Enable gradient_checkpointing
Browse files- codeparrot_training.py +1 -1
codeparrot_training.py
CHANGED
@@ -133,7 +133,7 @@ config = {"train_batch_size": 1,
|
|
133 |
"lr_scheduler_type": "cosine",
|
134 |
"num_warmup_steps": 750,
|
135 |
"gradient_accumulation_steps": 16,
|
136 |
-
"gradient_checkpointing":
|
137 |
"max_train_steps": 50_000,
|
138 |
"max_eval_steps": -1,
|
139 |
"seq_length": 1024,
|
|
|
133 |
"lr_scheduler_type": "cosine",
|
134 |
"num_warmup_steps": 750,
|
135 |
"gradient_accumulation_steps": 16,
|
136 |
+
"gradient_checkpointing": True,
|
137 |
"max_train_steps": 50_000,
|
138 |
"max_eval_steps": -1,
|
139 |
"seq_length": 1024,
|