fix : checkpoint path

fix : cpu per task

image_ref/config.py

@@ -17,7 +17,7 @@ def load_args_contrastive():
     parser.add_argument('--dataset_val_dir', type=str, default='data/processed_data/npy_image/data_test_contrastive')
     parser.add_argument('--dataset_test_dir', type=str, default=None)
     parser.add_argument('--base_out', type=str, default='output/baseline')
-    parser.add_argument('--dataset_ref_dir', type=str, default='image_ref/img_ref')
+    parser.add_argument('-- ', type=str, default='image_ref/img_ref')
     parser.add_argument('--output', type=str, default='output/out_contrastive.csv')
     parser.add_argument('--save_path', type=str, default='output/best_model_constrastive.pt')
     parser.add_argument('--pretrain_path', type=str, default=None)

image_ref/main_ray.py

@@ -133,7 +133,7 @@ def train_model(config,args):
         # Note to save a file like checkpoint, you still need to put it under a directory
         # to construct a checkpoint.
         with tempfile.TemporaryDirectory(
-                dir='lustre/fswork/projects/rech/bun/ucg81ws/these/pseudo_image/checkpoints') as temp_checkpoint_dir:
+                dir='/lustre/fswork/projects/rech/bun/ucg81ws/these/pseudo_image/checkpoints') as temp_checkpoint_dir:
             path = os.path.join(temp_checkpoint_dir, "checkpoint.pt")
 
             torch.save(
@@ -212,7 +212,7 @@ def test_model(best_result, args):
 def main(args, gpus_per_trial=1):
     config = {
         "lr": tune.loguniform(1e-4, 1e-2),
-        "noise": tune.loguniform(1e-7, 500),
+        "noise": tune.loguniform(1, 1000),
         "positive_prop": tune.uniform(0, 100),
         "optimizer": tune.choice(['Adam', 'SGD']),
         "sampler": tune.choice(['random', 'balanced']),
@@ -228,7 +228,7 @@ def main(args, gpus_per_trial=1):
     tuner = tune.Tuner(
         tune.with_resources(
             tune.with_parameters(train_model, args=args),
-            resources={"cpu": 80, "gpu": gpus_per_trial}
+            resources={"cpu": 20, "gpu": gpus_per_trial}
         ),
         tune_config=tune.TuneConfig(
             time_budget_s=3600 * 23.5,