_head_k(uint32_t il) const { } uint32_t llama_hparams::n_embd_head_v(uint32_t il) const { - if (il < n_layer) { + if (il < n_layer_all) { return is_swa(il) ? n_embd_head_v_swa : n_embd_head_v_full; } @@ -127,7 +134,7 @@ uint32_t llama_hparams::n_embd_v_gqa(uint32_t il) const { bool llama_hparams::is_n_embd_k_gqa_variable() const { const uint32_t val = n_embd_k_gqa(); - for (uint32_t il = 0; il < n_layer; ++il) { + for (uint32_t il = 0; il < n_layer_all; ++il) { if (val != n_embd_k_gqa(il)) { return true; } @@ -138,7 +145,7 @@ bool llama_hparams::is_n_embd_k_gqa_variable() const { bool llama_hparams::is_n_embd_v_gqa_variable() const { const uint32_t val = n_embd_v_gqa(); - for (uint32_t il = 0; il < n_layer; ++il) { + for (uint32_t il = 0; il < n_layer_all; ++il) { if (val != n_embd_v_gqa(il)) { return true; } @@ -149,7 +156,7 @@ bool llama_hparams::is_n_embd_v_gqa_variable() const { uint32_t llama_hparams::n_embd_k_gqa_max() const { uint32_t val = n_embd_k_gqa(); - for (uint32_t il = 0; il < n_layer; ++il) { + for (uint32_t il = 0; il < n_layer_all; ++il) { val = std::max(val, n_embd_k_gqa(il)); } @@ -158,7 +165,7 @@ uint32_t llama_hparams::n_embd_k_gqa_max() const { uint32_t llama_hparams::n_embd_v_gqa_max() const { uint32_t val = n_embd_v_gqa(); - for (uint32_t il = 0; il < n_layer; ++il) { + for (uint32_t il = 0; il < n_layer_all; ++il) { val = std::max(val, n_embd_v_gqa(il)); } @@ -207,11 +214,11 @@ uint32_t llama_hparams::n_embd_s() const { } bool llama_hparams::is_recr(uint32_t il) const { - if (il < n_layer) { + if (il < n_layer_all) { return is_recr_impl[il]; } - GGML_ABORT("%s: il (%u) out of bounds (n_layer: %u)\n", __func__, il, n_layer); + GGML_ABORT("%s: il (%u) out of bounds (n_layer_all: %u)\n", __func__, il, n_layer_all); } uint32_t llama_hparams::n_pos_per_embd() const { @@ -219,11 +226,11 @@ uint32_t llama_hparams::n_pos_per_embd() const { } bool llama_hparams::is_swa(uint32_t il) const { - if (il < n_layer) { + if (il < n_layer_all) { return is_swa_impl[il]; } - GGML_ABORT("fatal error"); + GGML_ABORT("%s: il (%u) out of bounds (n_layer_all: %u)\n", __func__, il, n_layer_all); } bool llama_hparams::is_mla() const { @@ -242,12 +249,6 @@ uint32_t llama_hparams::n_embd_head_v_mla() const { } bool llama_hparams::has_kv(uint32_t il) const { - if (kv_only_nextn) { - // MTP head: only the trailing nextn_predict_layers blocks own a KV cache; - // the leading trunk blocks are not executed in this graph. - return nextn_predict_layers > 0 && il >= (n_layer - nextn_predict_layers); - } - if (n_layer_kv_from_start >= 0) { if (il < (uint32_t) n_layer_kv_from_start) { return true; @@ -260,16 +261,8 @@ bool llama_hparams::has_kv(uint32_t il) const { return true; } -uint32_t llama_hparams::n_layer_kv() const { - uint32_t res = 0; - - for (uint32_t il = 0; il < n_layer; ++il) { - if (has_kv(il)) { - res++; - } - } - - return res; +uint32_t llama_hparams::n_layer() const { + return n_layer_all - n_layer_nextn; } bool llama_hparams::use_mrope() const { diff --git a/src/llama-hparams.h b/src/llama-hparams.h index e8ed4dd74..fde6183e8 100644 --- a/src/llama-hparams.h +++ b/src/llama-hparams.h @@ -48,12 +48,15 @@ struct llama_hparams { uint32_t n_ctx_train; // context size the model was trained on uint32_t n_embd; - uint32_t n_layer; - int32_t n_layer_kv_from_start = -1; // if non-negative, the first n_layer_kv_from_start layers have KV cache + uint32_t n_layer_all; + uint32_t n_layer_nextn = 0; uint32_t n_expert = 0; uint32_t n_expert_used = 0; uint32_t n_rel_attn_bkts = 0; + // TODO: this needs to be reworked + int32_t n_layer_kv_from_start = -1; // if non-negative, the first n_layer_kv_from_start layers have KV cache + // different head size for full_attention and SWA layers uint32_t n_embd_head_k_full; // dimension of keys (d_k). d_q is assumed to be the same, but there are n_head q heads, and only n_head_kv k-v heads uint32_t n_embd_head_v_full; // dimension of values (d_v) aka n_embd_head @@ -96,9 +99,6 @@ struct llama_hparams { uint32_t expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_NONE; uint32_t moe_every_n_layers = 0; uint32_t moe_latent_size = 0; - uint32_t nextn_predict_layers = 0; - - bool kv_only_nextn = false; // if true, only the last nextn_predict_layers blocks have a KV cache (MTP head arches) float f_norm_eps; float f_norm_rms_eps; @@ -272,8 +272,7 @@ struct llama_hparams { bool is_swa(uint32_t il) const; - // TODO: implement - //void set_recr_pattern(uint32_t n_pattern, bool dense_first = false); + void set_recr_pattern(uint32_t n_pattern, bool dense_first = false); // whether or not the given layer is recurrent (for hybrid models) bool is_recr(uint32_t il) const; @@ -329,8 +328,8 @@ struct llama_hparams { bool has_kv(uint32_t il) const; - // number of layers for which has_kv() returns true - uint32_t n_layer_kv() const; + // number of effective layers (excludes nextn layers) + uint32_t n_layer() const; // note that this function uses different SWA parameters from those in the hparams // note: inlined on purpose for performance reasons diff --git a/src/llama-kv-cache.cpp b/src/llama-kv-cache.cpp index 015974084..31cbd905f 100644 --- a/src/llama-kv-cache.cpp +++ b/src/llama-kv-cache.cpp @@ -101,7 +101,7 @@ llama_kv_cache::llama_kv_cache( GGML_ASSERT(kv_size % n_pad == 0); - const uint32_t n_layer_kv = hparams.n_layer_kv(); + const uint32_t n_layer = hparams.n_layer_all; // define a comparator for the buft -> ctx map to ensure that the order is well-defined: struct ggml_backend_buft_comparator { @@ -116,7 +116,7 @@ llama_kv_cache::llama_kv_cache( auto it = ctx_map.find(buft); if (it == ctx_map.end()) { ggml_init_params params = { - /*.mem_size =*/ size_t(2u*(1 + n_stream)*n_layer_kv*ggml_tensor_overhead()), + /*.mem_size =*/ size_t(2u*(1 + n_stream)*n_layer*ggml_tensor_overhead()), /*.mem_buffer =*/ NULL, /*.no_alloc =*/ true, }; @@ -164,7 +164,7 @@ llama_kv_cache::llama_kv_cache( const bool is_mla = hparams.is_mla(); - for (uint32_t il = 0; il < hparams.n_layer; il++) { + for (uint32_t il = 0; il < n_layer; il++) { if (!hparams.has_kv(il)) { // LLAMA_LOG_DEBUG("%s: layer %3d: does not have KV cache\n", __func__, il); continue; @@ -234,7 +234,7 @@ llama_kv_cache::llama_kv_cache( if (reuse) { LLAMA_LOG_DEBUG("%s: reusing layers:\n", __func__); - for (uint32_t il = 0; il < hparams.n_layer; il++) { + for (uint32_t il = 0; il < n_layer; il++) { const int32_t il_reuse = reuse(il); if (il_reuse < 0) { diff --git a/src/llama-memory-recurrent.cpp b/src/llama-memory-recurrent.cpp index dacb2933d..43d0cca48 100644 --- a/src/llama-memory-recurrent.cpp +++ b/src/llama-memory-recurrent.cpp @@ -26,7 +26,7 @@ llama_memory_recurrent::llama_memory_recurrent( uint32_t n_seq_max, uint32_t n_rs_seq, const layer_filter_cb & filter) : hparams(model.hparams), n_seq_max(n_seq_max) { - const int32_t n_layer = hparams.n_layer; + const int32_t n_layer = hparams.n_layer(); head = 0; size = mem_size; @@ -863,7 +863,7 @@ void llama_memory_recurrent::state_write_meta(llama_io_write_i & io, const std:: void llama_memory_recurrent::state_write_data(llama_io_write_i & io, const std::vector> & cell_ranges) const { const uint32_t s_trans = 0; - const uint32_t n_layer = hparams.n_layer; + const uint32_t n_layer = hparams.n_layer(); io.write(&s_trans, sizeof(s_trans)); io.write(&n_layer, sizeof(n_layer)); @@ -1047,8 +1047,8 @@ bool llama_memory_recurrent::state_read_data(llama_io_read_i & io, uint32_t cell io.read(&s_trans, sizeof(s_trans)); io.read(&n_layer, sizeof(n_layer)); - if (n_layer != hparams.n_layer) { - LLAMA_LOG_ERROR("%s: mismatched layer count (%u instead of %u)\n", __func__, n_layer, hparams.n_layer); + if (n_layer != hparams.n_layer()) { + LLAMA_LOG_ERROR("%s: mismatched layer count (%u instead of %u)\n", __func__, n_layer, hparams.n_layer()); return false; } if (cell_count > size) { diff --git a/src/llama-model-loader.cpp b/src/llama-model-loader.cpp index 6c8a72735..68fb73c96 100644 --- a/src/llama-model-loader.cpp +++ b/src/llama-model-loader.cpp @@ -1051,10 +1051,10 @@ struct ggml_tensor * llama_model_loader::create_tensor( if (it == ctx_map.end()) { // one ggml context per buffer type int max_n_tensors = n_tensors; - max_n_tensors += 1; // duplicated output tensor - max_n_tensors += hparams.n_layer*2; // duplicated rope freq tensors + max_n_tensors += 1; // duplicated output tensor + max_n_tensors += hparams.n_layer()*2; // duplicated rope freq tensors if (files.empty()) { - max_n_tensors += hparams.n_layer*256; // this should be well above what any model actually uses + max_n_tensors += hparams.n_layer()*256; // this should be well above what any model actually uses } const size_t ctx_size = ggml_tensor_overhead()*max_n_tensors; diff --git a/src/llama-model-saver.cpp b/src/llama-model-saver.cpp index 26fda1abf..b05228780 100644 --- a/src/llama-model-saver.cpp +++ b/src/llama-model-saver.cpp @@ -77,7 +77,7 @@ void llama_model_saver::add_kv(const enum llm_kv key, const char value) { template void llama_model_saver::add_kv(const enum llm_kv key, const Container & value, const bool per_layer) { GGML_ASSERT(model != nullptr || !per_layer); - const size_t n_values = per_layer ? size_t(model->hparams.n_layer) : value.size(); + const size_t n_values = per_layer ? size_t(model->hparams.n_layer()) : value.size(); GGML_ASSERT(n_values <= value.size()); if (n_values == 0) { @@ -206,7 +206,7 @@ void llama_model_saver::add_kv_from_model() { if (hparams.n_embd_out_impl > 0) { add_kv(LLM_KV_EMBEDDING_LENGTH_OUT, hparams.n_embd_out_impl); } - add_kv(LLM_KV_BLOCK_COUNT, hparams.n_layer); + add_kv(LLM_KV_BLOCK_COUNT, hparams.n_layer_all); add_kv(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead); add_kv(LLM_KV_FEED_FORWARD_LENGTH, hparams.n_ff_arr, true); add_kv(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp); @@ -227,7 +227,7 @@ void llama_model_saver::add_kv_from_model() { add_kv(LLM_KV_EXPERT_GROUP_SCALE, hparams.expert_group_scale); add_kv(LLM_KV_EXPERTS_PER_GROUP, hparams.n_group_experts); add_kv(LLM_KV_MOE_EVERY_N_LAYERS, hparams.moe_every_n_layers); - add_kv(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers); + add_kv(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn); add_kv(LLM_KV_NUM_DEEPSTACK_LAYERS, hparams.n_deepstack_layers); add_kv(LLM_KV_POOLING_TYPE, uint32_t(hparams.pooling_type)); add_kv(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); diff --git a/src/llama-model.cpp b/src/llama-model.cpp index 994c29122..86097057e 100644 --- a/src/llama-model.cpp +++ b/src/llama-model.cpp @@ -531,7 +531,7 @@ struct ggml_backend_meta_split_state llama_meta_device_get_split_state(const str rotation = get_il_eff(il) % ud->n_devices; } else { il = 0; - rotation = hparams.n_layer % ud->n_devices; + rotation = hparams.n_layer() % ud->n_devices; } const ggml_tensor * tensor_axis_0 = suffix.empty() ? tensor : ud->model->get_tensor((prefix + suffix).c_str()); if (tensor_axis_0 == nullptr) { @@ -686,10 +686,12 @@ struct ggml_backend_meta_split_state llama_meta_device_get_split_state(const str }; auto get_split_granularity = [&](int64_t blck_size, uint32_t il, const std::vector> & segments) -> std::vector { + // for better performance it may make sense to round up blck_size to a higher power of 2 so that more efficient kernels can be used if (hparams.is_recr(il)) { // linear attention - const int64_t head_dim = hparams.ssm_d_state; - const int64_t granularity_qkv = std::lcm(blck_size, head_dim); + const int64_t head_dim = hparams.ssm_d_state; + const int64_t blck_size_perf = std::lcm(blck_size, 128); + const int64_t granularity_qkv = std::lcm(blck_size_perf, head_dim); if (std::regex_match(tensor_name, pattern_qkv_weight) || std::regex_match(tensor_name, pattern_attn_gate_weight) || std::regex_match(tensor_name, pattern_ssm_conv1d) || std::regex_match(tensor_name, pattern_ssm_out_weight)) { return std::vector(segments.size(), granularity_qkv); @@ -711,17 +713,24 @@ struct ggml_backend_meta_split_state llama_meta_device_get_split_state(const str // regular attention const uint32_t n_gqa = hparams.n_gqa(il); const uint32_t n_embd_q = n_gqa * hparams.n_embd_head_k(il); - if (std::regex_match(tensor_name, pattern_attn_sinks)) { - GGML_ASSERT(segments.size() == 1); - return {std::lcm(n_embd_q, blck_size)/n_embd_q * n_gqa}; + + // to handle head sizes like 80, only increase granularity while it doesn't cause underutilization + int64_t blck_size_perf = blck_size; + while (blck_size_perf < 128 && blck_size_perf*ud->n_devices < n_embd_q) { + blck_size_perf *= 2; } - const int64_t granularity_q = std::lcm(n_embd_q, blck_size); + if (std::regex_match(tensor_name, pattern_attn_sinks)) { + GGML_ASSERT(segments.size() == 1); + return {std::lcm(n_embd_q, blck_size_perf)/n_embd_q * n_gqa}; + } + + const int64_t granularity_q = std::lcm(n_embd_q, blck_size_perf); if (std::regex_match(tensor_name, pattern_q_weight) || std::regex_match(tensor_name, pattern_q_bias)) { GGML_ASSERT(segments.size() == 1); // some models have Q gate tensors, for those cases the granularity needs to be doubled: if (ud->model->arch == LLM_ARCH_QWEN3NEXT || ud->model->arch == LLM_ARCH_QWEN35 || ud->model->arch == LLM_ARCH_QWEN35MOE) { - return {std::lcm(2*n_embd_q, blck_size)}; + return {std::lcm(2*n_embd_q, blck_size_perf)}; } return {granularity_q}; } @@ -746,8 +755,9 @@ struct ggml_backend_meta_split_state llama_meta_device_get_split_state(const str // FFN if (std::regex_match(tensor_name, pattern_ffn_up_gate_weight) || std::regex_match(tensor_name, pattern_ffn_up_gate_bias) || std::regex_match(tensor_name, pattern_ffn_gate_up_weight) || std::regex_match(tensor_name, pattern_ffn_down_weight)) { + const int64_t blck_size_perf = std::lcm(blck_size, 128); GGML_ASSERT(segments.size() == 1); - return {blck_size}; + return {blck_size_perf}; } // everything else @@ -760,7 +770,6 @@ struct ggml_backend_meta_split_state llama_meta_device_get_split_state(const str tensor_config tc = get_tensor_config(); split_state.axis = tc.axis; if (split_state.axis >= 0 && split_state.axis < GGML_MAX_DIMS) { - const int64_t ne_full = tensor->ne[split_state.axis]; const int64_t blck_size = ggml_blck_size(tc.tensor_axis_0->type); const float * tensor_split = ud->model->tensor_split(); std::vector tensor_split_scan; @@ -777,7 +786,6 @@ struct ggml_backend_meta_split_state llama_meta_device_get_split_state(const str const int64_t ne_s = segments[is].first; const uint32_t nr_s = segments[is].second; const int64_t g_s = granularity[is]; - GGML_ASSERT(ne_full % g_s == 0); int64_t low = 0; size_t j = 0; for (; j < ud->n_devices - 1; j++) { @@ -1167,7 +1175,7 @@ void llama_model_base::load_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EMBEDDING_LENGTH_OUT, hparams.n_embd_out_impl, false); ml.get_key(LLM_KV_ATTENTION_CAUSAL, hparams.causal_attn, false); ml.get_key(LLM_KV_POOLING_TYPE, hparams.pooling_type, false); - ml.get_key(LLM_KV_BLOCK_COUNT, hparams.n_layer); + ml.get_key(LLM_KV_BLOCK_COUNT, hparams.n_layer_all); ml.get_key(LLM_KV_EXPERT_COUNT, hparams.n_expert, false); ml.get_key(LLM_KV_EXPERT_USED_COUNT, hparams.n_expert_used, false); ml.get_key(LLM_KV_EXPERT_GROUP_COUNT, hparams.n_expert_groups, false); @@ -1222,13 +1230,13 @@ void llama_model_base::load_hparams(llama_model_loader & ml) { std::fill(hparams.swiglu_clamp_exp.begin(), hparams.swiglu_clamp_exp.end(), 0.0f); std::fill(hparams.swiglu_clamp_shexp.begin(), hparams.swiglu_clamp_shexp.end(), 0.0f); - ml.get_key_or_arr(LLM_KV_FEED_FORWARD_LENGTH, hparams.n_ff_arr, hparams.n_layer, false); - ml.get_key_or_arr(LLM_KV_ATTENTION_HEAD_COUNT, hparams.n_head_arr, hparams.n_layer, false); + ml.get_key_or_arr(LLM_KV_FEED_FORWARD_LENGTH, hparams.n_ff_arr, hparams.n_layer(), false); + ml.get_key_or_arr(LLM_KV_ATTENTION_HEAD_COUNT, hparams.n_head_arr, hparams.n_layer(), false); // n_head_kv is optional, default to n_head hparams.n_head_kv_arr = hparams.n_head_arr; - ml.get_key_or_arr(LLM_KV_ATTENTION_HEAD_COUNT_KV, hparams.n_head_kv_arr, hparams.n_layer, false); + ml.get_key_or_arr(LLM_KV_ATTENTION_HEAD_COUNT_KV, hparams.n_head_kv_arr, hparams.n_layer(), false); bool rope_finetuned = false; ml.get_key(LLM_KV_ROPE_SCALING_FINETUNED, rope_finetuned, false); @@ -1327,7 +1335,7 @@ bool llama_model_base::load_tensors(llama_model_loader & ml) { const auto & use_mlock = params.use_mlock; const auto & tensor_split = params.tensor_split; - const int n_layer = hparams.n_layer; + const int n_layer = hparams.n_layer_all; const int n_gpu_layers = this->n_gpu_layers(); bool use_mmap_buffer = true; @@ -1384,11 +1392,10 @@ bool llama_model_base::load_tensors(llama_model_loader & ml) { splits[i] /= split_sum; } - int i_gpu_start = std::max(int(hparams.n_layer) + 1 - n_gpu_layers, 0); - - const int act_gpu_layers = devices.empty() ? 0 : std::min(n_gpu_layers, int(n_layer) + 1); + const int i_gpu_start = std::max(n_layer + 1 - n_gpu_layers, 0); + const int act_gpu_layers = devices.empty() ? 