biondizzle/deepseek-v4-quant
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix: use hf_ptq.py arg names (--pyt_ckpt_path, --qformat, --inference_tensor_parallel)

Browse Source
This commit is contained in:
biondizzle
2026-05-09 14:57:28 +00:00
parent 1a36a655ea
commit a300302486
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
scripts/quantize_nvfp4.py
View File

@@ -212,12 +212,12 @@ def run_calibration(model_path, export_dir, calib_save_path, amax_snapshot_path,
saved_argv = sys.argv
sys.argv = [
"hf_ptq.py",
"--model", model_path,
"--quant", QUANT,
"--pyt_ckpt_path", model_path,
"--qformat", QUANT,
"--calib_size", str(calib_size),
"--calib_seq", str(calib_seq),
"--kv_cache_qformat", KV_CACHE_QUANT,
"--tp", str(TP),
"--inference_tensor_parallel", str(TP),
"--export_path", export_dir,
"--trust_remote_code",
"--use_seq_device_map",