Pushed by DataDreamer

Update training_args.json

training_args.json

@@ -6,11 +6,11 @@
     "do_predict": false,
     "evaluation_strategy": "steps",
     "prediction_loss_only": false,
-    "per_device_train_batch_size": 128,
-    "per_device_eval_batch_size": 128,
+    "per_device_train_batch_size": 64,
+    "per_device_eval_batch_size": 64,
     "per_gpu_train_batch_size": null,
     "per_gpu_eval_batch_size": null,
-    "gradient_accumulation_steps": 1,
+    "gradient_accumulation_steps": 2,
     "eval_accumulation_steps": 1,
     "eval_delay": 0,
     "learning_rate": 0.0001,
@@ -19,16 +19,16 @@
     "adam_beta2": 0.999,
     "adam_epsilon": 1e-08,
     "max_grad_norm": 1.0,
-    "num_train_epochs": 3,
+    "num_train_epochs": 2,
     "max_steps": -1,
-    "lr_scheduler_type": "linear",
+    "lr_scheduler_type": "constant",
     "lr_scheduler_kwargs": {},
     "warmup_ratio": 0.0,
     "warmup_steps": 0,
     "log_level": "passive",
     "log_level_replica": "warning",
     "log_on_each_node": true,
-    "logging_dir": "./output/train-styledistance-model/_checkpoints/runs/Nov21_07-04-08_nlpgpu04.seas.upenn.edu",
+    "logging_dir": "./output/train-styledistance-model/_checkpoints/runs/Nov29_20-48-36_nlpgpu10.seas.upenn.edu",
     "logging_strategy": "steps",
     "logging_first_step": false,
     "logging_steps": 10,