nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 19 Packages Projects Releases Wiki Activity

revert : "[Model] Qwen3.5 dense and MoE support (no vision) (#19435)" (#19453)

Browse Source 
This reverts commit 39bf692af1.
This commit is contained in:
Georgi Gerganov
2026-02-09 14:57:51 +02:00
committed by GitHub
parent f5e7734ff2
commit 972f323e73
14 changed files with 399 additions and 1532 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gguf-py/gguf/tensor_mapping.py
+2 -4
View File
@@ -228,7 +228,6 @@ class TensorNameMap:
"transformer_encoder.{bid}.qkv", # neobert
"layers.{bid}.attn.Wqkv", # modern-bert
"model.layers.{bid}.self_attn.language_expert_query_key_value", # cogvlm
"model.layers.{bid}.linear_attn.in_proj_qkv", # qwen3.5
),
# Attention query
@@ -359,9 +358,8 @@ class TensorNameMap:
),
MODEL_TENSOR.ATTN_GATE: (
"model.layers.{bid}.self_attn.gate_proj", # afmoe
"model.layers.{bid}.self_attn.g_proj", # step3.5 head-wise attention gate
"model.layers.{bid}.linear_attn.in_proj_z", # qwen3.5
"model.layers.{bid}.self_attn.gate_proj", # afmoe
"model.layers.{bid}.self_attn.g_proj", # step3.5 head-wise attention gate
),
# Feed-forward norm