vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-11 17:44:38 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

implment vision model architecture, gguf convertor

Browse source
This commit is contained in:
HimariO 2025-02-02 18:02:07 +08:00
parent 558a764713
commit d9f2d71bc2
3 changed files with 219 additions and 83 deletions
Download patch file Download diff file Expand all files Collapse all files

5
examples/llava/clip-impl.h
View file

@ -26,6 +26,8 @@
#define KEY_HAS_QWEN2VL_MERGER "clip.has_qwen2vl_merger"
#define KEY_USE_GELU "clip.use_gelu"
#define KEY_USE_SILU "clip.use_silu"
#define KEY_USE_GLU_MLP "clip.use_glu_mlp"
#define KEY_USE_RMS_NORM "clip.use_rms_norm"
#define KEY_N_EMBD "clip.%s.embedding_length"
#define KEY_N_FF "clip.%s.feed_forward_length"
#define KEY_N_BLOCK "clip.%s.block_count"
@ -44,6 +46,8 @@
#define KEY_MM_PATCH_MERGE_TYPE "clip.vision.mm_patch_merge_type"
#define KEY_IMAGE_GRID_PINPOINTS "clip.vision.image_grid_pinpoints"
#define KEY_IMAGE_CROP_RESOLUTION "clip.vision.image_crop_resolution"
#define KEY_FULLATTN_BLK_IDX "clip.vision.fullatt_block_indexes"
#define KEY_ATTN_WINDOW_SIZE "clip.vision.window_size"
//
@ -62,6 +66,7 @@
#define TN_FFN_DOWN "%s.blk.%d.ffn_down.%s"
#define TN_FFN_GATE "%s.blk.%d.ffn_gate.%s"
#define TN_FFN_UP "%s.blk.%d.ffn_up.%s"
#define TN_FFN_GATE "%s.blk.%d.ffn_gate.%s"
#define TN_LN_1 "%s.blk.%d.ln1.%s"
#define TN_LN_2 "%s.blk.%d.ln2.%s"
#define TN_LN_PRE "%s.pre_ln.%s"

99
examples/llava/clip.cpp
View file

@ -169,6 +169,7 @@ struct clip_hparams {
std::vector<int32_t> image_grid_pinpoints;
int32_t image_crop_resolution;
std::unordered_set<int32_t> vision_feature_layer;
std::vector<int32_t> full_attn_layers;
};
struct clip_layer {
@ -200,6 +201,9 @@ struct clip_layer {
struct ggml_tensor * ff_down_w = nullptr;
struct ggml_tensor * ff_down_b = nullptr;
struct ggml_tensor * ff_g_w = NULL;
struct ggml_tensor * ff_g_b = NULL;
// layernorm 2
struct ggml_tensor * ln_2_w = nullptr;
struct ggml_tensor * ln_2_b = nullptr;
@ -324,6 +328,9 @@ struct clip_ctx {
float image_std[3];
bool use_gelu = false;
bool use_silu = false;
bool use_glu_mlp = false;
bool use_rms_norm = false;
int32_t ftype = 1;
gguf_context_ptr ctx_gguf;
ggml_context_ptr ctx_data;
@ -810,6 +817,7 @@ static ggml_cgraph * clip_image_build_graph_legacy(clip_ctx * ctx, const clip_im
const int n_head = hparams.n_head;
const int d_head = hidden_size / n_head;
const float eps = hparams.eps;
const bool use_window_attn = hparams.full_attn_layers.size() > 0;
int mrope_sections[4] = {d_head/4, d_head/4, d_head/4, d_head/4};
const int batch_size = imgs.entries.size();
@ -864,6 +872,8 @@ static ggml_cgraph * clip_image_build_graph_legacy(clip_ctx * ctx, const clip_im
}
struct ggml_tensor * embeddings = inp;
struct ggml_tensor * pos_embed = nullptr;
struct ggml_tensor * window_mask = nullptr;
struct ggml_tensor * window_idx = nullptr;
if (ctx->has_llava_projector) {
// concat class_embeddings and patch_embeddings
@ -915,6 +925,28 @@ static ggml_cgraph * clip_image_build_graph_legacy(clip_ctx * ctx, const clip_im
const auto & vision_feature_layer = hparams.vision_feature_layer;
// loop over layers
if (use_window_attn) {
window_idx = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, num_positions / 4);
ggml_set_name(window_idx, "window_idx");
ggml_set_input(window_idx);
// mask for window attention
window_mask = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, num_positions, num_positions);
ggml_set_name(window_mask, "window_mask");
ggml_set_input(window_mask);
// embeddings shape: [hidden_size, patches_w * patches_h, batch_size]
GGML_ASSERT(batch_size == 1);
embeddings = ggml_reshape_2d(ctx0, embeddings, hidden_size * 4, patches_w * patches_h * batch_size / 4);
embeddings = ggml_get_rows(ctx0, embeddings, window_idx);
embeddings = ggml_reshape_3d(ctx0, embeddings, hidden_size, patches_w * patches_h, batch_size);
positions = ggml_reshape_2d(ctx0, positions, 16, num_position_ids / 4 / 4);
positions = ggml_get_rows(ctx0, positions, window_idx);
positions = ggml_reshape_1d(ctx0, positions, num_position_ids);
}
for (int il = 0; il < ctx->max_feature_layer; il++) {
struct ggml_tensor * cur = embeddings; // embeddings = residual, cur = hidden_states
@ -927,9 +959,12 @@ static ggml_cgraph * clip_image_build_graph_legacy(clip_ctx * ctx, const clip_im
//const size_t nb_q_w = model.layers[il].q_w->nb[0];
// layernorm1
{
if (ctx->use_rms_norm) {
cur = ggml_rms_norm(ctx0, cur, eps);
cur = ggml_mul(ctx0, cur, model.layers[il].ln_1_w);
}
else {
cur = ggml_norm(ctx0, cur, eps);
cur = ggml_add(ctx0, ggml_mul(ctx0, cur, model.layers[il].ln_1_w),
model.layers[il].ln_1_b);
}
@ -969,7 +1004,14 @@ static ggml_cgraph * clip_image_build_graph_legacy(clip_ctx * ctx, const clip_im
V = ggml_reshape_3d(ctx0, V, num_positions, d_head, n_head * batch_size);
struct ggml_tensor * KQ = ggml_mul_mat(ctx0, K, Q);
const bool inlist = std::find(hparams.full_attn_layers.begin(), hparams.full_attn_layers.end(), il) != hparams.full_attn_layers.end();
const bool full_attn = use_window_attn ? inlist : true;
if (full_attn) {
KQ = ggml_soft_max_ext(ctx0, KQ, nullptr, 1.0f / sqrtf((float)d_head), 0.0f);
} else {
KQ = ggml_soft_max_ext(ctx0, KQ, window_mask, 1.0f, 0.0f);
}
struct ggml_tensor * KQV = ggml_mul_mat(ctx0, V, KQ);
KQV = ggml_reshape_4d(ctx0, KQV, d_head, num_positions, n_head, batch_size);
KQV = ggml_permute(ctx0, KQV, 0, 2, 1, 3);
@ -986,12 +1028,36 @@ static ggml_cgraph * clip_image_build_graph_legacy(clip_ctx * ctx, const clip_im
embeddings = cur; // embeddings = residual, cur = hidden_states
// layernorm2
{
if (ctx->use_rms_norm) {
cur = ggml_rms_norm(ctx0, cur, eps);
cur = ggml_mul(ctx0, cur, model.layers[il].ln_2_w);
} else {
cur = ggml_norm(ctx0, cur, eps);
