CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

gguf-py : fix some metadata name extraction edge cases

Browse Source 
* convert_lora : use the lora dir for the model card path
This commit is contained in:
Francis Couture-Harpin
2024-07-19 12:30:37 -04:00
parent 87e397d00b
commit 2164c9deb3
4 changed files with 57 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
convert_hf_to_gguf.py
View File

@@ -62,6 +62,7 @@ class Model:
gguf_writer: gguf.GGUFWriter
model_name: str | None
metadata_override: Path | None
dir_model_card: Path
# subclasses should define this!
model_arch: gguf.MODEL_ARCH
@@ -90,6 +91,7 @@ class Model:
self.tensor_names = None
self.metadata_override = metadata_override
self.model_name = model_name
self.dir_model_card = dir_model # overridden in convert_lora_to_gguf.py
# Apply heuristics to figure out typical tensor encoding based on first layer tensor encoding type
if self.ftype == gguf.LlamaFileType.GUESSED:
@@ -345,7 +347,7 @@ class Model:
total_params, shared_params, expert_params, expert_count = self.gguf_writer.get_total_parameter_count()
self.metadata = gguf.Metadata.load(self.metadata_override, self.dir_model, self.model_name, total_params)
self.metadata = gguf.Metadata.load(self.metadata_override, self.dir_model, self.model_name, self.dir_model_card, total_params)
# Fallback to model directory name if metadata name is still missing
if self.metadata.name is None: