423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-29 02:00:36 +08:00
Code Issues Packages Projects Releases Wiki Activity

[misc] fix ci with uv (#9676)

Browse Source
This commit is contained in:
Yaowei Zheng
2025-12-27 01:39:13 +08:00
committed by GitHub
parent a1b1931b4a
commit 55590f5ece
22 changed files with 118 additions and 121 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
src/llamafactory/model/patcher.py
View File

@@ -156,16 +156,13 @@ def patch_config(
# deepspeed zero3 is not compatible with low_cpu_mem_usage
init_kwargs["low_cpu_mem_usage"] = model_args.low_cpu_mem_usage and (not is_deepspeed_zero3_enabled())
# do not cast data type of the model deepspeed zero3 without qlora
if not (is_deepspeed_zero3_enabled() and model_args.quantization_bit is None):
init_kwargs["torch_dtype"] = "auto"
# fsdp/deepspeed zero3 does not need device map
if not (is_deepspeed_zero3_enabled() or is_fsdp_enabled()) and init_kwargs["low_cpu_mem_usage"]:
if "device_map" not in init_kwargs and model_args.device_map:
init_kwargs["device_map"] = model_args.device_map # device map requires low_cpu_mem_usage=True
if init_kwargs["low_cpu_mem_usage"] and not is_fsdp_enabled(): # fsdp does not need device map
if "device_map" not in init_kwargs and model_args.device_map:
init_kwargs["device_map"] = model_args.device_map # device map requires low_cpu_mem_usage=True
if init_kwargs.get("device_map", None) == "auto":
init_kwargs["offload_folder"] = model_args.offload_folder
if init_kwargs.get("device_map", None) == "auto":
init_kwargs["offload_folder"] = model_args.offload_folder
def patch_model(