cur = ggml_add(ctx0, ggml_mul(ctx0, cur, model.layers[il].ln_2_w), model.layers[il].ln_2_b);
}
// mlp
if (ctx->use_glu_mlp) {
// ffn_up
auto cur_up = ggml_mul_mat(ctx0, model.layers[il].ff_o_w, cur);
cur_up = ggml_add(ctx0, cur_up, model.layers[il].ff_o_b);
auto cur_gate = ggml_mul_mat(ctx0, model.layers[il].ff_g_w, cur);
cur_gate = ggml_add(ctx0, cur_gate, model.layers[il].ff_g_b);
if (ctx->use_gelu) {
cur_gate = ggml_gelu_inplace(ctx0, cur_gate);
} else if (ctx->use_silu) {
cur_gate = ggml_silu_inplace(ctx0, cur_gate);
} else {
cur_gate = ggml_gelu_quick_inplace(ctx0, cur_gate);
}
cur = ggml_mul(ctx0, cur_gate, cur_up);
// ffn_down
cur = ggml_mul_mat(ctx0, model.layers[il].ff_i_w, cur);
cur = ggml_add(ctx0, cur, model.layers[il].ff_i_b);
}
else {
cur = ggml_mul_mat(ctx0, model.layers[il].ff_i_w, cur);
cur = ggml_add(ctx0, cur, model.layers[il].ff_i_b);
@ -1005,6 +1071,7 @@ static ggml_cgraph * clip_image_build_graph_legacy(clip_ctx * ctx, const clip_im
cur = ggml_mul_mat(ctx0, model.layers[il].ff_o_w, cur);
cur = ggml_add(ctx0, cur, model.layers[il].ff_o_b);
}
// residual 2
cur = ggml_add(ctx0, embeddings, cur);
@ -1014,11 +1081,18 @@ static ggml_cgraph * clip_image_build_graph_legacy(clip_ctx * ctx, const clip_im
// post-layernorm
if (model.post_ln_w) {
if (ctx->use_rms_norm) {
embeddings = ggml_rms_norm(ctx0, embeddings, eps);
ggml_set_name(embeddings, "post_ln");
embeddings = ggml_mul(ctx0, embeddings, model.post_ln_w);
} else {
embeddings = ggml_norm(ctx0, embeddings, eps);
ggml_set_name(embeddings, "post_ln");
embeddings = ggml_add(ctx0, ggml_mul(ctx0, embeddings, model.post_ln_w), model.post_ln_b);
}
}
// final layer is a vision feature layer
if (vision_feature_layer.find(ctx->max_feature_layer) != vision_feature_layer.end()) {
@ -1331,6 +1405,18 @@ static ggml_cgraph * clip_image_build_graph_legacy(clip_ctx * ctx, const clip_im
embeddings = ggml_add(ctx0, embeddings, model.mm_1_b);
}
if (use_window_attn) {
struct ggml_tensor * inv_window_idx = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, num_positions / 4);
ggml_set_name(inv_window_idx, "inv_window_idx");
ggml_set_input(inv_window_idx);
// embeddings shape: [hidden_size, patches_w * patches_h, batch_size]
GGML_ASSERT(batch_size == 1);
embeddings = ggml_reshape_2d(ctx0, embeddings, hparams.projection_dim, patches_w * patches_h / 4);
embeddings = ggml_get_rows(ctx0, embeddings, inv_window_idx);
embeddings = ggml_reshape_3d(ctx0, embeddings, hparams.projection_dim, patches_w * patches_h / 4, batch_size);
}
// build the graph
ggml_build_forward_expand(gf, embeddings);
@ -1447,6 +1533,8 @@ struct clip_model_loader {
get_bool(KEY_USE_GELU, ctx_clip.use_gelu, false);
get_bool(KEY_USE_SILU, ctx_clip.use_silu, false);
get_bool(KEY_USE_GLU_MLP, ctx_clip.use_glu_mlp, false);
get_bool(KEY_USE_RMS_NORM, ctx_clip.use_rms_norm, false);
get_u32(string_format(KEY_N_EMBD, "vision"), hparams.hidden_size);
