Datasets:

CodeIsAbstract
/

sanskrit-sandhi-boundaries-v2

Tasks:

Token Classification

Modalities:

Formats:

Languages:

Size:

Tags:

boundary-detection

Libraries:

License:

Dataset card Data Studio Files Files and versions

CodeIsAbstract commited on Aug 25

Commit

78a2ce3

·

verified ·

1 Parent(s): 823f46d

Update train_split_model_cuda.py

Files changed (1) hide show

train_split_model_cuda.py +0 -1

train_split_model_cuda.py CHANGED Viewed

@@ -371,7 +371,6 @@ def main():
         per_device_eval_batch_size=args.batch,
         gradient_accumulation_steps=args.grad_accum,
         learning_rate=args.lr,
-        warmup_ratio=0.05,
         lr_scheduler_type="cosine",
         weight_decay=0.01,
         max_grad_norm=0.5,

         per_device_eval_batch_size=args.batch,
         gradient_accumulation_steps=args.grad_accum,
         learning_rate=args.lr,
         lr_scheduler_type="cosine",
         weight_decay=0.01,
         max_grad_norm=0.5,