Skip to content

Fix offload false #2054

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Oct 17, 2025
Merged

Fix offload false #2054

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
20b53aa
alreary convert_gptq_v2_to_v1_format_module
LRL2-ModelCloud Oct 17, 2025
c344d26
cleanup
LRL2-ModelCloud Oct 17, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 4 additions & 16 deletions gptqmodel/models/writer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -225,20 +225,8 @@ def save_quantized(
f"Using 'format = {FORMAT.GPTQ_V2}': the serialized model is only supported by GPTQModel version >= {MIN_VERSION_WITH_V2}."
)

if not self.load_quantized_model:
model = self.model
# internal is always gptq v2 but allow users to pass gptq (v1) via config
if (
quantize_config.format == FORMAT.GPTQ
and quantize_config.quant_method == METHOD.GPTQ
and self.qlinear_kernel.REQUIRES_FORMAT_V2
):
# Model qzeros may be edited in place for export compatibility.
model = convert_gptq_v2_to_v1_format(
model, quantize_config=quantize_config, qlinear_kernel=self.qlinear_kernel
)
else:
model = self.get_model_with_quantize(
if self.load_quantized_model:
self.model = self.get_model_with_quantize(
qcfg=quantize_config,
model_id_or_path=self.model_local_path,
)
Expand Down Expand Up @@ -286,10 +274,10 @@ def debug_saved_config(path):

# Due to shell/turtle state, we need to sync the modules from turtle to shell
if not self.load_quantized_model:
alias_all_from_turtle_if_meta(shell_model=model, turtle_model=self.turtle_model)
alias_all_from_turtle_if_meta(shell_model=self.model, turtle_model=self.turtle_model)

offload_root = self.quantize_config.offload_to_disk_path if getattr(self.quantize_config, "offload_to_disk", False) else None
state_dict = get_state_dict_for_save(model, offload_root=offload_root)
state_dict = get_state_dict_for_save(self.model, offload_root=offload_root)

model_base_name = "model"
model_save_name = model_base_name + ".safetensors"
Expand Down