0 : std::min(n_gpu_layers, n_layer + 1); auto get_layer_buft_list = [&](int il) -> llama_model::impl::layer_dev { - const bool is_swa = il < int(hparams.n_layer) && hparams.is_swa(il); + const bool is_swa = il < n_layer && hparams.is_swa(il); if (il < i_gpu_start || (il - i_gpu_start) >= act_gpu_layers) { // LLAMA_LOG_DEBUG("load_tensors: layer %3d assigned to device %s, is_swa = %d\n", il, ggml_backend_dev_name(cpu_dev), is_swa); return {cpu_dev, &pimpl->cpu_buft_list}; @@ -1691,7 +1698,7 @@ bool llama_model_base::load_tensors(llama_model_loader & ml) { } if (llama_supports_gpu_offload()) { - const int n_gpu = std::min(n_gpu_layers, int(hparams.n_layer)); + const int n_gpu = std::min(n_gpu_layers, n_layer); int n_repeating = n_gpu; if (n_repeating > 0) { @@ -1700,8 +1707,8 @@ bool llama_model_base::load_tensors(llama_model_loader & ml) { } LLAMA_LOG_INFO("%s: offloading %d repeating layers to GPU\n", __func__, n_repeating); - const int max_backend_supported_layers = hparams.n_layer + 1; - const int max_offloadable_layers = hparams.n_layer + 1; + const int max_backend_supported_layers = n_layer + 1; + const int max_offloadable_layers = n_layer + 1; LLAMA_LOG_INFO("%s: offloaded %d/%d layers to GPU\n", __func__, std::min(n_gpu_layers, max_offloadable_layers), max_backend_supported_layers); } @@ -1770,7 +1777,8 @@ const float * llama_model::tensor_split() const { } uint32_t llama_model::n_gpu_layers() const { - return params.n_gpu_layers >= 0 ? params.n_gpu_layers : hparams.n_layer + 1; + // note: plus 1 for the "output" layer + return params.n_gpu_layers >= 0 ? params.n_gpu_layers : hparams.n_layer_all + 1; } llama_split_mode llama_model::split_mode() const { @@ -1841,17 +1849,17 @@ void llama_model::print_info() const { LLAMA_LOG_INFO("%s: n_ctx_train = %u\n", __func__, hparams.n_ctx_train); LLAMA_LOG_INFO("%s: n_embd = %u\n", __func__, hparams.n_embd); LLAMA_LOG_INFO("%s: n_embd_inp = %u\n", __func__, hparams.n_embd_inp()); - LLAMA_LOG_INFO("%s: n_layer = %u\n", __func__, hparams.n_layer); - LLAMA_LOG_INFO("%s: n_head = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_head(il); }, hparams.n_layer).c_str()); - LLAMA_LOG_INFO("%s: n_head_kv = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_head_kv(il); }, hparams.n_layer).c_str()); + LLAMA_LOG_INFO("%s: n_layer = %u\n", __func__, hparams.n_layer()); + LLAMA_LOG_INFO("%s: n_head = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_head(il); }, hparams.n_layer()).c_str()); + LLAMA_LOG_INFO("%s: n_head_kv = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_head_kv(il); }, hparams.n_layer()).c_str()); LLAMA_LOG_INFO("%s: n_rot = %u\n", __func__, hparams.n_rot_full); LLAMA_LOG_INFO("%s: n_swa = %u\n", __func__, hparams.n_swa); LLAMA_LOG_INFO("%s: is_swa_any = %u\n", __func__, hparams.is_swa_any()); LLAMA_LOG_INFO("%s: n_embd_head_k = %u\n", __func__, hparams.n_embd_head_k_full); LLAMA_LOG_INFO("%s: n_embd_head_v = %u\n", __func__, hparams.n_embd_head_v_full); - LLAMA_LOG_INFO("%s: n_gqa = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_gqa(il); }, hparams.n_layer).c_str()); - LLAMA_LOG_INFO("%s: n_embd_k_gqa = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_embd_k_gqa(il); }, hparams.n_layer).c_str()); - LLAMA_LOG_INFO("%s: n_embd_v_gqa = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_embd_v_gqa(il); }, hparams.n_layer).c_str()); + LLAMA_LOG_INFO("%s: n_gqa = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_gqa(il); }, hparams.n_layer()).c_str()); + LLAMA_LOG_INFO("%s: n_embd_k_gqa = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_embd_k_gqa(il); }, hparams.n_layer()).c_str()); + LLAMA_LOG_INFO("%s: n_embd_v_gqa = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_embd_v_gqa(il); }, hparams.n_layer()).c_str()); LLAMA_LOG_INFO("%s: f_norm_eps = %.1e\n", __func__, hparams.f_norm_eps); LLAMA_LOG_INFO("%s: f_norm_rms_eps = %.1e\n", __func__, hparams.f_norm_rms_eps); LLAMA_LOG_INFO("%s: f_clamp_kqv = %.1e\n", __func__, hparams.f_clamp_kqv); @@ -1859,7 +1867,7 @@ void llama_model::print_info() const { LLAMA_LOG_INFO("%s: f_logit_scale = %.1e\n", __func__, hparams.f_logit_scale); LLAMA_LOG_INFO("%s: f_attn_scale = %.1e\n", __func__, hparams.f_attention_scale); LLAMA_LOG_INFO("%s: f_attn_value_scale = %.4f\n", __func__, hparams.f_attn_value_scale); - LLAMA_LOG_INFO("%s: n_ff = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_ff(il); }, hparams.n_layer).c_str()); + LLAMA_LOG_INFO("%s: n_ff = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_ff(il); }, hparams.n_layer()).c_str()); LLAMA_LOG_INFO("%s: n_expert = %u\n", __func__, hparams.n_expert); LLAMA_LOG_INFO("%s: n_expert_used = %u\n", __func__, hparams.n_expert_used); LLAMA_LOG_INFO("%s: n_expert_groups = %d\n", __func__, hparams.n_expert_groups); @@ -1986,7 +1994,7 @@ void llama_model::print_info() const { LLAMA_LOG_INFO("%s: expert_weights_scale = %.1f\n", __func__, hparams.expert_weights_scale); LLAMA_LOG_INFO("%s: expert_weights_norm = %d\n", __func__, hparams.expert_weights_norm); LLAMA_LOG_INFO("%s: expert_gating_func = %s\n", __func__, llama_expert_gating_func_name((llama_expert_gating_func_type) hparams.expert_gating_func)); - LLAMA_LOG_INFO("%s: nextn_predict_layers = %d\n", __func__, hparams.nextn_predict_layers); + LLAMA_LOG_INFO("%s: n_layer_nextn = %d\n", __func__, hparams.n_layer_nextn); } if (arch == LLM_ARCH_SMALLTHINKER || arch == LLM_ARCH_LFM2MOE) { @@ -2168,22 +2176,21 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params, llama_memory_hybrid::layer_filter_cb filter_attn = nullptr; llama_memory_hybrid::layer_filter_cb filter_recr = nullptr; if (arch == LLM_ARCH_FALCON_H1) { - filter_attn = [&](int32_t) { return true; }; - filter_recr = [&](int32_t) { return true; }; + filter_attn = [&](uint32_t) { return true; }; + filter_recr = [&](uint32_t) { return true; }; } else if (arch == LLM_ARCH_NEMOTRON_H || arch == LLM_ARCH_NEMOTRON_H_MOE) { - filter_attn = [&](int32_t il) { + filter_attn = [&](uint32_t il) { return !hparams.is_recr(il) && hparams.n_ff(il) == 0; }; - filter_recr = [&](int32_t il) { + filter_recr = [&](uint32_t il) { return hparams.is_recr(il) && hparams.n_ff(il) == 0; }; } else if (arch == LLM_ARCH_QWEN35 || arch == LLM_ARCH_QWEN35MOE) { - const uint32_t n_main = hparams.n_layer - hparams.nextn_predict_layers; - filter_attn = [&, n_main](int32_t il) { - return (uint32_t)il < n_main && !hparams.is_recr(il); + filter_attn = [&](uint32_t il) { + return il < hparams.n_layer() && !hparams.is_recr(il); }; - filter_recr = [&, n_main](int32_t il) { - return (uint32_t)il < n_main && hparams.is_recr(il); + filter_recr = [&](uint32_t il) { + return il < hparams.n_layer() && hparams.is_recr(il); }; } @@ -2232,9 +2239,11 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params, llama_kv_cache::layer_filter_cb filter = nullptr; if (arch == LLM_ARCH_GEMMA3N || arch == LLM_ARCH_GEMMA4) { - reuse = [&](int32_t il) { - if (il >= (int32_t) hparams.n_layer_kv_from_start) { - return (int32_t) hparams.n_layer_kv_from_start - (hparams.is_swa(il) ? 2 : 1); + reuse = [&](uint32_t il) { + GGML_ASSERT(hparams.n_layer_kv_from_start >= 2); + + if (il >= (uint32_t)hparams.n_layer_kv_from_start) { + return hparams.n_layer_kv_from_start - (hparams.is_swa(il) ? 2 : 1); } return -1; @@ -2242,16 +2251,14 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params, } if (mtp_on_hybrid_qwen35) { - const uint32_t n_main = hparams.n_layer - hparams.nextn_predict_layers; - filter = [n_main](int32_t il) { return (uint32_t)il >= n_main; }; + filter = [&](uint32_t il) { return il >= hparams.n_layer(); }; } - if (arch == LLM_ARCH_STEP35 && hparams.nextn_predict_layers > 0) { - const uint32_t n_main = hparams.n_layer - hparams.nextn_predict_layers; + if (arch == LLM_ARCH_STEP35 && hparams.n_layer_nextn > 0) { if (params.ctx_type == LLAMA_CONTEXT_TYPE_MTP) { - filter = [n_main](int32_t il) { return (uint32_t)il >= n_main; }; + filter = [&](uint32_t il) { return il >= hparams.n_layer(); }; } else { - filter = [n_main](int32_t il) { return (uint32_t)il < n_main; }; + filter = [&](uint32_t il) { return il < hparams.n_layer(); }; } } @@ -2376,7 +2383,7 @@ int32_t llama_model_n_embd_out(const llama_model * model) { } int32_t llama_model_n_layer(const llama_model * model) { - return model->hparams.n_layer; + return model->hparams.n_layer(); } int32_t llama_model_n_head(const llama_model * model) { diff --git a/src/llama-model.h b/src/llama-model.h index a561374ed..884cfdf5c 100644 --- a/src/llama-model.h +++ b/src/llama-model.h @@ -700,7 +700,8 @@ const char * llm_type_name(llm_type type); // convenience macro for loading local variables for load_tensors() in llama_model_base // note: cast to int64_t since we will use these for the tensor dimensions #define LLAMA_LOAD_LOCALS \ - const int n_layer = hparams.n_layer; GGML_UNUSED(n_layer); \ + const int n_layer = hparams.n_layer(); GGML_UNUSED(n_layer); \ + const int n_layer_all = hparams.n_layer_all; GGML_UNUSED(n_layer_all); \ const int64_t n_head = hparams.n_head(); GGML_UNUSED(n_head); \ const int64_t n_head_kv = hparams.n_head_kv(); GGML_UNUSED(n_head_kv); \ const int64_t n_embd = hparams.n_embd; GGML_UNUSED(n_embd); \ diff --git a/src/llama-quant.cpp b/src/llama-quant.cpp index 46629e71c..74de4f3a0 100644 --- a/src/llama-quant.cpp +++ b/src/llama-quant.cpp @@ -849,7 +849,7 @@ static void init_quantize_state_counters(quantize_state_impl & qs, std::vectorhparams.n_embd = desc->n_embd; model->hparams.n_embd_head_k_full = desc->n_embd_head_k; model->hparams.n_embd_head_v_full = desc->n_embd_head_v; - model->hparams.n_layer = desc->n_layer; + model->hparams.n_layer_all = desc->n_layer; model->hparams.n_expert = desc->n_expert; for (uint32_t i = 0; i < desc->n_layer; i++) { diff --git a/src/models/afmoe.cpp b/src/models/afmoe.cpp index a7c77ee5d..063b21425 100644 --- a/src/models/afmoe.cpp +++ b/src/models/afmoe.cpp @@ -30,7 +30,7 @@ void llama_model_afmoe::load_arch_hparams(llama_model_loader & ml) { hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID; } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 56: type = LLM_TYPE_6B; break; case 32: type = LLM_TYPE_26B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/apertus.cpp b/src/models/apertus.cpp index bec713652..6dfb8905f 100644 --- a/src/models/apertus.cpp +++ b/src/models/apertus.cpp @@ -2,12 +2,13 @@ void llama_model_apertus::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - ml.get_key_or_arr(LLM_KV_XIELU_ALPHA_N, hparams.xielu_alpha_n, hparams.n_layer); - ml.get_key_or_arr(LLM_KV_XIELU_ALPHA_P, hparams.xielu_alpha_p, hparams.n_layer); - ml.get_key_or_arr(LLM_KV_XIELU_BETA, hparams.xielu_beta, hparams.n_layer); - ml.get_key_or_arr(LLM_KV_XIELU_EPS, hparams.xielu_eps, hparams.n_layer); - switch (hparams.n_layer) { + ml.get_key_or_arr(LLM_KV_XIELU_ALPHA_N, hparams.xielu_alpha_n, hparams.n_layer()); + ml.get_key_or_arr(LLM_KV_XIELU_ALPHA_P, hparams.xielu_alpha_p, hparams.n_layer()); + ml.get_key_or_arr(LLM_KV_XIELU_BETA, hparams.xielu_beta, hparams.n_layer()); + ml.get_key_or_arr(LLM_KV_XIELU_EPS, hparams.xielu_eps, hparams.n_layer()); + + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_8B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/arcee.cpp b/src/models/arcee.cpp index d086c4717..9536e7c5d 100644 --- a/src/models/arcee.cpp +++ b/src/models/arcee.cpp @@ -4,7 +4,7 @@ void llama_model_arcee::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); // Arcee uses the same structure as Llama - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 36: type = LLM_TYPE_4B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/arctic.cpp b/src/models/arctic.cpp index 27deadffe..09ee0f752 100644 --- a/src/models/arctic.cpp +++ b/src/models/arctic.cpp @@ -4,7 +4,7 @@ void llama_model_arctic::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); if (hparams.n_expert == 128) { - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 35: type = LLM_TYPE_10B_128x3_66B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/arwkv7.cpp b/src/models/arwkv7.cpp index 9bd04127b..b38b20647 100644 --- a/src/models/arwkv7.cpp +++ b/src/models/arwkv7.cpp @@ -10,7 +10,7 @@ void llama_model_arwkv7::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_GATE_LORA_RANK, hparams.n_lora_gate, false); ml.get_key(LLM_KV_TOKEN_SHIFT_COUNT, hparams.token_shift_count, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 12: switch (hparams.n_embd) { case 768: type = LLM_TYPE_190M; break; diff --git a/src/models/baichuan.cpp b/src/models/baichuan.cpp index 4d26081cd..585f36141 100644 --- a/src/models/baichuan.cpp +++ b/src/models/baichuan.cpp @@ -2,7 +2,7 @@ void llama_model_baichuan::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_7B; break; case 40: type = LLM_TYPE_13B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/bailingmoe.cpp b/src/models/bailingmoe.cpp index fe1ae1086..7faf73c83 100644 --- a/src/models/bailingmoe.cpp +++ b/src/models/bailingmoe.cpp @@ -8,7 +8,7 @@ void llama_model_bailingmoe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale, false); ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM, hparams.expert_weights_norm, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 28: type = LLM_TYPE_16B; break; case 88: type = LLM_TYPE_290B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/bailingmoe2.cpp b/src/models/bailingmoe2.cpp index 2f0d44a62..5000e9c6d 100644 --- a/src/models/bailingmoe2.cpp +++ b/src/models/bailingmoe2.cpp @@ -9,17 +9,13 @@ void llama_model_bailingmoe2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale, false); ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM, hparams.expert_weights_norm, false); ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func); - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); - // TODO: when MTP is implemented, this should probably be updated if needed - hparams.n_layer_kv_from_start = hparams.n_layer - hparams.nextn_predict_layers; + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl"); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 20: type = LLM_TYPE_16B_A1B; break; - case 21: type = LLM_TYPE_16B_A1B; break; case 32: type = LLM_TYPE_100B_A6B; break; - case 33: type = LLM_TYPE_100B_A6B; break; default: type = LLM_TYPE_UNKNOWN; } } @@ -39,9 +35,9 @@ void llama_model_bailingmoe2::load_arch_tensors(llama_model_loader &) { GGML_ASSERT(n_expert > 0 && "n_expert must be > 0 for bailingmoe2"); GGML_ASSERT(n_expert_used > 0 && "n_expert_used must be > 0 for bailingmoe2"); - for (int i = 0; i < n_layer; ++i) { + for (int i = 0; i < n_layer_all; ++i) { int flags = 0; - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { // skip all tensors in the NextN layers flags |= TENSOR_SKIP; } @@ -78,7 +74,7 @@ void llama_model_bailingmoe2::load_arch_tensors(llama_model_loader &) { } // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { layer.nextn.eh_proj = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags); layer.nextn.embed_tokens = create_tensor(tn(LLM_TENSOR_NEXTN_EMBED_TOKENS, "weight", i), { n_embd, n_vocab }, TENSOR_NOT_REQUIRED | flags); layer.nextn.enorm = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags); @@ -112,8 +108,7 @@ llama_model_bailingmoe2::graph::graph(const llama_model & model, const llm_graph ggml_tensor * inp_out_ids = build_inp_out_ids(); - const int n_transformer_layers = n_layer - hparams.nextn_predict_layers; - for (int il = 0; il < n_transformer_layers; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; // norm @@ -146,7 +141,7 @@ llama_model_bailingmoe2::graph::graph(const llama_model & model, const llm_graph Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il); } - if (il == n_transformer_layers - 1 && inp_out_ids) { + if (il == n_layer - 1 && inp_out_ids) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } diff --git a/src/models/bert.cpp b/src/models/bert.cpp index 3c28f419c..53ce29f23 100644 --- a/src/models/bert.cpp +++ b/src/models/bert.cpp @@ -1,9 +1,9 @@ #include "models.h" void llama_model_bert::load_arch_hparams(llama_model_loader & ml) { - ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 3: type = LLM_TYPE_17M; break; // bge-micro case 6: diff --git a/src/models/bitnet.cpp b/src/models/bitnet.cpp index 7e8125dee..c83302745 100644 --- a/src/models/bitnet.cpp +++ b/src/models/bitnet.cpp @@ -3,7 +3,7 @@ void llama_model_bitnet::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 26: type = LLM_TYPE_3B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/bloom.cpp b/src/models/bloom.cpp index 30b0f3d07..609d2ddf9 100644 --- a/src/models/bloom.cpp +++ b/src/models/bloom.cpp @@ -3,7 +3,7 @@ void llama_model_bloom::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_1B; break; case 30: switch (hparams.n_embd) { diff --git a/src/models/chameleon.cpp b/src/models/chameleon.cpp index 4bceaefd6..4f45acecf 100644 --- a/src/models/chameleon.cpp +++ b/src/models/chameleon.cpp @@ -6,7 +6,7 @@ void llama_model_chameleon::load_arch_hparams(llama_model_loader & ml) { hparams.f_norm_eps = 1e-5; // eps for qk-norm, torch default ml.get_key(LLM_KV_SWIN_NORM, hparams.swin_norm, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_7B; break; case 48: type = LLM_TYPE_34B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/chatglm.cpp b/src/models/chatglm.cpp index 6766fa71c..7ae5b938f 100644 --- a/src/models/chatglm.cpp +++ b/src/models/chatglm.cpp @@ -2,7 +2,8 @@ void llama_model_chatglm::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 28: { if (hparams.n_head(0) == 16) { type = LLM_TYPE_1_5B; diff --git a/src/models/codeshell.cpp b/src/models/codeshell.cpp index 274dd3342..de53bb981 100644 --- a/src/models/codeshell.cpp +++ b/src/models/codeshell.cpp @@ -2,7 +2,8 @@ void llama_model_codeshell::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 42: type = LLM_TYPE_7B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/cogvlm.cpp b/src/models/cogvlm.cpp index 2e231bb3f..750f57a39 100644 --- a/src/models/cogvlm.cpp +++ b/src/models/cogvlm.cpp @@ -2,7 +2,8 @@ void llama_model_cogvlm::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_13B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/cohere2.cpp b/src/models/cohere2.cpp index a514cf88f..61a5945a1 100644 --- a/src/models/cohere2.cpp +++ b/src/models/cohere2.cpp @@ -5,6 +5,7 @@ void llama_model_cohere2::load_arch_hparams(llama_model_loader & ml) { uint32_t swa_period = 4; ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, swa_period, false); hparams.set_swa_pattern(swa_period); + hparams.rope_freq_base_train_swa = hparams.rope_freq_base_train; hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train; @@ -12,7 +13,8 @@ void llama_model_cohere2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa); ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_8B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/command-r.cpp b/src/models/command-r.cpp index adf7fcaa2..94a46188b 100644 --- a/src/models/command-r.cpp +++ b/src/models/command-r.cpp @@ -3,7 +3,8 @@ void llama_model_command_r::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale, false); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 40: type = LLM_TYPE_35B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/dbrx.cpp b/src/models/dbrx.cpp index af71c7753..4f5ac4d06 100644 --- a/src/models/dbrx.cpp +++ b/src/models/dbrx.cpp @@ -1,14 +1,14 @@ #include "models.h" void llama_model_dbrx::load_arch_hparams(llama_model_loader & ml) { -ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); -ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV, hparams.f_clamp_kqv); + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); + ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV, hparams.f_clamp_kqv); -switch (hparams.n_layer) { - case 40: type = LLM_TYPE_16x12B; break; - default: type = LLM_TYPE_UNKNOWN; + switch (hparams.n_layer()) { + case 40: type = LLM_TYPE_16x12B; break; + default: type = LLM_TYPE_UNKNOWN; + } } - } void llama_model_dbrx::load_arch_tensors(llama_model_loader &) { LLAMA_LOAD_LOCALS; diff --git a/src/models/deci.cpp b/src/models/deci.cpp index 567e35352..cdfcf29e0 100644 --- a/src/models/deci.cpp +++ b/src/models/deci.cpp @@ -2,7 +2,8 @@ void llama_model_deci::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_7B; break; case 80: type = LLM_TYPE_70B; break; case 162: type = LLM_TYPE_405B; break; diff --git a/src/models/deepseek2.cpp b/src/models/deepseek2.cpp index 1fe54adc1..a9e8bc514 100644 --- a/src/models/deepseek2.cpp +++ b/src/models/deepseek2.cpp @@ -5,7 +5,7 @@ void llama_model_deepseek2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_VOCAB_SIZE, n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, n_vocab, false); // lite variants include DeepSeek-V2-Lite, GigaChat3-10B-A1.8B, Kanana-2-30B-A3B - const bool is_lite = (hparams.n_layer == 27 || hparams.n_layer == 26 || (hparams.n_layer == 48 && n_vocab == 128256)); + const bool is_lite = (hparams.n_layer() == 27 || hparams.n_layer() == 26 || (hparams.n_layer() == 48 && n_vocab == 128256)); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead, false); @@ -23,7 +23,7 @@ void llama_model_deepseek2::load_arch_hparams(llama_model_loader & ml) { if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) { // for compatibility with existing DeepSeek V2 and V2.5 GGUFs // that have no expert_gating_func model parameter set - if ((hparams.n_layer == 47 || hparams.n_layer == 48) && n_vocab == 154880) { + if ((hparams.n_layer() == 47 || hparams.n_layer() == 48) && n_vocab == 154880) { // GLM 4.7 Lite hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID; } else { @@ -43,7 +43,7 @@ void llama_model_deepseek2::load_arch_hparams(llama_model_loader & ml) { hparams.f_attn_temp_offset = 0.0f; - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 27: type = LLM_TYPE_16B; break; case 47: type = LLM_TYPE_30B_A3B; break; case 60: type = LLM_TYPE_236B; break; @@ -191,8 +191,7 @@ llama_model_deepseek2::graph::graph(const llama_model & model, const llm_graph_p ggml_tensor * inp_out_ids = build_inp_out_ids(); - int effective_n_layers = hparams.n_layer - hparams.nextn_predict_layers; - for (int il = 0; il < effective_n_layers; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; // norm @@ -366,7 +365,7 @@ llama_model_deepseek2::graph::graph(const llama_model & model, const llm_graph_p Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il); } } - if (il == effective_n_layers - 1 && inp_out_ids) { + if (il == n_layer - 1 && inp_out_ids) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } diff --git a/src/models/deepseek2ocr.cpp b/src/models/deepseek2ocr.cpp index f9e4c9878..65d31c31b 100644 --- a/src/models/deepseek2ocr.cpp +++ b/src/models/deepseek2ocr.cpp @@ -14,7 +14,7 @@ void llama_model_deepseek2ocr::load_arch_hparams(llama_model_loader & ml) { hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX; } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 12: type = LLM_TYPE_3B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/deepseek32.cpp b/src/models/deepseek32.cpp index c92ab60d1..9a20e2ce9 100644 --- a/src/models/deepseek32.cpp +++ b/src/models/deepseek32.cpp @@ -31,7 +31,7 @@ void llama_model_deepseek32::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_INDEXER_TOP_K, hparams.indexer_top_k); // Expert gating function - ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func); + ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func); if (ml.get_key(LLM_KV_ROPE_SCALING_YARN_LOG_MUL, hparams.rope_yarn_log_mul, 0.0f)) { // [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX] @@ -40,13 +40,10 @@ void llama_model_deepseek32::load_arch_hparams(llama_model_loader & ml) { } // NextN/MTP parameters - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer"); - // TODO: when MTP is implemented, this should probably be updated if needed - hparams.n_layer_kv_from_start = hparams.n_layer - hparams.nextn_predict_layers; - - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 62: type = LLM_TYPE_685B_A37B; break; default: type = LLM_TYPE_UNKNOWN; } @@ -82,9 +79,9 @@ void llama_model_deepseek32::load_arch_tensors(llama_model_loader &) { output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); } - for (int i = 0; i < n_layer; ++i) { + for (int i = 0; i < n_layer_all; ++i) { int flags = 0; - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { // skip all tensors in the NextN layers // TODO @ngxson : TENSOR_NOT_REQUIRED was a hack, need to remove it later flags |= TENSOR_SKIP | TENSOR_NOT_REQUIRED; @@ -142,7 +139,7 @@ void llama_model_deepseek32::load_arch_tensors(llama_model_loader &) { } // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { layer.nextn.eh_proj = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags); layer.nextn.enorm = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags); layer.nextn.hnorm = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), { n_embd }, flags); @@ -205,8 +202,7 @@ llama_model_deepseek32::graph::graph(const llama_model & model, const llm_graph_ ggml_tensor * inp_out_ids = build_inp_out_ids(); - int effective_n_layers = hparams.n_layer - hparams.nextn_predict_layers; - for (int il = 0; il < effective_n_layers; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; // norm @@ -427,7 +423,7 @@ llama_model_deepseek32::graph::graph(const llama_model & model, const llm_graph_ Qcur, Kcur, Vcur, nullptr, nullptr, model.layers[il].wv_b, top_k, kq_scale, il); } } - if (il == effective_n_layers - 1 && inp_out_ids) { + if (il == n_layer - 1 && inp_out_ids) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } diff --git a/src/models/dots1.cpp b/src/models/dots1.cpp index 435d27281..07d6ab1b7 100644 --- a/src/models/dots1.cpp +++ b/src/models/dots1.cpp @@ -8,7 +8,8 @@ void llama_model_dots1::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale, false); ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM, hparams.expert_weights_norm, false); ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func, false); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 62: type = LLM_TYPE_142B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/dream.cpp b/src/models/dream.cpp index 12ac6f1ce..abe737c33 100644 --- a/src/models/dream.cpp +++ b/src/models/dream.cpp @@ -2,8 +2,9 @@ void llama_model_dream::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); + // Dream models are primarily 7B with 28 layers - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 28: type = LLM_TYPE_7B; break; diff --git a/src/models/ernie4-5.cpp b/src/models/ernie4-5.cpp index 9b39c605e..895cf690b 100644 --- a/src/models/ernie4-5.cpp +++ b/src/models/ernie4-5.cpp @@ -12,7 +12,7 @@ void llama_model_ernie4_5::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead, false); } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 18: type = LLM_TYPE_0_3B; break; case 28: type = LLM_TYPE_21B_A3B; break; case 54: type = LLM_TYPE_300B_A47B; break; diff --git a/src/models/eurobert.cpp b/src/models/eurobert.cpp index ddf13c302..0948d7de6 100644 --- a/src/models/eurobert.cpp +++ b/src/models/eurobert.cpp @@ -3,7 +3,7 @@ void llama_model_eurobert::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - if (hparams.n_layer == 12) { + if (hparams.n_layer() == 12) { type = LLM_TYPE_SMALL; // 0.2B } } diff --git a/src/models/exaone-moe.cpp b/src/models/exaone-moe.cpp index 76d91982f..5aed93794 100644 --- a/src/models/exaone-moe.cpp +++ b/src/models/exaone-moe.cpp @@ -20,13 +20,12 @@ void llama_model_exaone_moe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM, hparams.expert_weights_norm, false); ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead, false); - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl"); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_30B_A3B; break; - case 48: - case 49: type = LLM_TYPE_235B_A22B; break; + case 48: type = LLM_TYPE_235B_A22B; break; default: type = LLM_TYPE_UNKNOWN; } } @@ -50,9 +49,9 @@ void llama_model_exaone_moe::load_arch_tensors(llama_model_loader &) { output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); } - for (int i = 0; i < n_layer; ++i) { + for (int i = 0; i < n_layer_all; ++i) { int flags = 0; - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { // skip all tensors in the NextN layers flags |= TENSOR_SKIP; } @@ -70,7 +69,7 @@ void llama_model_exaone_moe::load_arch_tensors(llama_model_loader &) { layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, flags); // dense layers for first n_layer_dense_lead layers or nextn_predict_layers layers at the end - if (i < (int) hparams.n_layer_dense_lead || (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers)) { + if (i < (int) hparams.n_layer_dense_lead || (i >= n_layer)) { layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, flags); layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, flags); layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, flags); @@ -95,7 +94,7 @@ void llama_model_exaone_moe::load_arch_tensors(llama_model_loader &) { } // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { layer.nextn.eh_proj = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), {2 * n_embd, n_embd}, flags); layer.nextn.enorm = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), {n_embd}, flags); layer.nextn.hnorm = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), {n_embd}, flags); @@ -130,8 +129,7 @@ llama_model_exaone_moe::graph::graph(const llama_model & model, const llm_graph_ ggml_tensor * inp_out_ids = build_inp_out_ids(); - const int n_transformer_layers = n_layer - hparams.nextn_predict_layers; - for (int il = 0; il < n_transformer_layers; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; // use RoPE for SWA layers @@ -170,7 +168,7 @@ llama_model_exaone_moe::graph::graph(const llama_model & model, const llm_graph_ Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il); cb(cur, "attn_out", il); } - if (il == n_transformer_layers - 1 && inp_out_ids) { + if (il == n_layer - 1 && inp_out_ids) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } diff --git a/src/models/exaone.cpp b/src/models/exaone.cpp index c7e9960d7..676fb37b5 100644 --- a/src/models/exaone.cpp +++ b/src/models/exaone.cpp @@ -3,7 +3,7 @@ void llama_model_exaone::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_8B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/exaone4.cpp b/src/models/exaone4.cpp index b5030eb05..863268abc 100644 --- a/src/models/exaone4.cpp +++ b/src/models/exaone4.cpp @@ -1,7 +1,7 @@ #include "models.h" void llama_model_exaone4::load_arch_hparams(llama_model_loader & ml) { - if (hparams.n_layer == 64) { // 32B + if (hparams.n_layer() == 64) { // 32B hparams.swa_type = LLAMA_SWA_TYPE_STANDARD; hparams.n_swa = 4096; uint32_t swa_period = 4; @@ -15,11 +15,11 @@ void llama_model_exaone4::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); - hparams.n_layer_kv_from_start = hparams.n_layer - hparams.nextn_predict_layers; + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); - switch (hparams.n_layer) { + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer"); + + switch (hparams.n_layer()) { case 30: type = LLM_TYPE_1_2B; break; case 64: type = LLM_TYPE_32B; break; default: type = LLM_TYPE_UNKNOWN; @@ -40,8 +40,8 @@ void llama_model_exaone4::load_arch_tensors(llama_model_loader &) { output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); } - for (int i = 0; i < n_layer; ++i) { - const bool is_nextn = hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers; + for (int i = 0; i < n_layer_all; ++i) { + const bool is_nextn = i >= n_layer; int flags = 0; if (is_nextn) { // NextN/MTP layers are preserved in GGUF but are not executed yet. @@ -109,11 +109,7 @@ llama_model_exaone4::graph::graph(const llama_model & model, const llm_gra } ggml_tensor * inp_out_ids = build_inp_out_ids(); - // MTP / NextN tail blocks are loaded for compatibility but not executed (same as exaone-moe). - const int n_layer_main = int(n_layer) - int(hparams.nextn_predict_layers); - GGML_ASSERT(n_layer_main > 0); - - for (int il = 0; il < n_layer_main; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; // use RoPE for SWA layers or non-SWA models @@ -149,7 +145,7 @@ llama_model_exaone4::graph::graph(const llama_model & model, const llm_gra Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il); cb(cur, "attn_out", il); } - if (il == n_layer_main - 1 && inp_out_ids) { + if (il == n_layer - 1 && inp_out_ids) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } diff --git a/src/models/falcon-h1.cpp b/src/models/falcon-h1.cpp index c130ccdd4..d6ef2d519 100644 --- a/src/models/falcon-h1.cpp +++ b/src/models/falcon-h1.cpp @@ -13,7 +13,7 @@ void llama_model_falcon_h1::load_arch_hparams(llama_model_loader & ml) { std::fill(hparams.is_recr_impl.begin(), hparams.is_recr_impl.end(), true); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 36: type = LLM_TYPE_0_5B; break; case 24: diff --git a/src/models/falcon.cpp b/src/models/falcon.cpp index ad546ef2d..b2ad90b32 100644 --- a/src/models/falcon.cpp +++ b/src/models/falcon.cpp @@ -3,7 +3,7 @@ void llama_model_falcon::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_7B; break; case 60: type = LLM_TYPE_40B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/gemma-embedding.cpp b/src/models/gemma-embedding.cpp index 4e07f5f2b..80ed3b1a4 100644 --- a/src/models/gemma-embedding.cpp +++ b/src/models/gemma-embedding.cpp @@ -21,7 +21,7 @@ void llama_model_gemma_embedding::load_arch_hparams(llama_model_loader & ml) { GGML_ASSERT((hparams.dense_2_feat_in == 0 || hparams.dense_2_feat_in == hparams.n_embd) && "dense_2_feat_in must be equal to n_embd"); GGML_ASSERT((hparams.dense_3_feat_out == 0 || hparams.dense_3_feat_out == hparams.n_embd) && "dense_3_feat_out must be equal to n_embd"); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_0_3B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/gemma.cpp b/src/models/gemma.cpp index 1519682fd..651cd7e64 100644 --- a/src/models/gemma.cpp +++ b/src/models/gemma.cpp @@ -3,7 +3,7 @@ void llama_model_gemma::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 18: type = LLM_TYPE_2B; break; case 28: type = LLM_TYPE_7B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/gemma2.cpp b/src/models/gemma2.cpp index ae3f9ffb5..2fbfb15a9 100644 --- a/src/models/gemma2.cpp +++ b/src/models/gemma2.cpp @@ -16,7 +16,7 @@ void llama_model_gemma2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTN_LOGIT_SOFTCAPPING, hparams.f_attn_logit_softcapping, false); ml.get_key(LLM_KV_FINAL_LOGIT_SOFTCAPPING, hparams.f_final_logit_softcapping, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 26: type = LLM_TYPE_2B; break; case 42: type = LLM_TYPE_9B; break; case 46: type = LLM_TYPE_27B; break; diff --git a/src/models/gemma3.cpp b/src/models/gemma3.cpp index 63a2b380e..690194529 100644 --- a/src/models/gemma3.cpp +++ b/src/models/gemma3.cpp @@ -17,7 +17,7 @@ void llama_model_gemma3::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_FINAL_LOGIT_SOFTCAPPING, hparams.f_final_logit_softcapping, false); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 18: type = LLM_TYPE_270M; break; case 26: type = LLM_TYPE_1B; break; case 32: type = LLM_TYPE_8B; break; // Rnj-1 diff --git a/src/models/gemma3n.cpp b/src/models/gemma3n.cpp index 6ec3a0060..83eb8250a 100644 --- a/src/models/gemma3n.cpp +++ b/src/models/gemma3n.cpp @@ -6,14 +6,14 @@ void llama_model_gemma3n::load_arch_hparams(llama_model_loader & ml) { hparams.swa_type = LLAMA_SWA_TYPE_STANDARD; hparams.set_swa_pattern(swa_period); - hparams.n_layer_kv_from_start = 20; - hparams.f_attention_scale = 1.0f; + hparams.n_layer_kv_from_start = 20; + hparams.f_attention_scale = 1.0f; ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false); ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 30: type = LLM_TYPE_E2B; break; case 35: type = LLM_TYPE_E4B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/gemma4.cpp b/src/models/gemma4.cpp index e21e5901a..e74e12f26 100644 --- a/src/models/gemma4.cpp +++ b/src/models/gemma4.cpp @@ -2,12 +2,12 @@ void llama_model_gemma4::load_arch_hparams(llama_model_loader & ml) { hparams.swa_type = LLAMA_SWA_TYPE_STANDARD; - ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer); + ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer()); uint32_t n_kv_shared_layers = 0; ml.get_key(LLM_KV_ATTENTION_SHARED_KV_LAYERS, n_kv_shared_layers, false); - hparams.n_layer_kv_from_start = hparams.n_layer - (int32_t)n_kv_shared_layers; + hparams.n_layer_kv_from_start = hparams.n_layer_all - (int32_t)n_kv_shared_layers; hparams.f_attention_scale = 1.0f; // Gemma4 uses self.scaling = 1.0 (no pre-attn scaling) ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false); @@ -19,7 +19,7 @@ void llama_model_gemma4::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH_SWA, hparams.n_embd_head_v_swa); ml.get_key(LLM_KV_FINAL_LOGIT_SOFTCAPPING, hparams.f_final_logit_softcapping, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 30: type = LLM_TYPE_26B_A4B; break; case 35: type = LLM_TYPE_E2B; break; case 42: type = LLM_TYPE_E4B; break; diff --git a/src/models/glm-dsa.cpp b/src/models/glm-dsa.cpp index af2b55ef5..11d91312d 100644 --- a/src/models/glm-dsa.cpp +++ b/src/models/glm-dsa.cpp @@ -33,13 +33,10 @@ void llama_model_glm_dsa::load_arch_hparams(llama_model_loader & ml) { } // NextN/MTP parameters - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl"); - // TODO: when MTP is implemented, this should probably be updated if needed - hparams.n_layer_kv_from_start = hparams.n_layer - hparams.nextn_predict_layers; - - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 79: type = LLM_TYPE_744B_A40B; break; default: type = LLM_TYPE_UNKNOWN; } @@ -76,9 +73,9 @@ void llama_model_glm_dsa::load_arch_tensors(llama_model_loader &) { output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); } - for (int i = 0; i < n_layer; ++i) { + for (int i = 0; i < n_layer_all; ++i) { int flags = 0; - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { // skip all tensors in the NextN layers // TODO @ngxson : TENSOR_NOT_REQUIRED was a hack, need to remove it later flags |= TENSOR_SKIP | TENSOR_NOT_REQUIRED; @@ -135,8 +132,8 @@ void llama_model_glm_dsa::load_arch_tensors(llama_model_loader &) { layer.ffn_up_shexp = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, flags); } - // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + // NextN/MTP tensors (preserved but unused) - conditionally load for last n_layer_nextn + if (i >= n_layer) { layer.nextn.eh_proj = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags); layer.nextn.enorm = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags); layer.nextn.hnorm = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), { n_embd }, flags); diff --git a/src/models/glm4-moe.cpp b/src/models/glm4-moe.cpp index 27654b8cb..d60e47ddf 100644 --- a/src/models/glm4-moe.cpp +++ b/src/models/glm4-moe.cpp @@ -20,16 +20,13 @@ void llama_model_glm4_moe::load_arch_hparams(llama_model_loader & ml) { } // NextN/MTP parameters - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl"); - // TODO: when MTP is implemented, this should probably be updated if needed - hparams.n_layer_kv_from_start = hparams.n_layer - hparams.nextn_predict_layers; - - switch (hparams.n_layer) { - case 47: type = LLM_TYPE_106B_A12B; break; // GLM-4.5-Air (46 layers + 1 NextN layer) + switch (hparams.n_layer()) { + case 46: type = LLM_TYPE_106B_A12B; break; // GLM-4.5-Air case 48: type = LLM_TYPE_102B_A12B; break; // Solar Open - case 93: type = LLM_TYPE_355B_A32B; break; // GLM-4.5 (92 layers + 1 NextN layer) + case 92: type = LLM_TYPE_355B_A32B; break; // GLM-4.5 default: type = LLM_TYPE_UNKNOWN; } } @@ -54,9 +51,9 @@ void llama_model_glm4_moe::load_arch_tensors(llama_model_loader &) { // Load ALL tensors including NextN layer to satisfy total tensor