mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 28 Packages Projects Releases Wiki Activity

models : move the token embedding norms to the first layer (#20943)

Browse Source 
* models : move the token embedding norms to the first layer

* cont : fix LLM_TENSOR_CONV1D + fix il indexing
This commit is contained in:
Georgi Gerganov
2026-03-24 17:00:30 +02:00
committed by GitHub
parent 3fc6f1aed1
commit 9f102a1407
8 changed files with 26 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/models/modern-bert.cpp
+2 -2
View File
@@ -15,8 +15,8 @@ llm_build_modern_bert::llm_build_modern_bert(const llama_model & model, const ll
cb(inpL, "inp_embd", -1);
// embed layer norm
inpL = build_norm(inpL, model.tok_norm, nullptr, LLM_NORM, -1);
cb(inpL, "inp_norm", -1);
inpL = build_norm(inpL, model.tok_norm, nullptr, LLM_NORM, 0);
cb(inpL, "inp_norm", 0);
ggml_tensor * inp_out_ids = build_inp_out_ids();