fix Trackio crash: push at end only
Browse files- train_sft.py +6 -6
train_sft.py
CHANGED
|
@@ -45,18 +45,18 @@ def main() -> None:
|
|
| 45 |
logging_steps=5,
|
| 46 |
eval_strategy="steps",
|
| 47 |
eval_steps=20,
|
| 48 |
-
|
|
|
|
| 49 |
max_length=1024,
|
| 50 |
bf16=True,
|
| 51 |
-
push_to_hub=
|
| 52 |
hub_model_id=HUB_MODEL_ID,
|
| 53 |
-
report_to="
|
| 54 |
-
project="cpp-compiler-rl",
|
| 55 |
-
run_name="sft-qwen25-1.5b",
|
| 56 |
),
|
| 57 |
)
|
| 58 |
trainer.train()
|
| 59 |
-
trainer.push_to_hub()
|
|
|
|
| 60 |
print(f"Pushed to {HUB_MODEL_ID}")
|
| 61 |
|
| 62 |
|
|
|
|
| 45 |
logging_steps=5,
|
| 46 |
eval_strategy="steps",
|
| 47 |
eval_steps=20,
|
| 48 |
+
save_strategy="epoch",
|
| 49 |
+
save_total_limit=1,
|
| 50 |
max_length=1024,
|
| 51 |
bf16=True,
|
| 52 |
+
push_to_hub=False, # push once at end (avoid Trackio parquet bug mid-save)
|
| 53 |
hub_model_id=HUB_MODEL_ID,
|
| 54 |
+
report_to="none",
|
|
|
|
|
|
|
| 55 |
),
|
| 56 |
)
|
| 57 |
trainer.train()
|
| 58 |
+
trainer.model.push_to_hub(HUB_MODEL_ID, private=False)
|
| 59 |
+
trainer.processing_class.push_to_hub(HUB_MODEL_ID, private=False)
|
| 60 |
print(f"Pushed to {HUB_MODEL_ID}")
|
| 61 |
|
| 62 |
|