count // but only PROCESS up to last layer (skipping final NextN layer) in forward pass - for (int i = 0; i < n_layer; ++i) { + for (int i = 0; i < n_layer_all; ++i) { int flags = 0; - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { // skip all tensors in the NextN layers flags |= TENSOR_SKIP; } @@ -116,7 +113,7 @@ void llama_model_glm4_moe::load_arch_tensors(llama_model_loader &) { } // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { layer.nextn.eh_proj = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags); layer.nextn.enorm = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags); layer.nextn.hnorm = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), { n_embd }, flags); @@ -161,8 +158,7 @@ llama_model_glm4_moe::graph::graph(const llama_model & model, const llm_graph_pa // Only process up to last layer (skip final NextN layer) // Final layer tensors are loaded but not processed in forward pass - const int n_transformer_layers = n_layer - hparams.nextn_predict_layers; - for (int il = 0; il < n_transformer_layers; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; // Pre-attention norm @@ -211,7 +207,7 @@ llama_model_glm4_moe::graph::graph(const llama_model & model, const llm_graph_pa model.layers[il].wo, NULL, model.layers[il].wo_s, Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il); } - if (il == n_transformer_layers - 1 && inp_out_ids) { + if (il == n_layer - 1 && inp_out_ids) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } diff --git a/src/models/glm4.cpp b/src/models/glm4.cpp index 7c242fed2..b4326c5f2 100644 --- a/src/models/glm4.cpp +++ b/src/models/glm4.cpp @@ -5,13 +5,10 @@ void llama_model_glm4::load_arch_hparams(llama_model_loader & ml) { ml.get_key_or_arr(LLM_KV_ROPE_DIMENSION_SECTIONS, hparams.rope_sections, 4, false); // NextN/MTP parameters (GLM-OCR) - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl"); - // TODO: when MTP is implemented, this should probably be updated if needed - hparams.n_layer_kv_from_start = hparams.n_layer - hparams.nextn_predict_layers; - - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 17: type = LLM_TYPE_1B; break; // GLM-OCR case 40: type = LLM_TYPE_9B; break; case 61: type = LLM_TYPE_32B; break; @@ -32,9 +29,9 @@ void llama_model_glm4::load_arch_tensors(llama_model_loader &) { output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); } - for (int i = 0; i < n_layer; ++i) { + for (int i = 0; i < n_layer_all; ++i) { int flags = 0; - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { // skip all tensors in the NextN layers flags |= TENSOR_SKIP; } @@ -55,7 +52,7 @@ void llama_model_glm4::load_arch_tensors(llama_model_loader &) { layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, flags); // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers - if (hparams.nextn_predict_layers > 0 && static_cast(i) >= n_layer - hparams.nextn_predict_layers) { + if (i >= n_layer) { layer.nextn.eh_proj = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags); layer.nextn.enorm = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags); layer.nextn.hnorm = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), { n_embd }, flags); @@ -100,8 +97,7 @@ llama_model_glm4::graph::graph(const llama_model & model, const llm_graph_params // Only process up to last layer (skip final NextN layer) // Final layer tensors are loaded but not processed in forward pass - const int n_transformer_layers = n_layer - hparams.nextn_predict_layers; - for (int il = 0; il < n_transformer_layers; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; // Pre-attention norm @@ -140,7 +136,7 @@ llama_model_glm4::graph::graph(const llama_model & model, const llm_graph_params model.layers[il].wo, NULL, model.layers[il].wo_s, Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, 1.0f / sqrtf(float(n_embd_head)), il); } - if (il == n_transformer_layers - 1 && inp_out_ids) { + if (il == n_layer - 1 && inp_out_ids) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } diff --git a/src/models/gpt2.cpp b/src/models/gpt2.cpp index e2dcc8b15..45afbccc1 100644 --- a/src/models/gpt2.cpp +++ b/src/models/gpt2.cpp @@ -2,7 +2,8 @@ void llama_model_gpt2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 12: type = LLM_TYPE_SMALL; break; case 24: type = LLM_TYPE_MEDIUM; break; case 36: type = LLM_TYPE_LARGE; break; diff --git a/src/models/gptneox.cpp b/src/models/gptneox.cpp index 443e35add..ed5e8c50d 100644 --- a/src/models/gptneox.cpp +++ b/src/models/gptneox.cpp @@ -3,7 +3,8 @@ void llama_model_gptneox::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); ml.get_key(LLM_KV_USE_PARALLEL_RESIDUAL, hparams.use_par_res); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 6: switch (hparams.n_ff()) { case 512: type = LLM_TYPE_14M; break; diff --git a/src/models/granite-hybrid.cpp b/src/models/granite-hybrid.cpp index 8740d9fc7..eb23095ae 100644 --- a/src/models/granite-hybrid.cpp +++ b/src/models/granite-hybrid.cpp @@ -19,7 +19,7 @@ void llama_model_granite_hybrid::load_arch_hparams(llama_model_loader & ml) { hparams.rope_finetuned = rope_finetuned; // A layer is recurrent IFF the n_head_kv value is set to 0 - for (uint32_t i = 0; i < hparams.n_layer; ++i) { + for (uint32_t i = 0; i < hparams.n_layer(); ++i) { hparams.is_recr_impl[i] = hparams.n_head_kv(i) == 0; } diff --git a/src/models/granite-moe.cpp b/src/models/granite-moe.cpp index 0d89bc1f3..115263c41 100644 --- a/src/models/granite-moe.cpp +++ b/src/models/granite-moe.cpp @@ -12,7 +12,7 @@ void llama_model_granite_moe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ROPE_SCALING_FINETUNED, rope_finetuned, false); hparams.rope_finetuned = rope_finetuned; - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_3B; break; case 40: type = LLM_TYPE_3B; break; // Add additional layer/vocab/etc checks here for other model sizes diff --git a/src/models/granite.cpp b/src/models/granite.cpp index cda4aa231..7aff942da 100644 --- a/src/models/granite.cpp +++ b/src/models/granite.cpp @@ -12,7 +12,7 @@ void llama_model_granite::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ROPE_SCALING_FINETUNED, rope_finetuned, false); hparams.rope_finetuned = rope_finetuned; - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_3B; break; case 40: type = LLM_TYPE_3B; break; // Add additional layer/vocab/etc checks here for other model sizes diff --git a/src/models/grok.cpp b/src/models/grok.cpp index 7c46ec1c0..42f38af67 100644 --- a/src/models/grok.cpp +++ b/src/models/grok.cpp @@ -26,7 +26,7 @@ void llama_model_grok::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ROPE_SCALING_YARN_BETA_FAST, hparams.yarn_beta_fast, false); ml.get_key(LLM_KV_ROPE_SCALING_YARN_BETA_SLOW, hparams.yarn_beta_slow, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 64: type = LLM_TYPE_314B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/grovemoe.cpp b/src/models/grovemoe.cpp index 1cab75adc..643a448e5 100644 --- a/src/models/grovemoe.cpp +++ b/src/models/grovemoe.cpp @@ -7,7 +7,7 @@ void llama_model_grovemoe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERTS_PER_GROUP, hparams.n_group_experts); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 48: type = LLM_TYPE_30B_A3B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/hunyuan-moe.cpp b/src/models/hunyuan-moe.cpp index deb3c9671..4d55f5e7f 100644 --- a/src/models/hunyuan-moe.cpp +++ b/src/models/hunyuan-moe.cpp @@ -5,7 +5,7 @@ void llama_model_hunyuan_moe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp); ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_A13B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/internlm2.cpp b/src/models/internlm2.cpp index f9ee37a24..f6cfdfb94 100644 --- a/src/models/internlm2.cpp +++ b/src/models/internlm2.cpp @@ -2,7 +2,8 @@ void llama_model_internlm2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_7B; break; case 48: type = LLM_TYPE_20B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/jais.cpp b/src/models/jais.cpp index 2ba162605..415103ce2 100644 --- a/src/models/jais.cpp +++ b/src/models/jais.cpp @@ -4,7 +4,7 @@ void llama_model_jais::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); ml.get_key(LLM_KV_ATTENTION_MAX_ALIBI_BIAS, hparams.f_max_alibi_bias, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_1_3B; break; case 40: type = LLM_TYPE_13B; break; /* TODO: add variants */ diff --git a/src/models/jais2.cpp b/src/models/jais2.cpp index 896613144..8610fcc9f 100644 --- a/src/models/jais2.cpp +++ b/src/models/jais2.cpp @@ -3,7 +3,7 @@ void llama_model_jais2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_8B; break; case 68: type = LLM_TYPE_70B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/jamba.cpp b/src/models/jamba.cpp index a62b121b3..dba160b01 100644 --- a/src/models/jamba.cpp +++ b/src/models/jamba.cpp @@ -8,11 +8,11 @@ void llama_model_jamba::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - for (uint32_t i = 0; i < hparams.n_layer; ++i) { + for (uint32_t i = 0; i < hparams.n_layer(); ++i) { hparams.is_recr_impl[i] = hparams.n_head_kv(i) == 0; } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { // TODO: Jamba layers are a bit heterogeneous, so naming this is hard. case 12: // 900M 8x???M case 32: // 51B 16x?B diff --git a/src/models/jina-bert-v2.cpp b/src/models/jina-bert-v2.cpp index 4f8866ece..86ff1c84d 100644 --- a/src/models/jina-bert-v2.cpp +++ b/src/models/jina-bert-v2.cpp @@ -4,7 +4,7 @@ void llama_model_jina_bert_v2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); hparams.f_max_alibi_bias = 8.0f; - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 4: type = LLM_TYPE_33M; break; // jina-embeddings-small case 12: type = LLM_TYPE_137M; break; // jina-embeddings-base default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/jina-bert-v3.cpp b/src/models/jina-bert-v3.cpp index e0527529f..1c974a6f1 100644 --- a/src/models/jina-bert-v3.cpp +++ b/src/models/jina-bert-v3.cpp @@ -3,7 +3,7 @@ void llama_model_jina_bert_v3::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_558M; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/kimi-linear.cpp b/src/models/kimi-linear.cpp index c13f71b5b..367f6990d 100644 --- a/src/models/kimi-linear.cpp +++ b/src/models/kimi-linear.cpp @@ -14,7 +14,7 @@ void llama_model_kimi_linear::load_arch_hparams(llama_model_loader & ml) { // Mark KDA layers as recurrent using n_head_kv pattern (like Jamba) // Set n_head_kv = 0 for KDA layers (recurrent), n_head_kv = n_head for MLA layers (attention) - for (uint32_t i = 0; i < hparams.n_layer; ++i) { + for (uint32_t i = 0; i < hparams.n_layer(); ++i) { hparams.is_recr_impl[i] = hparams.n_head_kv(i) == 0; // KDA layers are recurrent } @@ -25,7 +25,7 @@ void llama_model_kimi_linear::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale, false); ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 27: type = LLM_TYPE_48B_A3B; break; // Kimi-Linear-48B-A3B default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/lfm2.cpp b/src/models/lfm2.cpp index 3898b56bb..97da8a6ab 100644 --- a/src/models/lfm2.cpp +++ b/src/models/lfm2.cpp @@ -5,10 +5,13 @@ void llama_model_lfm2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_SHORTCONV_L_CACHE, hparams.n_shortconv_l_cache); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - for (uint32_t il = 0; il < hparams.n_layer; ++il) { + + for (uint32_t il = 0; il < hparams.n_layer(); ++il) { hparams.is_recr_impl[il] = hparams.n_head_kv(il) == 0; } - hparams.n_layer_dense_lead = hparams.n_layer; + + hparams.n_layer_dense_lead = hparams.n_layer(); + switch (hparams.n_ff()) { case 4608: type = LLM_TYPE_350M; break; case 6912: type = LLM_TYPE_700M; break; @@ -16,9 +19,10 @@ void llama_model_lfm2::load_arch_hparams(llama_model_loader & ml) { case 10752: type = LLM_TYPE_2_6B; break; default: type = LLM_TYPE_UNKNOWN; } + if (const auto is_swa = ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false); is_swa && hparams.n_swa > 0) { hparams.swa_type = LLAMA_SWA_TYPE_STANDARD; - for (uint32_t il = 0; il < hparams.n_layer; ++il) { + for (uint32_t il = 0; il < hparams.n_layer(); ++il) { hparams.is_swa_impl[il] = !hparams.is_recr_impl[il]; } } diff --git a/src/models/lfm2moe.cpp b/src/models/lfm2moe.cpp index 81ced2eab..490f5c223 100644 --- a/src/models/lfm2moe.cpp +++ b/src/models/lfm2moe.cpp @@ -9,11 +9,11 @@ void llama_model_lfm2moe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp); ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func); - for (uint32_t il = 0; il < hparams.n_layer; ++il) { + for (uint32_t il = 0; il < hparams.n_layer(); ++il) { hparams.is_recr_impl[il] = hparams.n_head_kv(il) == 0; } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_8B_A1B; break; case 40: type = LLM_TYPE_24B_A2B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/llada-moe.cpp b/src/models/llada-moe.cpp index 9722dde9f..2ae893864 100644 --- a/src/models/llada-moe.cpp +++ b/src/models/llada-moe.cpp @@ -2,11 +2,12 @@ void llama_model_llada_moe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp, false); - ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); + // diffusion language model uses non-causal attention hparams.causal_attn = false; - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 16: type = LLM_TYPE_A1_7B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/llada.cpp b/src/models/llada.cpp index 58b2c466e..87d4259f9 100644 --- a/src/models/llada.cpp +++ b/src/models/llada.cpp @@ -2,14 +2,16 @@ void llama_model_llada::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); + // LLaDA-8B has 32 layers, similar to LLaMA but for diffusion - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_8B; break; default: type = LLM_TYPE_UNKNOWN; } + // Set non-causal attention for diffusion models hparams.causal_attn = false; } diff --git a/src/models/llama.cpp b/src/models/llama.cpp index cef66d054..c0ec7e0a9 100644 --- a/src/models/llama.cpp +++ b/src/models/llama.cpp @@ -7,13 +7,13 @@ void llama_model_llama::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); if (hparams.n_expert == 8) { - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_8x7B; break; case 56: type = LLM_TYPE_8x22B; break; default: type = LLM_TYPE_UNKNOWN; } } else { - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 16: type = LLM_TYPE_1B; break; // Llama 3.2 1B case 22: type = LLM_TYPE_1B; break; case 26: type = LLM_TYPE_3B; break; diff --git a/src/models/llama4.cpp b/src/models/llama4.cpp index 8f39b3f59..7194c72a5 100644 --- a/src/models/llama4.cpp +++ b/src/models/llama4.cpp @@ -8,7 +8,7 @@ void llama_model_llama4::load_arch_hparams(llama_model_loader & ml) { const bool found_swa = ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false); if (found_swa && hparams.n_swa == 0) { hparams.swa_type = LLAMA_SWA_TYPE_NONE; - hparams.n_no_rope_layer_step = hparams.n_layer; // always use rope + hparams.n_no_rope_layer_step = hparams.n_layer(); // always use rope } else { hparams.swa_type = LLAMA_SWA_TYPE_CHUNKED; hparams.n_swa = 8192; diff --git a/src/models/maincoder.cpp b/src/models/maincoder.cpp index 84cfe3990..ae56a26a1 100644 --- a/src/models/maincoder.cpp +++ b/src/models/maincoder.cpp @@ -2,7 +2,8 @@ void llama_model_maincoder::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_1B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/mamba.cpp b/src/models/mamba.cpp index 887a1fa50..0d94e9828 100644 --- a/src/models/mamba.cpp +++ b/src/models/mamba.cpp @@ -9,7 +9,7 @@ void llama_model_mamba::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: switch (hparams.n_embd) { case 768: type = LLM_TYPE_SMALL; break; diff --git a/src/models/mamba2.cpp b/src/models/mamba2.cpp index 3277ca53e..c5951cf0f 100644 --- a/src/models/mamba2.cpp +++ b/src/models/mamba2.cpp @@ -9,7 +9,7 @@ void llama_model_mamba2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: switch (hparams.n_embd) { case 768: type = LLM_TYPE_SMALL; break; diff --git a/src/models/mellum.cpp b/src/models/mellum.cpp index 1e1e97e9f..28823018b 100644 --- a/src/models/mellum.cpp +++ b/src/models/mellum.cpp @@ -13,7 +13,7 @@ void llama_model_mellum::load_arch_hparams(llama_model_loader & ml) { if (res) { hparams.set_swa_pattern(swa_period); } else { - ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer); + ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer()); } hparams.rope_freq_base_train_swa = hparams.rope_freq_base_train; @@ -24,7 +24,7 @@ void llama_model_mellum::load_arch_hparams(llama_model_loader & ml) { hparams.swa_type = LLAMA_SWA_TYPE_NONE; } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 28: type = LLM_TYPE_12B_A2_5B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/mimo2.cpp b/src/models/mimo2.cpp index 1bcdf696f..889891605 100644 --- a/src/models/mimo2.cpp +++ b/src/models/mimo2.cpp @@ -9,18 +9,17 @@ void llama_model_mimo2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa); ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false); - ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer); + ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer()); float value_scale = 0.0f; if (ml.get_key(LLM_KV_ATTENTION_VALUE_SCALE, value_scale, false) && value_scale != 1.0f) { hparams.f_attn_value_scale = value_scale; } - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); - hparams.n_layer_kv_from_start = hparams.n_layer - hparams.nextn_predict_layers; + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl"); - switch (hparams.n_layer - hparams.nextn_predict_layers) { + switch (hparams.n_layer()) { case 48: type = LLM_TYPE_310B_A15B; break; default: type = LLM_TYPE_UNKNOWN; } @@ -35,16 +34,14 @@ void llama_model_mimo2::load_arch_tensors(llama_model_loader &) { output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0); output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, 0); - const uint32_t n_nextn = hparams.nextn_predict_layers; - - for (int i = 0; i < n_layer; ++i) { + for (int i = 0; i < n_layer_all; ++i) { auto & layer = layers[i]; uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(i); uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(i); uint32_t n_head = hparams.n_head(i); // NextN/MTP layers (the last n_nextn blocks) are preserved but disabled pending support - const bool is_nextn = (n_nextn > 0) && (static_cast(i) >= n_layer - n_nextn); + const bool is_nextn = i >= n_layer; const int skip = is_nextn ? TENSOR_SKIP : 0; create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, skip); @@ -93,10 +90,7 @@ llama_model_mimo2::graph::graph(const llama_model & model, const llm_graph_param const float v_scale = hparams.f_attn_value_scale; - // The last hparams.nextn_predict_layers blocks are MTP heads, currently inactive - const int n_transformer_layers = n_layer - hparams.nextn_predict_layers; - - for (int il = 0; il < n_transformer_layers; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; uint32_t n_head_l = hparams.n_head(il); @@ -174,7 +168,7 @@ llama_model_mimo2::graph::graph(const