ViuAI commited on
Commit
bb3bb0d
·
verified ·
1 Parent(s): dda27ab

Upload code/train_sft.py with huggingface_hub

Browse files
Files changed (1) hide show
  1. code/train_sft.py +13 -10
code/train_sft.py CHANGED
@@ -563,21 +563,24 @@ def main():
563
 
564
  root_dir = os.path.abspath(os.path.join(cur_dir, ".."))
565
  data_dir = args.data_dir or os.path.join(root_dir, "tokenized_data")
566
- stage_subfolder = "ViuTranslate" if str(args.version).lower() == "viutranslate" else f"sft_{args.version}"
 
 
 
 
 
 
 
 
 
567
  stage_data_dir = os.path.join(data_dir, stage_subfolder)
568
 
569
- output_dir = args.output_dir or (
570
- os.path.join(root_dir, "sft_checkpoints", "ViuTranslate")
571
- if str(args.version).lower() == "viutranslate"
572
- else os.path.join(root_dir, "sft_checkpoints", f"sft_{args.version}")
573
- )
574
  os.makedirs(output_dir, exist_ok=True)
575
 
576
- ckpt_filename = "viutranslate_final.pt" if str(args.version).lower() == "viutranslate" else f"sft_{args.version}_final.pt"
577
  save_path = os.path.join(output_dir, ckpt_filename)
578
- # Best-checkpoint is kept in a SEPARATE file so later (worse) evals can never
579
- # overwrite the best weights. Previously both shared save_path.
580
- best_filename = "viutranslate_best.pt" if str(args.version).lower() == "viutranslate" else f"sft_{args.version}_best.pt"
581
  best_save_path = os.path.join(output_dir, best_filename)
582
 
583
  if args.init_ckpt:
 
563
 
564
  root_dir = os.path.abspath(os.path.join(cur_dir, ".."))
565
  data_dir = args.data_dir or os.path.join(root_dir, "tokenized_data")
566
+ v_clean = str(args.version).lower().strip()
567
+ if v_clean in ("500m_sarus_sft", "sarus_sft", "500m"):
568
+ stage_subfolder = "500M_Sarus_SFT"
569
+ ckpt_tag = "500m_sarus_sft"
570
+ elif v_clean == "viutranslate":
571
+ stage_subfolder = "ViuTranslate"
572
+ ckpt_tag = "viutranslate"
573
+ else:
574
+ stage_subfolder = f"sft_{args.version}"
575
+ ckpt_tag = f"sft_{args.version}"
576
  stage_data_dir = os.path.join(data_dir, stage_subfolder)
577
 
578
+ output_dir = args.output_dir or os.path.join(root_dir, "sft_checkpoints", stage_subfolder)
 
 
 
 
579
  os.makedirs(output_dir, exist_ok=True)
580
 
581
+ ckpt_filename = f"{ckpt_tag}_final.pt"
582
  save_path = os.path.join(output_dir, ckpt_filename)
583
+ best_filename = f"{ckpt_tag}_best.pt"
 
 
584
  best_save_path = os.path.join(output_dir, best_filename)
585
 
586
  if args.init_ckpt: