Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests 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/llama-graph.h
+5
View File
@@ -892,6 +892,7 @@ struct llm_graph_context {
llm_graph_input_attn_no_cache * inp,
ggml_tensor * wo,
ggml_tensor * wo_b,
ggml_tensor * wo_s,
ggml_tensor * q_cur, // [n_embd_head_q, n_head_q, n_tokens]
ggml_tensor * k_cur, // [n_embd_head_k, n_head_k, n_tokens]
ggml_tensor * v_cur, // [n_embd_head_v, n_head_v, n_tokens]
@@ -907,6 +908,7 @@ struct llm_graph_context {
llm_graph_input_attn_kv * inp,
ggml_tensor * wo,
ggml_tensor * wo_b,
ggml_tensor * wo_s,
ggml_tensor * q_cur, // [n_embd_head_q, n_head_q, n_tokens]
ggml_tensor * k_cur, // [n_embd_head_k, n_head_k, n_tokens]
ggml_tensor * v_cur, // [n_embd_head_v, n_head_v, n_tokens]
@@ -922,6 +924,7 @@ struct llm_graph_context {
llm_graph_input_attn_k * inp,
ggml_tensor * wo,
ggml_tensor * wo_b,
ggml_tensor * wo_s,
ggml_tensor * q_cur, // [n_embd_head_q, n_head_q, n_tokens]
ggml_tensor * k_cur, // [n_embd_head_k, n_head_k, n_tokens]
ggml_tensor * v_cur, // [n_embd_head_v, n_head_v, n_tokens]
@@ -938,6 +941,7 @@ struct llm_graph_context {
llm_graph_input_attn_kv_iswa * inp,
ggml_tensor * wo,
ggml_tensor * wo_b,
ggml_tensor * wo_s,
ggml_tensor * q_cur, // [n_embd_head_q, n_head_q, n_tokens]
ggml_tensor * k_cur, // [n_embd_head_k, n_head_k, n_tokens] optional
ggml_tensor * v_cur, // [n_embd_head_v, n_head_v, n_tokens] optional
@@ -953,6 +957,7 @@ struct llm_graph_context {
llm_graph_input_attn_cross * inp,
ggml_tensor * wo,
ggml_tensor * wo_b,
ggml_tensor * wo_s,
ggml_tensor * q_cur, // [n_embd_head_q, n_head_q, n_tokens]
ggml_tensor * k_cur, // [n_embd_head_k, n_head_k, n_tokens]
ggml_tensor * v_cur, // [n_embd_head_v, n_head_v, n_tokens]