llama_model & model, const llm_graph_param } } - if (il == n_transformer_layers - 1 && inp_out_ids) { + if (il == n_layer - 1 && inp_out_ids) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } diff --git a/src/models/minicpm.cpp b/src/models/minicpm.cpp index 966d3af61..fc3e5b171 100644 --- a/src/models/minicpm.cpp +++ b/src/models/minicpm.cpp @@ -3,7 +3,7 @@ void llama_model_minicpm::load_arch_hparams(llama_model_loader & ml) { // Backward-compatible defaults for older MiniCPM GGUFs hparams.f_embedding_scale = 12.0f; - hparams.f_residual_scale = 1.4f / sqrtf(float(hparams.n_layer)); + hparams.f_residual_scale = 1.4f / sqrtf(float(hparams.n_layer())); hparams.f_logit_scale = hparams.n_embd ? (256.0f / float(hparams.n_embd)) : 1.0f; ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); @@ -16,7 +16,7 @@ void llama_model_minicpm::load_arch_hparams(llama_model_loader & ml) { // MiniCPM uses rope by default, unlike Granite which uses it as a switch hparams.rope_finetuned = true; - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 52: type = LLM_TYPE_1B; break; case 40: type = LLM_TYPE_2B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/minicpm3.cpp b/src/models/minicpm3.cpp index 1ffc54fa7..e011b1ff0 100644 --- a/src/models/minicpm3.cpp +++ b/src/models/minicpm3.cpp @@ -5,7 +5,7 @@ void llama_model_minicpm3::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_Q_LORA_RANK, hparams.n_lora_q); ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK, hparams.n_lora_kv); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 62: type = LLM_TYPE_4B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/minimax-m2.cpp b/src/models/minimax-m2.cpp index 22e291d73..b25435e4d 100644 --- a/src/models/minimax-m2.cpp +++ b/src/models/minimax-m2.cpp @@ -5,7 +5,7 @@ void llama_model_minimax_m2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp); ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 62: type = LLM_TYPE_230B_A10B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/mistral3.cpp b/src/models/mistral3.cpp index 1ac5a95cc..9a8e3f9a5 100644 --- a/src/models/mistral3.cpp +++ b/src/models/mistral3.cpp @@ -18,7 +18,7 @@ void llama_model_mistral3::load_arch_hparams(llama_model_loader & ml) { } } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 26: type = LLM_TYPE_3B; break; case 34: type = LLM_TYPE_8B; break; case 40: type = LLM_TYPE_14B; break; diff --git a/src/models/modern-bert.cpp b/src/models/modern-bert.cpp index 5ab51867c..f3e9407e0 100644 --- a/src/models/modern-bert.cpp +++ b/src/models/modern-bert.cpp @@ -22,7 +22,7 @@ void llama_model_modern_bert::load_arch_hparams(llama_model_loader & ml) { hparams.llm_ffn_op = llm_ffn_op_type_from_string(hidden_act, LLM_FFN_GEGLU); } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 12: type = LLM_TYPE_47M; break; // granite-embedding-small case 22: diff --git a/src/models/mpt.cpp b/src/models/mpt.cpp index 0229d20ed..d094fd9f8 100644 --- a/src/models/mpt.cpp +++ b/src/models/mpt.cpp @@ -5,7 +5,7 @@ void llama_model_mpt::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV, hparams.f_clamp_kqv, false); ml.get_key(LLM_KV_ATTENTION_MAX_ALIBI_BIAS, hparams.f_max_alibi_bias, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_7B; break; case 48: type = LLM_TYPE_30B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/nemotron-h.cpp b/src/models/nemotron-h.cpp index d2c811d24..a45626934 100644 --- a/src/models/nemotron-h.cpp +++ b/src/models/nemotron-h.cpp @@ -9,7 +9,7 @@ void llama_model_nemotron_h::load_arch_hparams(llama_model_loader & ml) { // A layer is recurrent IFF the n_head_kv value is set to 0 and // the n_ff value is set to 0 - for (uint32_t i = 0; i < hparams.n_layer; ++i) { + for (uint32_t i = 0; i < hparams.n_layer(); ++i) { hparams.is_recr_impl[i] = (hparams.n_head_kv(i) == 0 && hparams.n_ff(i) == 0); } @@ -22,7 +22,7 @@ void llama_model_nemotron_h::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale, false); ml.get_key(LLM_KV_MOE_LATENT_SIZE, hparams.moe_latent_size, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 52: type = LLM_TYPE_31B_A3_5B; break; // Nemotron-H_MOE 31B case 56: type = LLM_TYPE_9B; break; case 88: type = LLM_TYPE_120B_A12B; break; diff --git a/src/models/nemotron.cpp b/src/models/nemotron.cpp index 5d4a3b5c6..6e2bd9a33 100644 --- a/src/models/nemotron.cpp +++ b/src/models/nemotron.cpp @@ -2,7 +2,8 @@ void llama_model_nemotron::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_4B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/neo-bert.cpp b/src/models/neo-bert.cpp index f00d6eddf..4a08d7abd 100644 --- a/src/models/neo-bert.cpp +++ b/src/models/neo-bert.cpp @@ -3,7 +3,7 @@ void llama_model_neo_bert::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - if (hparams.n_layer == 28) { + if (hparams.n_layer() == 28) { type = LLM_TYPE_250M; } } diff --git a/src/models/nomic-bert-moe.cpp b/src/models/nomic-bert-moe.cpp index a17abe2c2..da4b62919 100644 --- a/src/models/nomic-bert-moe.cpp +++ b/src/models/nomic-bert-moe.cpp @@ -4,7 +4,7 @@ void llama_model_nomic_bert_moe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); ml.get_key(LLM_KV_MOE_EVERY_N_LAYERS, hparams.moe_every_n_layers, 0); - if (hparams.n_layer == 12 && hparams.n_embd == 768) { + if (hparams.n_layer() == 12 && hparams.n_embd == 768) { if (arch == LLM_ARCH_NOMIC_BERT) { type = LLM_TYPE_137M; } else if (arch == LLM_ARCH_NOMIC_BERT_MOE && hparams.moe_every_n_layers == 2) { diff --git a/src/models/nomic-bert.cpp b/src/models/nomic-bert.cpp index 5a8a55844..e7fc72286 100644 --- a/src/models/nomic-bert.cpp +++ b/src/models/nomic-bert.cpp @@ -4,7 +4,7 @@ void llama_model_nomic_bert::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); ml.get_key(LLM_KV_MOE_EVERY_N_LAYERS, hparams.moe_every_n_layers, 0); - if (hparams.n_layer == 12 && hparams.n_embd == 768) { + if (hparams.n_layer() == 12 && hparams.n_embd == 768) { if (arch == LLM_ARCH_NOMIC_BERT) { type = LLM_TYPE_137M; } else if (arch == LLM_ARCH_NOMIC_BERT_MOE && hparams.moe_every_n_layers == 2) { diff --git a/src/models/olmo.cpp b/src/models/olmo.cpp index cfcf17bcb..9f7a2ba60 100644 --- a/src/models/olmo.cpp +++ b/src/models/olmo.cpp @@ -4,7 +4,7 @@ void llama_model_olmo::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV, hparams.f_clamp_kqv, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 22: type = LLM_TYPE_1B; break; case 32: type = LLM_TYPE_7B; break; case 80: type = LLM_TYPE_70B; break; diff --git a/src/models/olmo2.cpp b/src/models/olmo2.cpp index 7cc262f55..cb52cdef7 100644 --- a/src/models/olmo2.cpp +++ b/src/models/olmo2.cpp @@ -17,7 +17,7 @@ void llama_model_olmo2::load_arch_hparams(llama_model_loader & ml) { hparams.swa_type = LLAMA_SWA_TYPE_NONE; } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 16: type = LLM_TYPE_1B; break; case 32: type = LLM_TYPE_7B; break; case 40: type = LLM_TYPE_13B; break; diff --git a/src/models/olmoe.cpp b/src/models/olmoe.cpp index 7976ae44a..1e2baeb20 100644 --- a/src/models/olmoe.cpp +++ b/src/models/olmoe.cpp @@ -2,7 +2,8 @@ void llama_model_olmoe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 16: type = LLM_TYPE_A1_7B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/openai-moe.cpp b/src/models/openai-moe.cpp index 15b6c8c12..3ab15d61f 100644 --- a/src/models/openai-moe.cpp +++ b/src/models/openai-moe.cpp @@ -14,7 +14,7 @@ void llama_model_openai_moe::load_arch_hparams(llama_model_loader & ml) { hparams.rope_freq_scale_train_swa = hparams.rope_freq_scale_train; ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_20B; break; case 36: type = LLM_TYPE_120B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/openelm.cpp b/src/models/openelm.cpp index 9f76350fd..13120bd32 100644 --- a/src/models/openelm.cpp +++ b/src/models/openelm.cpp @@ -3,12 +3,12 @@ void llama_model_openelm::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { - case 16: type = LLM_TYPE_270M; break; - case 20: type = LLM_TYPE_450M; break; - case 28: type = LLM_TYPE_1B; break; - case 36: type = LLM_TYPE_3B; break; - default: type = LLM_TYPE_UNKNOWN; + switch (hparams.n_layer()) { + case 16: type = LLM_TYPE_270M; break; + case 20: type = LLM_TYPE_450M; break; + case 28: type = LLM_TYPE_1B; break; + case 36: type = LLM_TYPE_3B; break; + default: type = LLM_TYPE_UNKNOWN; } } diff --git a/src/models/orion.cpp b/src/models/orion.cpp index bcb4bbba4..863a28222 100644 --- a/src/models/orion.cpp +++ b/src/models/orion.cpp @@ -3,7 +3,7 @@ void llama_model_orion::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 40: type = LLM_TYPE_14B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/pangu-embed.cpp b/src/models/pangu-embed.cpp index 7593f879b..90f05c088 100644 --- a/src/models/pangu-embed.cpp +++ b/src/models/pangu-embed.cpp @@ -2,7 +2,8 @@ void llama_model_pangu_embed::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 26: type = LLM_TYPE_1B; break; // openPangu-Embedded-1B-V1.1 case 34: type = LLM_TYPE_7B; break; // openPangu-Embedded-7B-V1.1 default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/phi2.cpp b/src/models/phi2.cpp index 8f3ed5f7b..81b1ad12c 100644 --- a/src/models/phi2.cpp +++ b/src/models/phi2.cpp @@ -3,7 +3,7 @@ void llama_model_phi2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_1B; break; case 32: type = LLM_TYPE_3B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/phi3.cpp b/src/models/phi3.cpp index f8a4a4d5a..716ff814c 100644 --- a/src/models/phi3.cpp +++ b/src/models/phi3.cpp @@ -3,7 +3,7 @@ void llama_model_phi3::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_1B; break; case 32: type = LLM_TYPE_3B; break; case 40: type = LLM_TYPE_14B; break; diff --git a/src/models/phimoe.cpp b/src/models/phimoe.cpp index 4575d6139..c332553bc 100644 --- a/src/models/phimoe.cpp +++ b/src/models/phimoe.cpp @@ -3,7 +3,7 @@ void llama_model_phimoe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_16x3_8B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/plamo.cpp b/src/models/plamo.cpp index c7ed1211c..246144519 100644 --- a/src/models/plamo.cpp +++ b/src/models/plamo.cpp @@ -3,7 +3,7 @@ void llama_model_plamo::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 40: type = LLM_TYPE_13B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/plamo2.cpp b/src/models/plamo2.cpp index 2ffa0898f..b93cf48bc 100644 --- a/src/models/plamo2.cpp +++ b/src/models/plamo2.cpp @@ -11,11 +11,11 @@ void llama_model_plamo2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_SSM_TIME_STEP_RANK, hparams.ssm_dt_rank); ml.get_key(LLM_KV_SSM_GROUP_COUNT, hparams.ssm_n_group); - for (uint32_t i = 0; i < hparams.n_layer; ++i) { + for (uint32_t i = 0; i < hparams.n_layer(); ++i) { hparams.is_recr_impl[i] = hparams.n_head_kv(i) == 0; } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 16: type = LLM_TYPE_1B; break; case 32: if (hparams.n_embd == 2048) { diff --git a/src/models/plamo3.cpp b/src/models/plamo3.cpp index 29f3e803d..16d0b1dce 100644 --- a/src/models/plamo3.cpp +++ b/src/models/plamo3.cpp @@ -13,7 +13,7 @@ void llama_model_plamo3::load_arch_hparams(llama_model_loader & ml) { hparams.swa_type = LLAMA_SWA_TYPE_NONE; } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_2B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/plm.cpp b/src/models/plm.cpp index ce050919e..8ca325f5e 100644 --- a/src/models/plm.cpp +++ b/src/models/plm.cpp @@ -3,7 +3,8 @@ void llama_model_plm::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK, hparams.n_lora_kv); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_1_8B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/qwen.cpp b/src/models/qwen.cpp index 00467dbad..1f5dff384 100644 --- a/src/models/qwen.cpp +++ b/src/models/qwen.cpp @@ -3,7 +3,7 @@ void llama_model_qwen::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_7B; break; case 40: type = LLM_TYPE_13B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/qwen2.cpp b/src/models/qwen2.cpp index a5147460b..e9c2ea80a 100644 --- a/src/models/qwen2.cpp +++ b/src/models/qwen2.cpp @@ -2,7 +2,8 @@ void llama_model_qwen2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 24: type = hparams.n_embd == 1024 ? LLM_TYPE_0_5B : LLM_TYPE_1B; break; case 28: type = hparams.n_embd == 1536 ? LLM_TYPE_1_5B : LLM_TYPE_7B; break; case 32: type = LLM_TYPE_7B; break; diff --git a/src/models/qwen2moe.cpp b/src/models/qwen2moe.cpp index 7cb03859d..e831ed11a 100644 --- a/src/models/qwen2moe.cpp +++ b/src/models/qwen2moe.cpp @@ -5,7 +5,8 @@ void llama_model_qwen2moe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_A2_7B; break; case 28: type = LLM_TYPE_57B_A14B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/qwen3.cpp b/src/models/qwen3.cpp index 41b97fed9..1d0d2fab3 100644 --- a/src/models/qwen3.cpp +++ b/src/models/qwen3.cpp @@ -2,7 +2,8 @@ void llama_model_qwen3::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 28: type = hparams.n_embd == 1024 ? LLM_TYPE_0_6B : LLM_TYPE_1_7B; break; case 36: type = hparams.n_embd == 2560 ? LLM_TYPE_4B : LLM_TYPE_8B; break; case 40: type = LLM_TYPE_14B; break; diff --git a/src/models/qwen35.cpp b/src/models/qwen35.cpp index 348650b37..4b642cff4 100644 --- a/src/models/qwen35.cpp +++ b/src/models/qwen35.cpp @@ -13,22 +13,20 @@ void llama_model_qwen35::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_SSM_GROUP_COUNT, hparams.ssm_n_group); // NextN/MTP (Qwen3.5/3.6): extra decoder block appended beyond the main stack - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl"); // Mark recurrent layers (linear attention layers). MTP layers are dense // attention-only and must be flagged non-recurrent. - if (!ml.get_key_or_arr(LLM_KV_ATTENTION_RECURRENT_LAYERS, hparams.is_recr_impl, hparams.n_layer, false)) { - const uint32_t n_main = hparams.n_layer - hparams.nextn_predict_layers; - + if (!ml.get_key_or_arr(LLM_KV_ATTENTION_RECURRENT_LAYERS, hparams.is_recr_impl, hparams.n_layer_all, false)) { uint32_t full_attn_interval = 4; ml.get_key(LLM_KV_FULL_ATTENTION_INTERVAL, full_attn_interval, false); - for (uint32_t i = 0; i < hparams.n_layer; ++i) { - hparams.is_recr_impl[i] = (i < n_main) && ((i + 1) % full_attn_interval != 0); + for (uint32_t i = 0; i < hparams.n_layer_all; ++i) { + hparams.is_recr_impl[i] = (i < hparams.n_layer()) && ((i + 1) % full_attn_interval != 0); } } - switch (hparams.n_layer - hparams.nextn_predict_layers) { + switch (hparams.n_layer()) { case 24: type = hparams.n_embd == 1024 ? LLM_TYPE_0_8B : LLM_TYPE_2B; break; case 32: type = hparams.n_embd == 2560 ? LLM_TYPE_4B : LLM_TYPE_9B; break; case 64: type = LLM_TYPE_27B; break; @@ -39,9 +37,7 @@ void llama_model_qwen35::load_arch_hparams(llama_model_loader & ml) { void llama_model_qwen35::load_arch_tensors(llama_model_loader & ml) { LLAMA_LOAD_LOCALS; - const uint32_t n_main = n_layer - hparams.nextn_predict_layers; - const bool mtp_only = (hparams.nextn_predict_layers > 0) && - (ml.get_weight("blk.0.attn_norm.weight") == nullptr); + const bool mtp_only = (hparams.n_layer_nextn > 0) && (ml.get_weight("blk.0.attn_norm.weight") == nullptr); const int trunk_flags = mtp_only ? TENSOR_NOT_REQUIRED : 0; tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0); @@ -122,10 +118,10 @@ void llama_model_qwen35::load_arch_tensors(llama_model_loader & ml) { layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", il), { n_embd }, TENSOR_NOT_REQUIRED); }; - for (int i = 0; i < (int) n_main; ++i) { + for (int i = 0; i < n_layer; ++i) { load_block_trunk(i, trunk_flags); } - for (int i = (int) n_main; i < n_layer; ++i) { + for (int i = n_layer; i < n_layer_all; ++i) { load_block_mtp(i); } } @@ -159,8 +155,7 @@ llama_model_qwen35::graph::graph(const llama_model & model, const llm_graph_para ggml_tensor * inp_out_ids = build_inp_out_ids(); // MTP/NextN layers are loaded as extra decoder blocks but not executed in the main pass. - const int n_transformer_layers = n_layer - (int) hparams.nextn_predict_layers; - for (int il = 0; il < n_transformer_layers; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; cur = build_norm(inpL, model.layers[il].attn_norm, nullptr, LLM_NORM_RMS, il); @@ -177,7 +172,7 @@ llama_model_qwen35::graph::graph(const llama_model & model, const llm_graph_para cur = build_layer_attn(inp->get_attn(), cur, inp_pos, sections, il); } - if (il == n_transformer_layers - 1 && inp_out_ids && cparams.embeddings_nextn_masked) { + if (il == n_layer - 1 && inp_out_ids && cparams.embeddings_nextn_masked) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } @@ -490,15 +485,15 @@ ggml_tensor * llama_model_qwen35::graph::build_layer_ffn(ggml_tensor * cur, cons // LLM_GRAPH_TYPE_DECODER_MTP draft head for Qwen3.5/3.6 dense series llama_model_qwen35::graph_mtp::graph_mtp(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) { - GGML_ASSERT(hparams.nextn_predict_layers > 0 && "QWEN35 MTP requires nextn_predict_layers > 0"); - GGML_ASSERT(hparams.nextn_predict_layers == 1 && "QWEN35 MTP currently only supports a single MTP block"); + GGML_ASSERT(hparams.n_layer_nextn > 0 && "QWEN35 MTP requires n_layer_nextn > 0"); + GGML_ASSERT(hparams.n_layer_nextn == 1 && "QWEN35 MTP currently only supports a single MTP block"); const int64_t n_embd_head = hparams.n_embd_head_v(); GGML_ASSERT(n_embd_head == hparams.n_embd_head_k()); // hparams.n_layer includes both main model layers and MTP layers. The MTP // layer is stored immediately after the main layers in model.layers[]. - const int il = (int) hparams.n_layer - (int) hparams.nextn_predict_layers; + const int il = hparams.n_layer(); const auto & layer = model.layers[il]; GGML_ASSERT(layer.nextn.eh_proj && "MTP block missing nextn.eh_proj"); diff --git a/src/models/qwen35moe.cpp b/src/models/qwen35moe.cpp index 7d906191c..eb5e9a406 100644 --- a/src/models/qwen35moe.cpp +++ b/src/models/qwen35moe.cpp @@ -16,22 +16,20 @@ void llama_model_qwen35moe::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_SSM_GROUP_COUNT, hparams.ssm_n_group); // NextN/MTP (Qwen3.5/3.6): extra decoder block appended beyond the main stack - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl"); // Mark recurrent layers (linear attention layers). MTP layers are dense // attention-only and must be flagged non-recurrent. - if (!ml.get_key_or_arr(LLM_KV_ATTENTION_RECURRENT_LAYERS, hparams.is_recr_impl, hparams.n_layer, false)) { - const uint32_t n_main = hparams.n_layer - hparams.nextn_predict_layers; - + if (!ml.get_key_or_arr(LLM_KV_ATTENTION_RECURRENT_LAYERS, hparams.is_recr_impl, hparams.n_layer_all, false)) { uint32_t full_attn_interval = 4; ml.get_key(LLM_KV_FULL_ATTENTION_INTERVAL, full_attn_interval, false); - for (uint32_t i = 0; i < hparams.n_layer; ++i) { - hparams.is_recr_impl[i] = (i < n_main) && ((i + 1) % full_attn_interval != 0); + for (uint32_t i = 0; i < hparams.n_layer_all; ++i) { + hparams.is_recr_impl[i] = (i < hparams.n_layer()) && ((i + 1) % full_attn_interval != 0); } } - switch (hparams.n_layer - hparams.nextn_predict_layers) { + switch (hparams.n_layer()) { case 40: type = LLM_TYPE_35B_A3B; break; case 48: type = LLM_TYPE_122B_A10B; break; case 60: type = LLM_TYPE_397B_A17B; break; @@ -42,9 +40,7 @@ void llama_model_qwen35moe::load_arch_hparams(llama_model_loader & ml) { void llama_model_qwen35moe::load_arch_tensors(llama_model_loader & ml) { LLAMA_LOAD_LOCALS; - const uint32_t n_main = n_layer - hparams.nextn_predict_layers; - const bool mtp_only = (hparams.nextn_predict_layers > 0) && - (ml.get_weight("blk.0.attn_norm.weight") == nullptr); + const bool mtp_only = (hparams.n_layer_nextn > 0) && (ml.get_weight("blk.0.attn_norm.weight") == nullptr); const int trunk_flags = mtp_only ? TENSOR_NOT_REQUIRED : 0; tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0); @@ -145,10 +141,10 @@ void llama_model_qwen35moe::load_arch_tensors(llama_model_loader & ml) { layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", il), { n_embd }, TENSOR_NOT_REQUIRED); }; - for (int i = 0; i < (int) n_main; ++i) { + for (int i = 0; i < n_layer; ++i) { load_block_trunk(i, trunk_flags); } - for (int i = (int) n_main; i < n_layer; ++i) { + for (int i = n_layer; i < n_layer_all; ++i) { load_block_mtp(i); } } @@ -182,8 +178,7 @@ llama_model_qwen35moe::graph::graph(const llama_model & model, const llm_graph_p ggml_tensor * inp_out_ids = build_inp_out_ids(); // MTP/NextN layers are loaded as extra decoder blocks but not executed in the main pass. - const int n_transformer_layers = n_layer - (int) hparams.nextn_predict_layers; - for (int il = 0; il < n_transformer_layers; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; cur = build_norm(inpL, model.layers[il].attn_norm, nullptr, LLM_NORM_RMS, il); @@ -200,7 +195,7 @@ llama_model_qwen35moe::graph::graph(const llama_model & model, const llm_graph_p cur = build_layer_attn(inp->get_attn(), cur, inp_pos, sections, il); } - if (il == n_transformer_layers - 1 && inp_out_ids && cparams.embeddings_nextn_masked) { + if (il == n_layer - 1 && inp_out_ids && cparams.embeddings_nextn_masked) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } @@ -555,13 +550,13 @@ ggml_tensor * llama_model_qwen35moe::graph::build_layer_ffn(ggml_tensor * cur, c // LLM_GRAPH_TYPE_DECODER_MTP draft head for Qwen3.5/3.6 MoE llama_model_qwen35moe::graph_mtp::graph_mtp(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) { - GGML_ASSERT(hparams.nextn_predict_layers > 0 && "QWEN35MOE MTP requires nextn_predict_layers > 0"); - GGML_ASSERT(hparams.nextn_predict_layers == 1 && "QWEN35MOE MTP currently only supports a single MTP block"); + GGML_ASSERT(hparams.n_layer_nextn > 0 && "QWEN35MOE MTP requires n_layer_nextn > 0"); + GGML_ASSERT(hparams.n_layer_nextn == 1 && "QWEN35MOE MTP currently only supports a single MTP block"); const int64_t n_embd_head = hparams.n_embd_head_v(); GGML_ASSERT(n_embd_head == hparams.n_embd_head_k()); - const int il = (int) hparams.n_layer - (int) hparams.nextn_predict_layers; + const int il = hparams.n_layer(); const auto & layer = model.layers[il]; GGML_ASSERT(layer.nextn.eh_proj && "MTP block missing nextn.eh_proj"); diff --git a/src/models/qwen3moe.cpp b/src/models/qwen3moe.cpp index a4f8e1379..317e668be 100644 --- a/src/models/qwen3moe.cpp +++ b/src/models/qwen3moe.cpp @@ -1,10 +1,10 @@ #include "models.h" void llama_model_qwen3moe::load_arch_hparams(llama_model_loader & ml) { - ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp, false); - + ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp, false); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 48: type = LLM_TYPE_30B_A3B; break; case 94: type = LLM_TYPE_235B_A22B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/qwen3next.cpp b/src/models/qwen3next.cpp index c1e9b88a4..4505c92b0 100644 --- a/src/models/qwen3next.cpp +++ b/src/models/qwen3next.cpp @@ -14,15 +14,15 @@ void llama_model_qwen3next::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_SSM_GROUP_COUNT, hparams.ssm_n_group); // Mark recurrent layers (linear attention layers) - if (!ml.get_key_or_arr(LLM_KV_ATTENTION_RECURRENT_LAYERS, hparams.is_recr_impl, hparams.n_layer, false)) { + if (!ml.get_key_or_arr(LLM_KV_ATTENTION_RECURRENT_LAYERS, hparams.is_recr_impl, hparams.n_layer_all, false)) { uint32_t full_attn_interval = 4; ml.get_key(LLM_KV_FULL_ATTENTION_INTERVAL, full_attn_interval, false); - for (uint32_t i = 0; i < hparams.n_layer; ++i) { - hparams.is_recr_impl[i] = ((i + 1) % full_attn_interval != 0); + for (uint32_t i = 0; i < hparams.n_layer_all; ++i) { + hparams.is_recr_impl[i] = (i < hparams.n_layer()) && ((i + 1) % full_attn_interval != 0); } } - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 48: type = LLM_TYPE_80B_A3B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/qwen3vl.cpp b/src/models/qwen3vl.cpp index 5defd8939..724d6140d 100644 --- a/src/models/qwen3vl.cpp +++ b/src/models/qwen3vl.cpp @@ -4,7 +4,8 @@ void llama_model_qwen3vl::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_NUM_DEEPSTACK_LAYERS, hparams.n_deepstack_layers, false); ml.get_key_or_arr(LLM_KV_ROPE_DIMENSION_SECTIONS, hparams.rope_sections, 4, true); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 28: type = LLM_TYPE_1_7B; break; case 36: type = hparams.n_embd == 2560 ? LLM_TYPE_4B : LLM_TYPE_8B; break; case 64: type = LLM_TYPE_32B; break; diff --git a/src/models/qwen3vlmoe.cpp b/src/models/qwen3vlmoe.cpp index 5b77df571..7c41592f7 100644 --- a/src/models/qwen3vlmoe.cpp +++ b/src/models/qwen3vlmoe.cpp @@ -5,7 +5,8 @@ void llama_model_qwen3vlmoe::load_arch_hparams(llama_model_loader & ml) { ml.get_key_or_arr(LLM_KV_ROPE_DIMENSION_SECTIONS, hparams.rope_sections, 4, true); ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp, false); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 48: type = LLM_TYPE_30B_A3B; break; case 94: type = LLM_TYPE_235B_A22B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/refact.cpp b/src/models/refact.cpp index bf3949a90..a46c358fa 100644 --- a/src/models/refact.cpp +++ b/src/models/refact.cpp @@ -2,7 +2,8 @@ void llama_model_refact::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_1B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/rnd1.cpp b/src/models/rnd1.cpp index ca8e00961..fc276ce59 100644 --- a/src/models/rnd1.cpp +++ b/src/models/rnd1.cpp @@ -2,12 +2,13 @@ void llama_model_rnd1::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp, false); - ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 48: type = LLM_TYPE_30B_A3B; break; default: type = LLM_TYPE_UNKNOWN; } + // Set non-causal attention for diffusion models hparams.causal_attn = false; } diff --git a/src/models/rwkv6.cpp b/src/models/rwkv6.cpp index ba2a9dfa0..0b5013dc7 100644 --- a/src/models/rwkv6.cpp +++ b/src/models/rwkv6.cpp @@ -9,7 +9,7 @@ void llama_model_rwkv6::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_RESCALE_EVERY_N_LAYERS, hparams.rescale_every_n_layers, false); ml.get_key(LLM_KV_TOKEN_SHIFT_COUNT, hparams.token_shift_count, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_1_6B; break; case 32: switch (hparams.n_embd) { diff --git a/src/models/rwkv6qwen2.cpp b/src/models/rwkv6qwen2.cpp index 566b8cdcb..6c7db5144 100644 --- a/src/models/rwkv6qwen2.cpp +++ b/src/models/rwkv6qwen2.cpp @@ -9,7 +9,7 @@ void llama_model_rwkv6qwen2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_RESCALE_EVERY_N_LAYERS, hparams.rescale_every_n_layers, false); ml.get_key(LLM_KV_TOKEN_SHIFT_COUNT, hparams.token_shift_count, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_1_6B; break; case 32: switch (hparams.n_embd) { diff --git a/src/models/rwkv7.cpp b/src/models/rwkv7.cpp index 7574b2526..67c51f5b5 100644 --- a/src/models/rwkv7.cpp +++ b/src/models/rwkv7.cpp @@ -10,7 +10,7 @@ void llama_model_rwkv7::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_GATE_LORA_RANK, hparams.n_lora_gate, false); ml.get_key(LLM_KV_TOKEN_SHIFT_COUNT, hparams.token_shift_count, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 12: switch (hparams.n_embd) { case 768: type = LLM_TYPE_190M; break; diff --git a/src/models/seed-oss.cpp b/src/models/seed-oss.cpp index 806cba574..57de881a0 100644 --- a/src/models/seed-oss.cpp +++ b/src/models/seed-oss.cpp @@ -2,7 +2,8 @@ void llama_model_seed_oss::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 64: type = LLM_TYPE_36B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/smallthinker.cpp b/src/models/smallthinker.cpp index 4231cccc6..a8e3d957f 100644 --- a/src/models/smallthinker.cpp +++ b/src/models/smallthinker.cpp @@ -15,14 +15,14 @@ void llama_model_smallthinker::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false); } else { hparams.swa_type = LLAMA_SWA_TYPE_NONE; - hparams.n_no_rope_layer_step = hparams.n_layer; + hparams.n_no_rope_layer_step = hparams.n_layer(); } ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp, false); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_4B; break; case 52: type = LLM_TYPE_20B; break; default: type = LLM_TYPE_UNKNOWN; diff --git a/src/models/smollm3.cpp b/src/models/smollm3.cpp index 90e7d473e..c67d967b2 100644 --- a/src/models/smollm3.cpp +++ b/src/models/smollm3.cpp @@ -4,7 +4,7 @@ void llama_model_smollm3::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); hparams.n_no_rope_layer_step = 4; - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 36: type = LLM_TYPE_3B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/stablelm.cpp b/src/models/stablelm.cpp index 4da7f7aef..bf6087b87 100644 --- a/src/models/stablelm.cpp +++ b/src/models/stablelm.cpp @@ -3,7 +3,7 @@ void llama_model_stablelm::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_1B; break; case 32: type = LLM_TYPE_3B; break; case 40: type = LLM_TYPE_12B; break; diff --git a/src/models/starcoder.cpp b/src/models/starcoder.cpp index e131af058..f73a88fd4 100644 --- a/src/models/starcoder.cpp +++ b/src/models/starcoder.cpp @@ -2,7 +2,8 @@ void llama_model_starcoder::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 24: type = LLM_TYPE_1B; break; case 36: type = LLM_TYPE_3B; break; case 42: type = LLM_TYPE_7B; break; diff --git a/src/models/starcoder2.cpp b/src/models/starcoder2.cpp index 9c207c028..b81b46937 100644 --- a/src/models/starcoder2.cpp +++ b/src/models/starcoder2.cpp @@ -2,7 +2,8 @@ void llama_model_starcoder2::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 30: type = LLM_TYPE_3B; break; case 32: type = LLM_TYPE_7B; break; case 40: type = LLM_TYPE_15B; break; diff --git a/src/models/step35.cpp b/src/models/step35.cpp index cf9942b20..e2218c587 100644 --- a/src/models/step35.cpp +++ b/src/models/step35.cpp @@ -23,16 +23,16 @@ void llama_model_step35::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa); ml.get_key(LLM_KV_ROPE_FREQ_BASE_SWA, hparams.rope_freq_base_train_swa, false); - ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer); + ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer()); - ml.get_key_or_arr(LLM_KV_SWIGLU_CLAMP_EXP, hparams.swiglu_clamp_exp, hparams.n_layer, false); - ml.get_key_or_arr(LLM_KV_SWIGLU_CLAMP_SHEXP, hparams.swiglu_clamp_shexp, hparams.n_layer, false); + ml.get_key_or_arr(LLM_KV_SWIGLU_CLAMP_EXP, hparams.swiglu_clamp_exp, hparams.n_layer(), false); + ml.get_key_or_arr(LLM_KV_SWIGLU_CLAMP_SHEXP, hparams.swiglu_clamp_shexp, hparams.n_layer(), false); // NextN/MTP (Step3p5): extra decoder block appended beyond the main stack. - ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.nextn_predict_layers, false); - GGML_ASSERT(hparams.nextn_predict_layers < hparams.n_layer && "nextn_predict_layers must be < n_layer"); + ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false); + GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all && "n_layer_nextn must be < n_layer_impl"); - switch (hparams.n_layer - hparams.nextn_predict_layers) { + switch (hparams.n_layer()) { case 45: type = LLM_TYPE_196B_A11B; break; default: type = LLM_TYPE_UNKNOWN; } @@ -41,15 +41,12 @@ void llama_model_step35::load_arch_hparams(llama_model_loader & ml) { void llama_model_step35::load_arch_tensors(llama_model_loader & ml) { LLAMA_LOAD_LOCALS; - const uint32_t n_main = n_layer - hparams.nextn_predict_layers; - const bool mtp_only = (hparams.nextn_predict_layers > 0) && - (ml.get_weight("blk.0.attn_norm.weight") == nullptr); + const bool mtp_only = (hparams.n_layer_nextn > 0) && (ml.get_weight("blk.0.attn_norm.weight") == nullptr); // Trunk-only: the GGUF declares MTP layers in metadata but the actual MTP // tensors live in a separate file (e.g. user split target/draft). Mark // MTP tensors NOT_REQUIRED so the trunk loads cleanly. - const std::string mtp_probe = "blk." + std::to_string(n_main) + ".nextn.eh_proj.weight"; - const bool trunk_only = (hparams.nextn_predict_layers > 0) && - (ml.get_weight(mtp_probe.c_str()) == nullptr); + const std::string mtp_probe = "blk." + std::to_string(n_layer) + ".nextn.eh_proj.weight"; + const bool trunk_only = (hparams.n_layer_nextn > 0) && (ml.get_weight(mtp_probe.c_str()) == nullptr); const int trunk_flags = mtp_only ? TENSOR_NOT_REQUIRED : 0; const int mtp_flags = trunk_only ? TENSOR_NOT_REQUIRED : 0; @@ -176,7 +173,7 @@ void llama_model_step35::load_arch_tensors(llama_model_loader & ml) { layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), { n_embd }, TENSOR_NOT_REQUIRED); }; - for (int i = 0; i < (int) n_main; ++i) { + for (int i = 0; i < n_layer; ++i) { load_block_trunk(i, trunk_flags); } // Only the first MTP block (i == n_main) is required at runtime — the @@ -184,8 +181,8 @@ void llama_model_step35::load_arch_tensors(llama_model_loader & ml) { // Trailing MTP blocks are loaded if present (so an un-pruned GGUF with // all MTP layers still works) but tolerated when absent via the pruning // path. See scripts/prune_step35_extra_mtp.py for the pruner. - for (int i = (int) n_main; i < n_layer; ++i) { - load_block_mtp(i, /*is_first_mtp=*/ i == (int) n_main); + for (int i = n_layer; i < n_layer_all; ++i) { + load_block_mtp(i, /*is_first_mtp=*/ i == n_layer); } } @@ -206,8 +203,7 @@ llama_model_step35::graph::graph(const llama_model & model, const llm_graph_para ggml_tensor * inp_out_ids = build_inp_out_ids(); // MTP/NextN layers are loaded as extra decoder blocks but not executed in the main pass. - const int n_transformer_layers = n_layer - (int) hparams.nextn_predict_layers; - for (int il = 0; il < n_transformer_layers; ++il) { + for (int il = 0; il < n_layer; ++il) { ggml_tensor * inpSA = inpL; const uint32_t n_head_l = hparams.n_head(il); @@ -294,7 +290,7 @@ llama_model_step35::graph::graph(const llama_model & model, const llm_graph_para cb(cur, "attn_proj", il); } - if (il == n_transformer_layers - 1 && inp_out_ids && cparams.embeddings_nextn_masked) { + if (il == n_layer - 1 && inp_out_ids && cparams.embeddings_nextn_masked) { cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } @@ -374,7 +370,7 @@ llama_model_step35::graph::graph(const llama_model & model, const llm_graph_para // LLM_GRAPH_TYPE_DECODER_MTP draft head for Step3p5 (MoE) llama_model_step35::graph_mtp::graph_mtp(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) { - GGML_ASSERT(hparams.nextn_predict_layers > 0 && "STEP35 MTP requires nextn_predict_layers > 0"); + GGML_ASSERT(hparams.n_layer_nextn > 0 && "STEP35 MTP requires n_layer_nextn > 0"); // Single-block MTP only: always run the first trained MTP block (Qwen // MTP / vLLM single-MTP-layer style). Multi-block round-robin proved to @@ -382,7 +378,7 @@ llama_model_step35::graph_mtp::graph_mtp(const llama_model & model, const llm_gr // blocks are loaded with TENSOR_NOT_REQUIRED so pruned GGUFs (with just // block 0) also work — see load_arch_tensors below and // scripts/prune_step35_extra_mtp.py. - const int il = (int) hparams.n_layer - (int) hparams.nextn_predict_layers; + const int il = hparams.n_layer(); const auto & layer = model.layers[il]; GGML_ASSERT(layer.nextn.eh_proj && "MTP block missing nextn.eh_proj"); diff --git a/src/models/t5.cpp b/src/models/t5.cpp index 73e327414..b0e3f0625 100644 --- a/src/models/t5.cpp +++ b/src/models/t5.cpp @@ -9,10 +9,10 @@ void llama_model_t5::load_arch_hparams(llama_model_loader & ml) { hparams.dec_start_token_id = dec_start_token_id; } - hparams.dec_n_layer = hparams.n_layer; + hparams.dec_n_layer = hparams.n_layer(); ml.get_key(LLM_KV_DECODER_BLOCK_COUNT, hparams.dec_n_layer, false); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 6: type = LLM_TYPE_60M; break; // t5-small case 8: type = LLM_TYPE_80M; break; // flan-t5-small case 12: diff --git a/src/models/talkie.cpp b/src/models/talkie.cpp index 1258eeb19..393e8f65b 100644 --- a/src/models/talkie.cpp +++ b/src/models/talkie.cpp @@ -4,7 +4,7 @@ void llama_model_talkie::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); - switch (hparams.n_layer) { + switch (hparams.n_layer()) { case 40: type = LLM_TYPE_13B; break; default: type = LLM_TYPE_UNKNOWN; } diff --git a/src/models/xverse.cpp b/src/models/xverse.cpp index d6d1c7a2e..313500129 100644 --- a/src/models/xverse.cpp +++ b/src/models/xverse.cpp @@ -2,7 +2,8 @@ void llama_model_xverse::load_arch_hparams(llama_model_loader & ml) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - switch (hparams.n_layer) { + + switch (hparams.n_layer()) { case 32: type = LLM_TYPE_7B; break; case 40: type = LLM_TYPE_13B; break; case 80: type = LLM_TYPE_65B; break; diff --git a/tools/quantize/README.md b/tools/quantize/README.md index b8c225124..27384bebf 100644 --- a/tools/quantize/README.md +++ b/tools/quantize/README.md @@ -5,62 +5,87 @@ Quantization reduces the precision of model weights (e.g., from 32-bit floats to This process however, may introduce some accuracy loss which is usually measured in [Perplexity](https://huggingface.co/docs/transformers/en/perplexity) (ppl) and/or [Kullback–Leibler Divergence](https://en.wikipedia.org/wiki/Kullback%E2%80%93Leibler_divergence) (kld). This can be minimized by using a suitable imatrix file. -You can also use the [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space on Hugging Face to build your own quants without any setup. +You can also use the [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space on Hugging Face to build your own quants without any setup. It syncs from llama.cpp `main` every 6 hours. -Note: It is synced from llama.cpp `main` every 6 hours. +## Overview -Example usage: +Quantization is done in two phases: +- Convert the original model to GGUF format. +- Quantize the converted GGUF file. -```./llama-quantize [options] input-model-f32.gguf [output-model-quant.gguf] type [threads]``` +If the model supports multimodal inputs (images or audio), you also need to convert and quantize the multimodal encoders and projectors. + +To perform these tasks, you need to install the Python requirements: ```bash -# from Hugginface, obtain the official meta-llama/Llama-3.1-8B model weights and place them in ./models -ls ./models -config.json model-00001-of-00004.safetensors