get_u32(string_format(KEY_N_HEAD, "vision"), hparams.n_head);
@ -1458,6 +1546,7 @@ struct clip_model_loader {
get_u32(KEY_PATCH_SIZE, hparams.patch_size);
get_u32(KEY_IMAGE_CROP_RESOLUTION, hparams.image_crop_resolution, false);
get_arr_int(KEY_IMAGE_GRID_PINPOINTS, hparams.image_grid_pinpoints, false);
get_arr_int(KEY_FULLATTN_BLK_IDX, hparams.full_attn_layers, false);
{
std::string mm_patch_merge_type;
@ -1603,8 +1692,10 @@ struct clip_model_loader {
// legacy naming (the in and out is reversed! don't ask me why)
layer.ff_i_w = layer.ff_down_w;
layer.ff_o_w = layer.ff_up_w;
layer.ff_g_w = layer.ff_gate_b;
layer.ff_i_b = layer.ff_down_b;
layer.ff_o_b = layer.ff_up_b;
layer.ff_g_b = layer.ff_gate_b;
}
switch (ctx_clip.proj_type) {

66
examples/llava/qwen2_vl_surgery.py
View file

@ -5,10 +5,12 @@ import torch
import numpy as np
from gguf import *
from transformers import (
Qwen2VLForConditionalGeneration,
Qwen2VLProcessor,
AutoProcessor,
Qwen2VLConfig
Qwen2VLForConditionalGeneration,
Qwen2_5_VLForConditionalGeneration,
Qwen2VLProcessor,
Qwen2VLConfig,
Qwen2_5_VLConfig,
)
@ -18,7 +20,9 @@ VISION = "clip.vision"
def k(raw_key: str, arch: str) -> str:
return raw_key.format(arch=arch)
class VL2:
@staticmethod
def to_gguf_name(name: str) -> str:
og = name
name = name.replace("text_model", "t").replace("vision_model", "v")
@ -31,8 +35,8 @@ def to_gguf_name(name: str) -> str:
print(f"[to_gguf_name] {og} --> {name}")
return name
def find_vision_tensors(qwen2vl, dtype) -> Dict[str, np.ndarray]:
@classmethod
def find_vision_tensors(cls, qwen2vl, dtype) -> Dict[str, np.ndarray]:
vision_model = qwen2vl.visual
tensor_map = {}
for name, ten in vision_model.state_dict().items():
@ -47,9 +51,9 @@ def find_vision_tensors(qwen2vl, dtype) -> Dict[str, np.ndarray]:
wq = ten[:c]
wk = ten[c: c * 2]
wv = ten[c * 2:]
tensor_map[to_gguf_name(f"vision_model.{name}").replace("qkv", "q")] = wq
tensor_map[to_gguf_name(f"vision_model.{name}").replace("qkv", "k")] = wk
tensor_map[to_gguf_name(f"vision_model.{name}").replace("qkv", "v")] = wv
tensor_map[cls.to_gguf_name(f"vision_model.{name}").replace("qkv", "q")] = wq
tensor_map[cls.to_gguf_name(f"vision_model.{name}").replace("qkv", "k")] = wk
tensor_map[cls.to_gguf_name(f"vision_model.{name}").replace("qkv", "v")] = wv
elif 'merger' in name:
if name.endswith("ln_q.weight"):
tensor_map['v.post_ln.weight'] = ten
@ -57,7 +61,7 @@ def find_vision_tensors(qwen2vl, dtype) -> Dict[str, np.ndarray]:
tensor_map['v.post_ln.bias'] = ten
else:
# "merger.mlp.%d.weight/bias" --> "mm.%d.weight/bias"
tensor_map[to_gguf_name(name)] = ten
tensor_map[cls.to_gguf_name(name)] = ten
elif 'patch_embed.proj.weight' in name:
# NOTE: split Conv3D into Conv2Ds
c1, c2, kt, kh, kw = ten.shape
@ -65,7 +69,7 @@ def find_vision_tensors(qwen2vl, dtype) -> Dict[str, np.ndarray]:
tensor_map["v.patch_embd.weight"] = ten[:, :, 0, ...]
tensor_map["v.patch_embd.weight.1"] = ten[:, :, 1, ...]
