Skip to content

Remove quantization_config in config.json from original deepseek models #753

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

Open
Edwardf0t1 wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Remove quantization_config in config.json from original deepseek models #753

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
30c6110
remove quantization_config in config.json from original deepseek models
Edwardf0t1 Jan 9, 2026
52a2230
simplify
Edwardf0t1 Jan 9, 2026
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
16 changes: 16 additions & 0 deletions examples/deepseek/quantize_to_nvfp4.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,6 +82,20 @@ def _remap_key(key_dict: dict[str, Any]):
key_dict.update(new_dict)


def remove_quantization_config_from_original_config(export_dir: str) -> None:
"""Remove `quantization_config` from exported HF `config.json`.

Assumes the exported checkpoint directory has a `config.json` containing `quantization_config`.
"""
config_path = os.path.join(export_dir, "config.json")
with open(config_path) as f:
cfg = json.load(f)
del cfg["quantization_config"]
with open(config_path, "w") as f:
json.dump(cfg, f, indent=2, sort_keys=True)
f.write("\n")


def load_and_preprocess_state_dict(modelopt_state_root, world_size=8):
state_dict_list = [
torch.load(f"{modelopt_state_root}/amax_dict_rank{rank}-mp{world_size}.pt")
Expand Down Expand Up @@ -302,3 +316,5 @@ def get_tensor(tensor_name):
save_root=args.fp4_path,
per_layer_quant_config=per_layer_quant_config,
)

remove_quantization_config_from_original_config(args.fp4_path)