model-00004-of-00004.safetensors README.md tokenizer.json -generation_config.json model-00002-of-00004.safetensors model.safetensors.index.json special_tokens_map.json USE_POLICY.md -LICENSE model-00003-of-00004.safetensors original tokenizer_config.json - -# [Optional] for PyTorch .bin models like Mistral-7B -ls ./models - - -# install Python dependencies python3 -m pip install -r requirements.txt - -# convert the model to ggml FP16 format -python3 convert_hf_to_gguf.py ./models/mymodel/ - -# quantize the model to 4-bits (using Q4_K_M method) -./llama-quantize ./models/mymodel/ggml-model-f16.gguf ./models/mymodel/ggml-model-Q4_K_M.gguf Q4_K_M - -# update the gguf filetype to current version if older version is now unsupported -./llama-quantize ./models/mymodel/ggml-model-Q4_K_M.gguf ./models/mymodel/ggml-model-Q4_K_M-v2.gguf COPY ``` -Run the quantized model: +Or if you use `uv`: ```bash -# start inference on a gguf model -./llama-cli -m ./models/mymodel/ggml-model-Q4_K_M.gguf -cnv -p "You are a helpful assistant" +uv pip install -r requirements.txt --index-strategy unsafe-best-match ``` +## Prepare the input GGUF file + +To convert a model from a Hugging Face repo, you can use a command like the following: + +``` +python convert_hf_to_gguf.py --outfile gemma-4-E2B-it-bf16.gguf --outtype bf16 --remote google/gemma-4-E2B-it +``` + +Notes: +- In the usual case where the model is distributed in 16-bit format, `--outtype auto` (or omitting `--outtype` entirely) also works well. +- If you have previously downloaded the model locally, specify the directory and remove the `--remote` flag. +- For compatibility reasons, the Python requirements install transformers 4, but more and more models (like Gemma 4) require transformers 5. You can safely `pip install -U transformers` to get the latest version. + +## Quantize the GGUF + +After you have created a high-quality GGUF version of the model, you use `llama-quantize` to apply quantization. For example, quantize to `Q4_K_M` using a command like the following: + +```bash +./build/bin/llama-quantize gemma-4-E2B-it-bf16.gguf gemma-4-E2B-it-Q4_K_M.gguf Q4_K_M +``` + +Various quantization methods are described [later in this document](#quantize). + Options: -* `--allow-requantize` allows requantizing tensors that have already been quantized. Warning: This can severely reduce quality compared to quantizing from 16bit or 32bit -* `--leave-output-tensor` will leave output.weight un(re)quantized. Increases model size but may also increase quality, especially when requantizing -* `--pure` disables k-quant mixtures and quantizes all tensors to the same type -* `--imatrix` uses data in file generated by `llama-imatrix` as importance matrix for quant optimizations (highly recommended) -* `--include-weights` use an importance matrix for tensor(s) in the list. Cannot be used with `--exclude-weights` -* `--exclude-weights` use an importance matrix for tensor(s) in the list. Cannot be used with `--include-weights` +* `--allow-requantize` allow requantizing tensors that have already been quantized. Warning: This can severely reduce quality compared to quantizing from 16bit or 32bit +* `--leave-output-tensor` leave output.weight un(re)quantized. Increases model size but may also increase quality, especially when requantizing +* `--pure` disable k-quant mixtures and quantizes all tensors to the same type +* `--imatrix file_name` use data in file_name as importance matrix for quant optimizations +* `--include-weights tensor_name` use importance matrix for this tensor (can be specified multiple times) +* `--exclude-weights tensor_name` use importance matrix for the tensors **not** specified (include/exclude cannot be mixed) * `--output-tensor-type` use a specific quant type for the output.weight tensor * `--token-embedding-type` use a specific quant type for the token embeddings tensor -* `--keep-split` will generate the quantized model in the same shards as the input file otherwise it will produce a single quantized file +* `--keep-split` generate the quantized model in the same shards as the input file instead of a single quantized file Advanced options: * `--tensor-type` quantize specific tensor(s) to specific quant types. Supports regex syntax. May be specified multiple times. * `--prune-layers` prune (remove) the layers in the list -* `--override-kv` option to override model metadata by key in the quantized model. May be specified multiple times +* `--override-kv` option to override model metadata by key in the quantized model. May be specified multiple times. -Examples: +## (Optional) Convert the multimodal components + +llama.cpp will convert the LLM portion of the source model, which is enough for conversational applications. If the model accepts multimodal inputs and you wish to take advantage of them, you need to create a separate GGUF file. This file is generically known as `mmproj`, for "multimedia projector"; however, it may contain various components such as vision or audio encoders in addition to projections. + +Multimodal components are usually much smaller than the LLMs they come with. In addition, their quality has a direct impact on the quality of LLM generations, because these components are in charge of preparing the inputs for the LLM: the closer inputs are to data seen during training, the better LLM results will be. + +For these reasons, multimodal components are usually kept in a high-quality format such as bf16 or q8. The impact on speed and memory from using a smaller quant is negligible, but overall quality could be impacted. + +```bash +python convert_hf_to_gguf.py --mmproj --outfile mmproj-gemma-4-E2B-it-Q8_0.gguf --outtype q8_0 --remote google/gemma-4-E2B-it +``` + +## Run the quantized model + + +```bash +./build/bin/llama cli -m ./gemma-4-E2B-it-Q4_K_M.gguf --mmproj ./mmproj-gemma-4-E2B-it-Q8_0.gguf --image --prompt "Describe this image" +``` + +## Quantization Examples ```bash # naive Q4_K_M quantization using default settings and 8 CPU threads. Output will be "ggml-model-Q4_K_M.gguf" diff --git a/tools/quantize/quantize.cpp b/tools/quantize/quantize.cpp index 154b2a9ae..6c444e08b 100644 --- a/tools/quantize/quantize.cpp +++ b/tools/quantize/quantize.cpp @@ -1,6 +1,7 @@ #include "llama.h" #include "build-info.h" #include "common/common.h" +#include "common/imatrix-loader.h" #include "gguf.h" @@ -13,7 +14,6 @@ #include #include #include -#include #include #include @@ -77,11 +77,6 @@ static const char * const LLM_KV_QUANTIZE_IMATRIX_DATASET = "quantize.imatrix static const char * const LLM_KV_QUANTIZE_IMATRIX_N_ENTRIES = "quantize.imatrix.entries_count"; static const char * const LLM_KV_QUANTIZE_IMATRIX_N_CHUNKS = "quantize.imatrix.chunks_count"; -// TODO: share with imatrix.cpp -static const char * const LLM_KV_IMATRIX_DATASETS = "imatrix.datasets"; -static const char * const LLM_KV_IMATRIX_CHUNK_COUNT = "imatrix.chunk_count"; -static const char * const LLM_KV_IMATRIX_CHUNK_SIZE = "imatrix.chunk_size"; - static bool striequals(const char * a, const char * b) { while (*a && *b) { if (std::tolower(*a) != std::tolower(*b)) { @@ -180,184 +175,84 @@ static void usage(const char * executable) { exit(1); } -static int load_legacy_imatrix(const std::string & imatrix_file, std::vector & imatrix_datasets, std::unordered_map> & imatrix_data) { - std::ifstream in(imatrix_file.c_str(), std::ios::binary); - if (!in) { - printf("%s: failed to open %s\n",__func__, imatrix_file.c_str()); - exit(1); - } - int n_entries; - in.read((char *)&n_entries, sizeof(n_entries)); - if (in.fail() || n_entries < 1) { - printf("%s: no data in file %s\n", __func__, imatrix_file.c_str()); - exit(1); - } - for (int i = 0; i < n_entries; ++i) { - int len; in.read((char *)&len, sizeof(len)); - std::vector name_as_vec(len+1); - in.read((char *)name_as_vec.data(), len); - if (in.fail()) { - printf("%s: failed reading name for entry %d from %s\n", __func__, i+1, imatrix_file.c_str()); - exit(1); - } - name_as_vec[len] = 0; - std::string name{name_as_vec.data()}; - auto & e = imatrix_data[name]; - int ncall; - in.read((char *)&ncall, sizeof(ncall)); - int nval; - in.read((char *)&nval, sizeof(nval)); - if (in.fail() || nval < 1) { - printf("%s: failed reading number of values for entry %d\n", __func__, i); - imatrix_data = {}; - exit(1); - } - e.resize(nval); - in.read((char *)e.data(), nval*sizeof(float)); - if (in.fail()) { - printf("%s: failed reading data for entry %d\n", __func__, i); - imatrix_data = {}; - exit(1); - } - if (ncall > 0) { - for (auto & v : e) { - v /= ncall; - } - } - - if (getenv("LLAMA_TRACE")) { - printf("%s: loaded data (size = %6d, ncall = %6d) for '%s'\n", __func__, int(e.size()), ncall, name.c_str()); - } - } - - // latest legacy imatrix version contains the dataset filename at the end of the file - int m_last_call = 0; - if (in.peek() != EOF) { - in.read((char *)&m_last_call, sizeof(m_last_call)); - int dataset_len; - in.read((char *)&dataset_len, sizeof(dataset_len)); - std::vector dataset_as_vec(dataset_len); - in.read(dataset_as_vec.data(), dataset_len); - imatrix_datasets.resize(1); - imatrix_datasets[0].assign(dataset_as_vec.begin(), dataset_as_vec.end()); - printf("%s: imatrix dataset='%s'\n", __func__, imatrix_datasets[0].c_str()); - } - printf("%s: loaded %d importance matrix entries from %s computed on %d chunks\n", __func__, int(imatrix_data.size()), imatrix_file.c_str(), m_last_call); - return m_last_call; -} - static int load_imatrix(const std::string & imatrix_file, std::vector & imatrix_datasets, std::unordered_map> & imatrix_data) { - - struct ggml_context * ctx = nullptr; - struct gguf_init_params meta_gguf_params = { - /* .no_alloc = */ false, // the data is needed - /* .ctx = */ &ctx, - }; - struct gguf_context * ctx_gguf = gguf_init_from_file(imatrix_file.c_str(), meta_gguf_params); - if (!ctx_gguf) { - fprintf(stderr, "%s: imatrix file '%s' is using old format\n", __func__, imatrix_file.c_str()); - return load_legacy_imatrix(imatrix_file, imatrix_datasets, imatrix_data); - } - const int32_t n_entries = gguf_get_n_tensors(ctx_gguf); - if (n_entries < 1) { - fprintf(stderr, "%s: no data in file %s\n", __func__, imatrix_file.c_str()); - gguf_free(ctx_gguf); - ggml_free(ctx); + common_imatrix loaded; + if (!common_imatrix_load(imatrix_file, loaded)) { + fprintf(stderr, "%s: failed to load imatrix from '%s'\n", __func__, imatrix_file.c_str()); exit(1); } - const int dataset_idx = gguf_find_key(ctx_gguf, LLM_KV_IMATRIX_DATASETS); - const int chunk_count_idx = gguf_find_key(ctx_gguf, LLM_KV_IMATRIX_CHUNK_COUNT); - const int chunk_size_idx = gguf_find_key(ctx_gguf, LLM_KV_IMATRIX_CHUNK_SIZE); - if (dataset_idx < 0 || chunk_count_idx < 0 || chunk_size_idx < 0) { + if (!loaded.is_legacy && !loaded.has_metadata) { fprintf(stderr, "%s: missing imatrix metadata in file %s\n", __func__, imatrix_file.c_str()); - gguf_free(ctx_gguf); - ggml_free(ctx); exit(1); } - const uint32_t chunk_size = gguf_get_val_u32(ctx_gguf, chunk_size_idx); - - const std::string sums_suffix{ ".in_sum2" }; - const std::string counts_suffix{ ".counts" }; - - // Using an ordered map to get a deterministic iteration order. - std::map> sums_counts_for; - - for (struct ggml_tensor * cur = ggml_get_first_tensor(ctx); cur; cur = ggml_get_next_tensor(ctx, cur)) { - std::string name = cur->name; - - if (name.empty()) { continue; } - - if (string_remove_suffix(name, sums_suffix)) { - // in_sum2 - sums_counts_for[std::move(name)].first = cur; - } else if (string_remove_suffix(name, counts_suffix)) { - // counts - sums_counts_for[std::move(name)].second = cur; - } else { - // ignore other tensors - } - } - - for (const auto & sc : sums_counts_for) { - const std::string & name = sc.first; - const struct ggml_tensor * sums = sc.second.first; - const struct ggml_tensor * counts = sc.second.second; - - if (!sums || !counts) { - fprintf(stderr, "%s: mismatched sums and counts for %s\n", __func__, name.c_str()); - gguf_free(ctx_gguf); - ggml_free(ctx); - exit(1); - } - - const int64_t ne0 = sums->ne[0]; - const int64_t ne1 = sums->ne[1]; - + for (const auto & [name, entry] : loaded.entries) { auto & e = imatrix_data[name]; - e.resize(ggml_nelements(sums)); - float max_count = 0.0f; - for (int64_t j = 0; j < ne1; ++j) { - const float count = ((const float *) counts->data)[j]; - if (count > 0.0f) { - for (int64_t i = 0; i < ne0; ++i) { - e[j*ne0 + i] = ((const float *) sums->data)[j*ne0 + i] / count; + e.resize(entry.sums.size()); + + if (!loaded.is_legacy) { + // GGUF format: normalize by per-expert counts + const int64_t ncounts = entry.counts.size(); + const int64_t ne0 = (int64_t) entry.sums.size() / ncounts; + + for (int64_t j = 0; j < ncounts; ++j) { + const float count = (float) entry.counts[j]; + if (count > 0.0f) { + for (int64_t i = 0; i < ne0; ++i) { + e[j*ne0 + i] = entry.sums[j*ne0 + i] / count; + } + } else { + for (int64_t i = 0; i < ne0; ++i) { + e[j*ne0 + i] = 1; + } + } + } + + if (getenv("LLAMA_TRACE")) { + float max_count = 0.0f; + for (int64_t j = 0; j < ncounts; ++j) { + const float count = (float) entry.counts[j]; + if (count > max_count) { + max_count = count; + } + } + printf("%s: loaded data (size = %6d, n_tokens = %6d, n_chunks = %6d) for '%s'\n", + __func__, int(e.size()), int(max_count), int(max_count / loaded.chunk_size), name.c_str()); + } + } else { + // Legacy format: sums contain (raw/count)*ncall, divide by ncall + const int64_t ncall = entry.counts.empty() ? 0 : entry.counts[0]; + if (ncall > 0) { + for (size_t i = 0; i < entry.sums.size(); ++i) { + e[i] = entry.sums[i] / ncall; } } else { - // Partial imatrix data, this tensor never got any input during calibration - for (int64_t i = 0; i < ne0; ++i) { - e[j*ne0 + i] = 1; + for (size_t i = 0; i < entry.sums.size(); ++i) { + e[i] = entry.sums[i]; } } - if (count > max_count) { - max_count = count; + + if (getenv("LLAMA_TRACE")) { + printf("%s: loaded data (size = %6d, ncall = %6d) for '%s'\n", + __func__, int(e.size()), int(ncall), name.c_str()); } } - if (getenv("LLAMA_TRACE")) { - printf("%s: loaded data (size = %6d, n_tokens = %6d, n_chunks = %6d) for '%s'\n", __func__, int(e.size()), int(max_count), int(max_count / chunk_size), name.c_str()); + } + + imatrix_datasets = std::move(loaded.datasets); + + if (!imatrix_datasets.empty()) { + printf("%s: imatrix datasets=['%s'", __func__, imatrix_datasets[0].c_str()); + for (size_t i = 1; i < imatrix_datasets.size(); ++i) { + printf(", '%s'", imatrix_datasets[i].c_str()); } + printf("]\n"); } - int m_last_chunk = gguf_get_val_u32(ctx_gguf, chunk_count_idx); + printf("%s: loaded %d importance matrix entries from %s computed on %d chunks\n", __func__, int(imatrix_data.size()), imatrix_file.c_str(), loaded.chunk_count); - int64_t n_datasets = gguf_get_arr_n(ctx_gguf, dataset_idx); - imatrix_datasets.reserve(n_datasets); - for (int64_t i = 0; i < n_datasets; ++i) { - imatrix_datasets.push_back(gguf_get_arr_str(ctx_gguf, dataset_idx, i)); - } - printf("%s: imatrix datasets=['%s'", __func__, imatrix_datasets[0].c_str()); - for (size_t i = 1; i < imatrix_datasets.size(); ++i) { - printf(", '%s'", imatrix_datasets[i].c_str()); - } - printf("]\n"); - - printf("%s: loaded %d importance matrix entries from %s computed on %d chunks\n", __func__, int(imatrix_data.size()), imatrix_file.c_str(), m_last_chunk); - - gguf_free(ctx_gguf); - ggml_free(ctx); - - return m_last_chunk; + return loaded.chunk_count; } static int prepare_imatrix(const std::string & imatrix_file, diff --git a/tools/server/server-context.cpp b/tools/server/server-context.cpp index 28f738c3f..ab0d59447 100644 --- a/tools/server/server-context.cpp +++ b/tools/server/server-context.cpp @@ -2512,7 +2512,7 @@ private: llama_memory_seq_pos_max(llama_get_memory(ctx_tgt), slot.id)); if (use_ckpt_dft) { - slot.spec_ckpt.update_dft(ctx_dft.get(), slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY | LLAMA_STATE_SEQ_FLAGS_ON_DEVICE); + slot.spec_ckpt.update_dft(ctx_dft.get(), slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY); } slot.spec_prompt = slot.prompt.tokens.get_text_tokens(); @@ -2551,7 +2551,7 @@ private: if (ctx_dft) { if (use_ckpt_dft) { - ckpt.load_dft(ctx_dft.get(), slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY | LLAMA_STATE_SEQ_FLAGS_ON_DEVICE); + ckpt.load_dft(ctx_dft.get(), slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY); } common_context_seq_rm(ctx_dft.get(), slot.id, ckpt.pos_max + 1, -1); @@ -2568,7 +2568,7 @@ private: if (use_ckpt_tgt) { //const int64_t t_start = ggml_time_us(); - ckpt.update_tgt(ctx_tgt, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY | LLAMA_STATE_SEQ_FLAGS_ON_DEVICE); + ckpt.update_tgt(ctx_tgt, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY); //const int64_t t_total = ggml_time_us() - t_start; //printf("checkpoint total: %f ms\n", t_total / 1000.0); @@ -2580,7 +2580,7 @@ private: } if (use_ckpt_dft) { - ckpt.update_dft(ctx_dft.get(), slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY | LLAMA_STATE_SEQ_FLAGS_ON_DEVICE); + ckpt.update_dft(ctx_dft.get(), slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY); } } } @@ -3447,13 +3447,13 @@ private: SLT_DBG(slot, "restoring speculative checkpoint (pos_min = %d, pos_max = %d, size = %zu)\n", ckpt.pos_min, ckpt.pos_max, ckpt.size()); { - ckpt.load_tgt(slot.ctx_tgt, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY | LLAMA_STATE_SEQ_FLAGS_ON_DEVICE); + ckpt.load_tgt(slot.ctx_tgt, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY); common_context_seq_rm(slot.ctx_tgt, slot.id, ckpt.pos_max + 1, -1); } if (slot.ctx_dft) { - ckpt.load_dft(slot.ctx_dft, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY | LLAMA_STATE_SEQ_FLAGS_ON_DEVICE); + ckpt.load_dft(slot.ctx_dft, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY); common_context_seq_rm(slot.ctx_dft, slot.id, ckpt.pos_max + 1, -1); } diff --git a/tools/ui/.npmrc b/tools/ui/.npmrc index b6f27f135..32e601270 100644 --- a/tools/ui/.npmrc +++ b/tools/ui/.npmrc @@ -1 +1,2 @@ engine-strict=true +ignore-scripts=true diff --git a/tools/ui/src/lib/components/app/actions/ActionIcon.svelte b/tools/ui/src/lib/components/app/actions/ActionIcon.svelte index 849b83b19..f156df669 100644 --- a/tools/ui/src/lib/components/app/actions/ActionIcon.svelte +++ b/tools/ui/src/lib/components/app/actions/ActionIcon.svelte @@ -35,23 +35,27 @@ - + onclick?.(e); + }} + class="h-6 w-6 p-0 {className} flex hover:bg-transparent data-[state=open]:bg-transparent!" + aria-label={ariaLabel || tooltip} + > + {#if icon} + {@const IconComponent = icon} + + {/if} + + {/snippet} diff --git a/tools/ui/src/lib/components/app/badges/BadgeInfo.svelte b/tools/ui/src/lib/components/app/badges/BadgeInfo.svelte index 25986082b..c87c94bc4 100644 --- a/tools/ui/src/lib/components/app/badges/BadgeInfo.svelte +++ b/tools/ui/src/lib/components/app/badges/BadgeInfo.svelte @@ -1,22 +1,22 @@ + {/snippet} + + + +

