Create training_args.json

training_args.json

@@ -0,0 +1,50 @@
+{
+  "output_dir": "./helion-v1.5-checkpoints",
+  "overwrite_output_dir": true,
+  "do_train": true,
+  "do_eval": true,
+  "evaluation_strategy": "steps",
+  "eval_steps": 500,
+  "per_device_train_batch_size": 4,
+  "per_device_eval_batch_size": 4,
+  "gradient_accumulation_steps": 8,
+  "learning_rate": 2e-05,
+  "weight_decay": 0.01,
+  "num_train_epochs": 3,
+  "lr_scheduler_type": "cosine",
+  "warmup_steps": 100,
+  "logging_dir": "./logs",
+  "logging_steps": 10,
+  "save_strategy": "steps",
+  "save_steps": 500,
+  "save_total_limit": 3,
+  "fp16": false,
+  "bf16": true,
+  "max_grad_norm": 1.0,
+  "optim": "adamw_torch",
+  "group_by_length": true,
+  "length_column_name": "length",
+  "report_to": [
+    "tensorboard",
+    "wandb"
+  ],
+  "load_best_model_at_end": true,
+  "metric_for_best_model": "eval_loss",
+  "greater_is_better": false,
+  "gradient_checkpointing": true,
+  "gradient_checkpointing_kwargs": {
+    "use_reentrant": false
+  },
+  "dataloader_num_workers": 4,
+  "dataloader_pin_memory": true,
+  "ddp_find_unused_parameters": false,
+  "torch_compile": false,
+  "max_steps": -1,
+  "save_safetensors": true,
+  "push_to_hub": false,
+  "hub_model_id": "DeepXR/Helion-V1.5",
+  "hub_strategy": "every_save",
+  "hub_token": null,
+  "auto_find_batch_size": false,
+  "include_inputs_for_metrics": false
+}