happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

convert : fix jamba conv1d shape squeezing

This commit is contained in:
Francis Couture-Harpin 2025-07-02 23:49:12 -04:00
parent 2bcaf64e8e
commit 908e6559d6
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
convert_hf_to_gguf.py
View File

@ -5055,6 +5055,9 @@ class JambaModel(TextModel):
new_name = self.map_tensor_name(name) new_name = self.map_tensor_name(name)
if self.match_model_tensor_name(new_name, gguf.MODEL_TENSOR.SSM_CONV1D, bid):
data_torch = data_torch.squeeze()
if name.endswith(".A_log"): if name.endswith(".A_log"):
logger.debug("A_log --> A ==> " + new_name) logger.debug("A_log --> A ==> " + new_name)
data_torch = -torch.exp(data_torch) data_torch = -torch.exp(data_torch)