Update app.py

app.py

@@ -14,29 +14,6 @@ import pandas as pd
 from sklearn.metrics import f1_score
 import seaborn as sns
 
-# Set a fixed random seed for reproducibility
-seed = 42
-np.random.seed(seed)
-torch.manual_seed(seed)  # Ensure PyTorch random seed for CPU
-
-# If running on GPU, set the seed for CUDA as well
-if torch.cuda.is_available():
-    torch.cuda.manual_seed(seed)
-    torch.cuda.manual_seed_all(seed)
-    torch.backends.cudnn.deterministic = True
-    torch.backends.cudnn.benchmark = False
-
-# Enable deterministic algorithms in PyTorch (slower, but ensures consistency)
-torch.use_deterministic_algorithms(True)
-
-# Limit the number of threads to prevent non-deterministic results from multithreading
-torch.set_num_threads(1)
-os.environ['MKL_NUM_THREADS'] = '1'
-os.environ['OMP_NUM_THREADS'] = '1'
-
-# Optional: Use for debugging to ensure intermediate values match across devices
-torch.set_printoptions(precision=10)
-
 #################### BEAM PREDICTION #########################}
 def beam_prediction_task(data_percentage, task_complexity):
     # Folder naming convention based on input_type, data_percentage, and task_complexity
@@ -438,8 +415,7 @@ def process_hdf5_file(uploaded_file, percentage):
         # Step 4: Load the model from lwm_model module
         device = 'cuda' if torch.cuda.is_available() else 'cpu'
         print(f"Loading the LWM model on {device}...")
-        model = lwm_model.LWM.from_pretrained(device=device)
-        model = model.float()
+        model = lwm_model.LWM.from_pretrained(device=device).float()
         #for name, param in model.state_dict().items():
         #    print(f"Layer: {name} | Weights: {param}")
 
@@ -454,7 +430,7 @@ def process_hdf5_file(uploaded_file, percentage):
         #print(preprocessed_chs[0][0][-1]) #CORRECT
 
         # Step 7: Perform inference using the functions from inference.py
-        output_emb = inference.lwm_inference(preprocessed_chs, 'cls_emb', model)
+        output_emb = inference.lwm_inference(preprocessed_chs, 'cls_emb', model, device)
         output_raw = inference.create_raw_dataset(preprocessed_chs, device)
 
         print(f"Output Embeddings Shape: {output_emb.shape}")