Pushed by DataDreamer

Update training_args.json

training_args.json

@@ -10,16 +10,16 @@
     "per_device_eval_batch_size": 16,
     "per_gpu_train_batch_size": null,
     "per_gpu_eval_batch_size": null,
-    "gradient_accumulation_steps": 8,
+    "gradient_accumulation_steps": 4,
     "eval_accumulation_steps": 1,
     "eval_delay": 0,
-    "learning_rate": 0.001,
+    "learning_rate": 0.0001,
     "weight_decay": 0.01,
     "adam_beta1": 0.9,
     "adam_beta2": 0.999,
     "adam_epsilon": 1e-08,
     "max_grad_norm": 1.0,
-    "num_train_epochs": 20,
+    "num_train_epochs": 10,
     "max_steps": -1,
     "lr_scheduler_type": "linear",
     "lr_scheduler_kwargs": {},
@@ -28,7 +28,7 @@
     "log_level": "passive",
     "log_level_replica": "warning",
     "log_on_each_node": true,
-    "logging_dir": "./output/train-wegmann--styledistance-model/_checkpoints/runs/Jul21_05-54-20_nlpgpu04.seas.upenn.edu",
+    "logging_dir": "./output/train-wegmann--styledistance-model/_checkpoints/runs/Jul23_14-22-58_nlpgpu04.seas.upenn.edu",
     "logging_strategy": "steps",
     "logging_first_step": false,
     "logging_steps": 1,