mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 32 Packages Projects Releases Wiki Activity

model : support NVFP4 tensors for Gemma4 (#21971)

Browse Source 
* support nvfp4 tensors for Gemma4

* add wo_s to build_attn

* add wo_s to build_attn

* fix glm4
This commit is contained in:
Sigbjørn Skjæret
2026-04-16 16:51:47 +02:00
committed by GitHub
parent b572d1ecd6
commit f772f6e434
105 changed files with 149 additions and 148 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/models/afmoe.cpp
+2 -2
View File
@@ -80,7 +80,7 @@ llm_build_afmoe::llm_build_afmoe(const llama_model & model, const llm_graph_para
Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
cur = build_attn(inp_attn,
NULL, NULL, // wo will be applied after gating
NULL, NULL, NULL, // wo will be applied after gating
Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
cb(cur, "attn_out", il);
@@ -91,7 +91,7 @@ llm_build_afmoe::llm_build_afmoe(const llama_model & model, const llm_graph_para
cb(cur, "attn_gated", il);
// now apply output projection
cur = build_lora_mm(model.layers[il].wo, cur);
cur = build_lora_mm(model.layers[il].wo, cur, model.layers[il].wo_s);
cb(cur, "attn_o_proj", il);
}