convert : add --fuse-qkv flag to fuse Q/K/V into QKV during HF-to-GGUF conversion (#22780)
This commit is contained in:
1 parent
5fdfa62829
commit
465e49b9ce
20 files changed
+396
-125
No files matched your search
@@ -36,12 +36,7 @@ void llama_model_llada::load_arch_tensors(llama_model_loader &) {
|
||||
|
||||
layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, 0);
|
||||
|
||||
// Use separate Q, K, V projections without bias, matching LLaDALlamaBlock
|
||||
layer.wq =
|
||||
create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), { n_embd, n_embd_head_k * n_head }, 0);
|
||||
layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K, "weight", i), { n_embd, n_embd_k_gqa }, 0);
|
||||
layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V, "weight", i), { n_embd, n_embd_v_gqa }, 0);
|
||||
// No bias for QKV projections as per config: include_bias=false, include_qkv_bias=false
|
||||
create_tensor_qkv(layer, i, n_embd, n_embd_head_k * n_head, n_embd_k_gqa, n_embd_v_gqa, 0);
|
||||
layer.wo =
|
||||
create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, 0);
|
||||
layer.wo_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), { n_embd }, TENSOR_NOT_REQUIRED);
|
||||
|
||||
Reference in new issue
Block a user