dev

app.py

@@ -6,6 +6,7 @@ from datasets import load_dataset, concatenate_datasets, load_from_disk
 import traceback
 from sklearn.metrics import accuracy_score
 import numpy as np
+import torch
 
 import os
 from huggingface_hub import login
@@ -16,6 +17,7 @@ os.environ['HF_HOME'] = '/data/.huggingface'
 @spaces.GPU(duration=120)
 def fine_tune_model(model_name, dataset_name, hub_id, api_key, num_epochs, batch_size, lr, grad):
     try:
+        torch.cuda.empty_cache()
         def compute_metrics(eval_pred):
             logits, labels = eval_pred
             predictions = np.argmax(logits, axis=1)
@@ -34,6 +36,7 @@ def fine_tune_model(model_name, dataset_name, hub_id, api_key, num_epochs, batch
     
         # Load the model and tokenizer
         model = AutoModelForSeq2SeqLM.from_pretrained(model_name.strip(), num_labels=2)
+        model.gradient_checkpointing_enable()
         #model = get_peft_model(model, lora_config)
         
     
@@ -91,7 +94,7 @@ def fine_tune_model(model_name, dataset_name, hub_id, api_key, num_epochs, batch
                 model_inputs = tokenizer(
                     examples['text'], 
                     max_length=max_length,  # Set to None for dynamic padding
-                    padding='max_length',     # Disable padding here, we will handle it later
+                    padding=True,     # Disable padding here, we will handle it later
                     truncation=True,
                 )
             
@@ -99,7 +102,7 @@ def fine_tune_model(model_name, dataset_name, hub_id, api_key, num_epochs, batch
                 labels = tokenizer(
                     examples['target'], 
                     max_length=max_length,  # Set to None for dynamic padding
-                    padding='max_length',     # Disable padding here, we will handle it later
+                    padding=True,     # Disable padding here, we will handle it later
                     truncation=True,
                     text_target=examples['target']  # Use text_target for target text
                 )
@@ -132,10 +135,12 @@ def fine_tune_model(model_name, dataset_name, hub_id, api_key, num_epochs, batch
 '''
 # Define Gradio interface
 def predict(text):
+    model = AutoModelForSeq2SeqLM.from_pretrained(model_name.strip(), num_labels=2)
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
     inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)
     outputs = model(inputs)
     predictions = outputs.logits.argmax(dim=-1)
-    return "Positive" if predictions.item() == 1 else "Negative"
+    return predictions.item()
 '''
 # Create Gradio interface
 try:
@@ -156,6 +161,17 @@ try:
         title="Fine-Tune Hugging Face Model",
         description="This interface allows you to fine-tune a Hugging Face model on a specified dataset."
     )
+    '''
+    iface = gr.Interface(
+        fn=predict,
+        inputs=[
+            gr.Textbox(label="Query"),
+        ],
+        outputs="text",
+        title="Fine-Tune Hugging Face Model",
+        description="This interface allows you to test a fine-tune Hugging Face model."
+    )
+    '''
     # Launch the interface
     iface.launch()    
 except Exception as e:

requirements.txt

@@ -5,3 +5,4 @@ peft
 huggingface_hub
 scikit-learn
 numpy
+import torch