vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

update convert-baichuan limit of authority

Browse source
This commit is contained in:
jameswu2014 2023-09-11 13:31:48 +08:00
parent 306d71be42
commit fa41aeb8ba
1 changed files with 1 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

3
convert-baichuan-hf-to-gguf.py Normal file → Executable file
View file

@ -1,5 +1,5 @@
#!/usr/bin/env python3 #!/usr/bin/env python3
# HF llama --> gguf conversion # HF baichuan --> gguf conversion
from __future__ import annotations from __future__ import annotations
@ -23,7 +23,6 @@ if TYPE_CHECKING:
NDArray: TypeAlias = 'np.ndarray[Any, Any]' NDArray: TypeAlias = 'np.ndarray[Any, Any]'
# reverse HF permute back to original pth layout # reverse HF permute back to original pth layout
# https://github.com/huggingface/transformers/blob/main/src/transformers/models/llama/convert_llama_weights_to_hf.py
def reverse_hf_permute(weights: NDArray, n_head: int, n_kv_head: int | None = None) -> NDArray: def reverse_hf_permute(weights: NDArray, n_head: int, n_kv_head: int | None = None) -> NDArray: