vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-05-05 23:41:45 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

model : add JAIS-2 architecture support (#19488)

Browse source 
* model: add JAIS-2 architecture support

Add support for the JAIS-2 family of Arabic-English bilingual models
from Inception AI (https://huggingface.co/inceptionai/Jais-2-8B-Chat).

Architecture characteristics:
- LayerNorm (not RMSNorm) with biases
- ReLU² (ReLU squared) activation function
- Separate Q/K/V projections with biases
- Simple MLP without gate projection (up -> act -> down)
- RoPE positional embeddings
- GPT-2 BPE tokenizer

Supported model sizes:
- Jais-2-8B (32 layers, 26 heads, 3328 hidden)
- Jais-2-70B (68 layers, 56 heads, 7168 hidden)

Tested with quantizations: BF16, Q8_0, Q6_K, Q5_K_M, Q5_0, Q4_K_M, Q4_0, Q3_K_M, Q2_K

Note: JAIS-2 requires F32 precision accumulators for numerical stability
and uses standard attention (not flash attention) on CUDA backends.

* fix: run convert_hf_to_gguf_update.py for jais-2 tokenizer hash

* fix: use NEOX RoPE type for JAIS2

* fix: remove Q/K permutation (NEOX RoPE doesn't need it)

* fix: enable flash attention for JAIS2 (fixed by #19115)

* fix: add dedicated JAIS2 pre-tokenizer type and control vector support

- Add LLAMA_VOCAB_PRE_TYPE_JAIS2 with cascading whitespace regex
- Include original regex from tokenizer.json as comment
- Add build_cvec call for control vector support

* no longer necessary to override set_vocab

---------

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
This commit is contained in:
3 a l i 2026-02-19 16:30:17 +04:00 • committed by GitHub
parent c78e682245
commit 2bf318fd2f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
12 changed files with 248 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

11
src/llama-graph.cpp
View file

@ -1128,8 +1128,8 @@ ggml_tensor * llm_graph_context::build_ffn(
if (down) {
cur = build_lora_mm(down, cur);
if (arch == LLM_ARCH_GLM4 || arch == LLM_ARCH_GLM4_MOE) {
// GLM4 and GLM4_MOE seem to have numerical issues with half-precision accumulators
if (arch == LLM_ARCH_GLM4 || arch == LLM_ARCH_GLM4_MOE || arch == LLM_ARCH_JAIS2) {
// GLM4, GLM4_MOE, and JAIS2 seem to have numerical issues with half-precision accumulators
ggml_mul_mat_set_prec(cur, GGML_PREC_F32);
}
}
@ -1724,7 +1724,8 @@ ggml_tensor * llm_graph_context::build_attn_mha(
ggml_tensor * cur;
if (cparams.flash_attn && kq_b == nullptr) {
const bool use_flash_attn = cparams.flash_attn && kq_b == nullptr;
if (use_flash_attn) {
GGML_ASSERT(kq_b == nullptr && "Flash attention does not support KQ bias yet");
if (v_trans) {
@ -1984,8 +1985,8 @@ ggml_tensor * llm_graph_context::build_attn(
if (wo) {
cur = build_lora_mm(wo, cur);
if (arch == LLM_ARCH_GLM4 || arch == LLM_ARCH_GLM4_MOE) {
// GLM4 and GLM4_MOE seem to have numerical issues with half-precision accumulators
if (arch == LLM_ARCH_GLM4 || arch == LLM_ARCH_GLM4_MOE || arch == LLM_ARCH_JAIS2) {
// GLM4, GLM4_MOE, and JAIS2 seem to have numerical issues with half-precision accumulators
ggml_mul_mat_set_prec(cur, GGML_PREC_F32);
}
}