mirror of
https://github.com/hiyouga/LLaMA-Factory.git
synced 2025-10-14 23:58:11 +08:00
fix export format
Former-commit-id: 7c82bd396b9e6ff395850ad544d95cbf1b7557cd
This commit is contained in:
parent
53d7c5109f
commit
af3f5b6e16
@ -758,6 +758,26 @@ register_template(
|
||||
)
|
||||
|
||||
|
||||
register_template(
|
||||
name="yuan",
|
||||
prefix=[
|
||||
"{{system}}"
|
||||
],
|
||||
prompt=[
|
||||
"{{query}}",
|
||||
{"token": "<sep>"}
|
||||
],
|
||||
system="",
|
||||
sep=[
|
||||
"\n"
|
||||
],
|
||||
stop_words=[
|
||||
"<eod>"
|
||||
],
|
||||
replace_eos=True
|
||||
)
|
||||
|
||||
|
||||
register_template(
|
||||
name="zephyr",
|
||||
prefix=[
|
||||
|
@ -102,6 +102,10 @@ class ModelArguments:
|
||||
default=1024,
|
||||
metadata={"help": "The maximum length of the model inputs used for quantization."}
|
||||
)
|
||||
export_lecacy_format: Optional[bool] = field(
|
||||
default=False,
|
||||
metadata={"help": "Whether or not to save the `.bin` files instead of `.safetensors`."}
|
||||
)
|
||||
|
||||
def __post_init__(self):
|
||||
self.compute_dtype = None
|
||||
|
@ -47,7 +47,11 @@ def export_model(args: Optional[Dict[str, Any]] = None):
|
||||
|
||||
model.config.use_cache = True
|
||||
model = model.to("cpu")
|
||||
model.save_pretrained(model_args.export_dir, max_shard_size="{}GB".format(model_args.export_size))
|
||||
model.save_pretrained(
|
||||
save_directory=model_args.export_dir,
|
||||
max_shard_size="{}GB".format(model_args.export_size),
|
||||
safe_serialization=(not model_args.export_lecacy_format)
|
||||
)
|
||||
|
||||
try:
|
||||
tokenizer.padding_side = "left" # restore padding side
|
||||
|
Loading…
x
Reference in New Issue
Block a user