kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-14 06:27:45 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

fix: GPT-Next convert failure (#3220)

Browse Source 
Signed-off-by: Enwei Zhu <21126786+syuoni@users.noreply.github.com>
This commit is contained in:
Enwei Zhu 2025-04-02 17:14:39 +08:00 committed by GitHub
parent e64c565750
commit d3948cd9b2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
examples/gpt/convert_checkpoint.py
View File

@ -180,9 +180,9 @@ def args_to_quant_config(args: argparse.Namespace) -> QuantConfig:
quant_config.kv_cache_quant_algo = QuantAlgo.INT8
# Check if model ckpt is pre-quantized to fp8.
hf_quant_config_file = Path(args.model_dir) / "hf_quant_config.json"
if hf_quant_config_file.exists():
with open(hf_quant_config_file, 'r') as f:
hf_quant_config_path = f"{args.model_dir}/hf_quant_config.json"
if os.path.exists(hf_quant_config_path):
with open(hf_quant_config_path, 'r') as f:
hf_quant_config = json.load(f)
if hf_quant_config.get("producer", {}).get("name") == "modelopt":
modelopt_quant_config = hf_quant_config.get("quantization", {})