{opts.tooltipText}

+ + +{/snippet} +
{#if hasPromptStats || isLive} - - - - - - -

Reading (prompt processing)

- - + {@render viewButton({ + view: ChatMessageStatsView.READING, + icon: BookOpenText, + label: 'Reading', + tooltipText: 'Reading (prompt processing)' + })} {/if} - - - - - - -

- {isGenerationDisabled - ? 'Generation (waiting for tokens...)' - : 'Generation (token output)'} -

- - + {@render viewButton({ + view: ChatMessageStatsView.GENERATION, + icon: Sparkles, + label: 'Generation', + tooltipText: isGenerationDisabled + ? 'Generation (waiting for tokens...)' + : 'Generation (token output)', + disabled: isGenerationDisabled + })} {#if hasAgenticStats} - - - - - - -

Tool calls

- - + {@render viewButton({ + view: ChatMessageStatsView.TOOLS, + icon: Wrench, + label: 'Tools', + tooltipText: 'Tool calls' + })} {#if !hideSummary} - - - - - - -

Agentic summary

- - + {@render viewButton({ + view: ChatMessageStatsView.SUMMARY, + icon: Layers, + label: 'Summary', + tooltipText: 'Agentic summary' + })} {/if} {/if}
diff --git a/tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessageStatistics/ChatMessageStatisticsBadge.svelte b/tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessageStatistics/ChatMessageStatisticsBadge.svelte index eea7da7b2..db7d01690 100644 --- a/tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessageStatistics/ChatMessageStatisticsBadge.svelte +++ b/tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessageStatistics/ChatMessageStatisticsBadge.svelte @@ -21,13 +21,16 @@ {#if tooltipLabel} - - {#snippet icon()} - - {/snippet} + + {#snippet child({ props })} + + {#snippet icon()} + + {/snippet} - {value} - + {value} + + {/snippet}

{tooltipLabel}

diff --git a/tools/ui/src/lib/components/app/chat/ChatScreen/ChatScreenActionScrollDown.svelte b/tools/ui/src/lib/components/app/chat/ChatScreen/ChatScreenActionScrollDown.svelte index c43bee3e3..a22c491ad 100644 --- a/tools/ui/src/lib/components/app/chat/ChatScreen/ChatScreenActionScrollDown.svelte +++ b/tools/ui/src/lib/components/app/chat/ChatScreen/ChatScreenActionScrollDown.svelte @@ -41,16 +41,13 @@ }); -
+
+ {/snippet} + - {#if ms.updating} - + {#if selectedOption} + +

{selectedOption.model}

+ {/if} - + {/if} {/if}
diff --git a/tools/ui/src/lib/components/app/navigation/DropdownMenuActions.svelte b/tools/ui/src/lib/components/app/navigation/DropdownMenuActions.svelte index 83d856d10..951831149 100644 --- a/tools/ui/src/lib/components/app/navigation/DropdownMenuActions.svelte +++ b/tools/ui/src/lib/components/app/navigation/DropdownMenuActions.svelte @@ -34,24 +34,28 @@ - e.stopPropagation()} - > - {#if triggerTooltip} - - + + + + {#snippet child({ props })} + e.stopPropagation()} + > {@render iconComponent(triggerIcon, 'h-3 w-3')} - {triggerTooltip} - - -

{triggerTooltip}

- - - {:else} - {@render iconComponent(triggerIcon, 'h-3 w-3')} + {#if triggerTooltip} + {triggerTooltip} + {/if} + + {/snippet} + + {#if triggerTooltip} + +

{triggerTooltip}

+ {/if} - + {#each actions as action, index (action.label)} diff --git a/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigationConversationItem.svelte b/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigationConversationItem.svelte index dad8d954c..e38a93738 100644 --- a/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigationConversationItem.svelte +++ b/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigationConversationItem.svelte @@ -105,6 +105,12 @@ onclick={handleSelect} onmouseover={handleMouseOver} onmouseleave={handleMouseLeave} + onfocusin={handleMouseOver} + onfocusout={(e) => { + if (!e.currentTarget.contains(e.relatedTarget as Node | null)) { + handleMouseLeave(); + } + }} >
0} - - - + + {#snippet child({ props })} + + + + {/snippet} @@ -195,7 +205,8 @@ opacity: 0; } - &:is(:hover) :global([data-slot='dropdown-menu-trigger']) { + &:is(:hover) :global([data-slot='dropdown-menu-trigger']), + &:focus-within :global([data-slot='dropdown-menu-trigger']) { opacity: 1; } @media (max-width: 768px) { diff --git a/tools/ui/tests/stories/a11y/ActionIcon.a11y.stories.svelte b/tools/ui/tests/stories/a11y/ActionIcon.a11y.stories.svelte new file mode 100644 index 000000000..20f5e057b --- /dev/null +++ b/tools/ui/tests/stories/a11y/ActionIcon.a11y.stories.svelte @@ -0,0 +1,34 @@ + + + { + const before = await canvas.findByRole('button', { name: 'before' }); + const target = await canvas.findByRole('button', { name: 'Copy' }); + + before.focus(); + await userEvent.tab(); + + await expect(target).toHaveFocus(); + }} +> +
+ + {}} /> +
+ diff --git a/tools/ui/tests/stories/a11y/ChatMessageStatistics.a11y.stories.svelte b/tools/ui/tests/stories/a11y/ChatMessageStatistics.a11y.stories.svelte new file mode 100644 index 000000000..4aaf60cd6 --- /dev/null +++ b/tools/ui/tests/stories/a11y/ChatMessageStatistics.a11y.stories.svelte @@ -0,0 +1,50 @@ + + + { + const reading = await canvas.findByRole('button', { name: 'Reading' }); + const generation = await canvas.findByRole('button', { name: 'Generation' }); + const tools = await canvas.findByRole('button', { name: 'Tools' }); + const summary = await canvas.findByRole('button', { name: 'Summary' }); + + reading.focus(); + await expect(reading).toHaveFocus(); + + await userEvent.tab(); + await expect(generation).toHaveFocus(); + + await userEvent.tab(); + await expect(tools).toHaveFocus(); + + await userEvent.tab(); + await expect(summary).toHaveFocus(); + }} +/> diff --git a/tools/ui/tests/stories/a11y/HorizontalScrollCarousel.a11y.stories.svelte b/tools/ui/tests/stories/a11y/HorizontalScrollCarousel.a11y.stories.svelte new file mode 100644 index 000000000..937d7ab10 --- /dev/null +++ b/tools/ui/tests/stories/a11y/HorizontalScrollCarousel.a11y.stories.svelte @@ -0,0 +1,69 @@ + + + { + const before = await canvas.findByRole('button', { name: 'before' }); + const after = await canvas.findByRole('button', { name: 'after' }); + const leftArrow = await canvas.findByRole('button', { name: 'Scroll left' }); + + await waitFor(() => { + expect(leftArrow).toBeDisabled(); + }); + + before.focus(); + await userEvent.tab(); + + await expect(after).toHaveFocus(); + }} +> +
+ + +
+
+ + +
+ + + { + const before = await canvas.findByRole('button', { name: 'before' }); + const rightArrow = await canvas.findByRole('button', { name: 'Scroll right' }); + + await waitFor(() => { + expect(rightArrow).not.toBeDisabled(); + }); + + before.focus(); + await userEvent.tab(); + + await expect(rightArrow).toHaveFocus(); + }} +> +
+ + + {#each [...Array(20).keys()] as i (i)} +
{i}
+ {/each} + +
+ diff --git a/tools/ui/tests/stories/a11y/SidebarNavigationConversationItem.a11y.stories.svelte b/tools/ui/tests/stories/a11y/SidebarNavigationConversationItem.a11y.stories.svelte new file mode 100644 index 000000000..1fc42608f --- /dev/null +++ b/tools/ui/tests/stories/a11y/SidebarNavigationConversationItem.a11y.stories.svelte @@ -0,0 +1,36 @@ + + + { + const row = await canvas.findByRole('button', { name: /Forked Conversation/ }); + const forkIcon = await canvas.findByRole('link'); + + row.focus(); + await userEvent.tab(); + + await expect(forkIcon).toHaveFocus(); + }} +/>