darthrubik/llama-cpp
1
0
Fork 0
Code Issues Pull Requests 1 Packages Projects Releases Wiki Activity

dflash : load optional FC input scale (#30945)
Copilot Setup Steps / copilot-setup-steps (push) Has been cancelled
Check Pre-Tokenizer Hashes / pre-tokenizer-hashes (push) Has been cancelled
Python check requirements.txt / check-requirements (push) Has been cancelled
Python Type-Check / python type-check (push) Has been cancelled
Update Operations Documentation / update-ops-docs (push) Has been cancelled

Browse Source 
Assisted-by: Codex
This commit is contained in:
Hrishith Thadicherla
2026-10-22 10:54:12 +05:30
committed by GitHub
parent 84746c5c94
commit 4912ac24f1
2 changed files with 4 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-model.h
+3 -2
View File
@@ -779,8 +779,9 @@ struct llama_model {
struct ggml_tensor * per_layer_proj_norm = nullptr;
// eagle3 / dflash feature fusion layer
struct ggml_tensor * fc = nullptr;
struct ggml_tensor * fc_s = nullptr;
struct ggml_tensor * fc = nullptr;
struct ggml_tensor * fc_s = nullptr;
struct ggml_tensor * fc_in_s = nullptr;
struct ggml_tensor * d2t = nullptr; // draft to target vocabulary mapping
// dspark
src/models/dflash.cpp
+1
View File
@@ -238,6 +238,7 @@ void llama_model_dflash::load_arch_tensors(llama_model_loader &) {
fc = create_tensor(tn(LLM_TENSOR_FC, "weight"), { n_embd_inp, n_embd }, 0);
fc_s = create_tensor(tn(LLM_TENSOR_FC, "scale"), { 1 }, TENSOR_NOT_REQUIRED);
fc_in_s = create_tensor(tn(LLM_TENSOR_FC, "input_scale"), { 1 }, TENSOR_NOT_REQUIRED);
output_norm_enc = create_tensor(tn(LLM_TENSOR_ENC_OUTPUT_NORM, "weight"), { n_embd }, 0); // encoder hidden_norm (after fc)
output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0); // decoder final norm