else:
tensor_map[to_gguf_name(f"vision_model.{name}")] = ten
tensor_map[cls.to_gguf_name(f"vision_model.{name}")] = ten
for new_name, ten in tensor_map.items():
if ten.ndim <= 1 or new_name.endswith("_norm.weight"):
@ -76,6 +80,22 @@ def find_vision_tensors(qwen2vl, dtype) -> Dict[str, np.ndarray]:
return tensor_map
class VL25(VL2):
@staticmethod
def to_gguf_name(name: str) -> str:
og = name
name = name.replace("text_model", "t").replace("vision_model", "v")
name = name.replace("blocks", "blk").replace("embeddings.", "")
name = name.replace("attn.", "attn_")
name = name.replace("mlp.down_proj", "ffn_down").replace("mlp.up_proj", "ffn_up")
name = name.replace("mlp.gate_proj", "ffn_gate").replace("proj.", "out.")
name = name.replace("norm1", "ln1").replace("norm2", "ln2")
name = name.replace("merger.mlp", 'mm')
print(f"[vl25][to_gguf_name] {og} --> {name}")
return name
def main(args):
if args.data_type == 'fp32':
dtype = torch.float32
@ -92,11 +112,18 @@ def main(args):
model_path = ""
model_name = args.model_name
print("model_name: ", model_name)
if args.model_type == "qwen2vl":
qwen2vl = Qwen2VLForConditionalGeneration.from_pretrained(
model_name, torch_dtype=dtype, device_map="cpu"
)
cfg: Qwen2VLConfig = qwen2vl.config # type: ignore[reportAssignmentType]
vcfg = cfg.vision_config
else:
qwen2vl = Qwen2_5_VLForConditionalGeneration.from_pretrained(
model_name, torch_dtype=dtype, device_map="cpu"
)
cfg: Qwen2_5_VLConfig = qwen2vl.config # type: ignore[reportAssignmentType]
vcfg = cfg.vision_config
if os.path.isdir(model_name):
local_model = True
@ -125,14 +152,26 @@ def main(args):
else:
raise ValueError()
tensor_map = find_vision_tensors(qwen2vl, np_dtype)
if args.model_type == "qwen2.5vl":
fout.add_bool("clip.use_glu_mlp", True) # gate linear unit MLP layer in vision model
fout.add_bool("clip.use_rms_norm", True)
fout.add_array("clip.vision.fullatt_block_indexes", vcfg.fullatt_block_indexes)
fout.add_uint32("clip.vision.window_size", vcfg.window_size)
fout.add_uint32(k(KEY_EMBEDDING_LENGTH, VISION), vcfg.hidden_size)
fout.add_uint32("clip.vision.projection_dim", vcfg.out_hidden_size)
else:
fout.add_uint32(k(KEY_EMBEDDING_LENGTH, VISION), vcfg.embed_dim)
fout.add_uint32("clip.vision.projection_dim", vcfg.hidden_size)
if args.model_type == "qwen2.5vl":
tensor_map = VL25.find_vision_tensors(qwen2vl, np_dtype)
else:
tensor_map = VL2.find_vision_tensors(qwen2vl, np_dtype)
for name, data in tensor_map.items():
fout.add_tensor(name, data)
fout.add_uint32("clip.vision.patch_size", vcfg.patch_size)
fout.add_uint32("clip.vision.image_size", 14 * 40) # some reasonable size that is divable by (14*2)
fout.add_uint32(k(KEY_EMBEDDING_LENGTH, VISION), vcfg.embed_dim)
fout.add_uint32("clip.vision.projection_dim", vcfg.hidden_size)
fout.add_uint32(k(KEY_ATTENTION_HEAD_COUNT, VISION), vcfg.num_heads)
fout.add_float32(k(KEY_ATTENTION_LAYERNORM_EPS, VISION), 1e-6)
fout.add_uint32(k(KEY_BLOCK_COUNT, VISION), vcfg.depth)
@ -160,6 +199,7 @@ def main(args):
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("model_name", nargs='?', default="Qwen/Qwen2-VL-2B-Instruct")
parser.add_argument("--model_type", nargs='?', choices=['qwen2vl', 'qwen2.5vl'], default="qwen2vl")
parser.add_argument("--data_type", nargs='?', choices=['fp32', 'fp16'], default="fp32")
args = parser.parse_args()
main(args)