Update app.py

app.py

@@ -104,17 +104,11 @@ def finetune_model(csv_file, sample_size=100, num_epochs=3, progress=gr.Progress
     # Load tokenizer
     progress(0.2, "Loading tokenizer...")
     
-    # Try the non-gated Mistral model first
+    # Use only the original Mistral model
     model_name = "mistralai/Mistral-7B-v0.1"
-    try:
-        tokenizer = AutoTokenizer.from_pretrained(model_name, token=hf_token)
-        print(f"Successfully loaded tokenizer for {model_name}")
-    except Exception as e:
-        print(f"Error loading {model_name}: {e}")
-        print("Falling back to original Mistral model with token authentication...")
-        model_name = "mistralai/Mistral-7B-v0.1"
-        tokenizer = AutoTokenizer.from_pretrained(model_name, token=hf_token)
+    print(f"Using model: {model_name}")
     
+    tokenizer = AutoTokenizer.from_pretrained(model_name, token=hf_token)
     tokenizer.pad_token = tokenizer.eos_token
     
     # Tokenize dataset
@@ -262,10 +256,10 @@ with gr.Blocks() as demo:
             # Check if we can access Mistral model
             try:
                 from huggingface_hub import model_info
-                info = model_info(" {model_name}", token=hf_token)
-                status.append(f"✅ Access to Mistral model verified: {info.modelId}")
+                info = model_info("mistralai/Mistral-7B-v0.1", token=hf_token)
+                status.append(f"✅ Access to Mistral-7B-v0.1 model verified: {info.modelId}")
             except Exception as e:
-                status.append(f"❌ Cannot access Mistral model: {str(e)}")
+                status.append(f"❌ Cannot access Mistral-7B-v0.1 model: {str(e)}")
             
             return "\n".join(status)
         
@@ -291,12 +285,12 @@ with gr.Blocks() as demo:
         gr.Markdown("""
         ## Fine-Tuning Mistral 7B for Student Bot
         
-        This app fine-tunes the Mistral 7B model to respond like a student to teacher prompts.
+        This app fine-tunes the original Mistral-7B-v0.1 model to respond like a student to teacher prompts.
         
         ### Requirements
         - CSV file with teacher-student conversation pairs
         - GPU acceleration (provided by this Space)
-        - Hugging Face authentication for accessing gated models
+        - Hugging Face authentication for accessing Mistral-7B-v0.1 (which is a gated model)
         
         ### Process
         1. Upload your CSV file
@@ -305,9 +299,8 @@ with gr.Blocks() as demo:
         4. View test results with sample prompts
         
         ### Important Notes
-        - The app tries to use Mistral-7B-Instruct-v0.2 which is not gated
-        - If that fails, it falls back to the original Mistral-7B-v0.1 model (which requires authentication)
         - Fine-tuning can take several hours depending on your sample size and epochs
+        - The model will be saved in the Space and can be downloaded for further use
         """)
 
 # Launch app