Add Gradio app for pashto-base-bloom with train/finetune/test

README.md

@@ -1,12 +1,22 @@
 ---
-title: Pashto Base Bloom Space
-emoji: 📉
+title: pashto-base-bloom Training Space
+emoji: 🚀
 colorFrom: blue
-colorTo: green
+colorTo: purple
 sdk: gradio
-sdk_version: 5.36.2
+sdk_version: 4.36.1
 app_file: app.py
 pinned: false
+license: apache-2.0
+hardware: zero-a10g
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# pashto-base-bloom Training Space
+
+This space provides three main functionalities for the pashto-base-bloom model:
+
+1. **Train**: Train the model from scratch
+2. **Fine-tune**: Fine-tune the existing model  
+3. **Test**: Test the model with sample inputs
+
+The space uses ZeroGPU for efficient GPU computation.

app.py

@@ -0,0 +1,145 @@
+import gradio as gr
+import spaces
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import os
+
+# Model configuration
+MODEL_NAME = "tasal9/pashto-base-bloom"
+
+@spaces.GPU
+def load_model():
+    """Load the model and tokenizer"""
+    try:
+        tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+        model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype=torch.float16)
+        if tokenizer.pad_token is None:
+            tokenizer.pad_token = tokenizer.eos_token
+        return model, tokenizer
+    except Exception as e:
+        return None, None
+
+@spaces.GPU
+def test_model(input_text, max_length=100, temperature=0.7):
+    """Test the model with given input"""
+    if not input_text.strip():
+        return "Please enter some text to test the model."
+    
+    model, tokenizer = load_model()
+    
+    if model is None or tokenizer is None:
+        return "❌ Failed to load model. Please check if the model exists on Hugging Face Hub."
+    
+    try:
+        inputs = tokenizer.encode(input_text, return_tensors="pt")
+        
+        with torch.no_grad():
+            outputs = model.generate(
+                inputs,
+                max_length=len(inputs[0]) + max_length,
+                temperature=temperature,
+                do_sample=True,
+                pad_token_id=tokenizer.eos_token_id
+            )
+        
+        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+        return response[len(input_text):].strip()
+        
+    except Exception as e:
+        return f"❌ Error during generation: {str(e)}"
+
+def train_model(dataset_text, epochs=1, learning_rate=2e-5):
+    """Train the model (placeholder implementation)"""
+    return f"🚀 Training started with {epochs} epochs and learning rate {learning_rate}\n\nNote: This is a placeholder. Actual training requires dataset preparation and more computational resources."
+
+def finetune_model(dataset_text, epochs=1, learning_rate=5e-5):
+    """Fine-tune the model (placeholder implementation)"""
+    return f"🔧 Fine-tuning started with {epochs} epochs and learning rate {learning_rate}\n\nNote: This is a placeholder. Actual fine-tuning requires dataset preparation and more computational resources."
+
+# Create Gradio interface
+with gr.Blocks(title="pashto-base-bloom Training Space", theme=gr.themes.Soft()) as iface:
+    gr.Markdown(f"# pashto-base-bloom Training Space")
+    gr.Markdown("Choose your operation: Train, Fine-tune, or Test the model")
+    
+    with gr.Tabs():
+        # Test Tab
+        with gr.TabItem("🧪 Test Model"):
+            gr.Markdown("### Test the model with your input")
+            with gr.Row():
+                with gr.Column():
+                    test_input = gr.Textbox(
+                        label="Input Text",
+                        placeholder="Enter text to test the model...",
+                        lines=3
+                    )
+                    max_length_slider = gr.Slider(
+                        minimum=10,
+                        maximum=500,
+                        value=100,
+                        label="Max Length"
+                    )
+                    temperature_slider = gr.Slider(
+                        minimum=0.1,
+                        maximum=2.0,
+                        value=0.7,
+                        label="Temperature"
+                    )
+                    test_btn = gr.Button("🚀 Generate", variant="primary")
+                
+                with gr.Column():
+                    test_output = gr.Textbox(
+                        label="Model Output",
+                        lines=5,
+                        interactive=False
+                    )
+            
+            test_btn.click(
+                fn=test_model,
+                inputs=[test_input, max_length_slider, temperature_slider],
+                outputs=test_output
+            )
+        
+        # Train Tab
+        with gr.TabItem("🏋️ Train Model"):
+            gr.Markdown("### Train the model from scratch")
+            train_dataset = gr.Textbox(
+                label="Training Dataset",
+                placeholder="Upload or paste your training data...",
+                lines=5
+            )
+            with gr.Row():
+                train_epochs = gr.Number(label="Epochs", value=1, minimum=1)
+                train_lr = gr.Number(label="Learning Rate", value=2e-5, minimum=1e-6)
+            
+            train_btn = gr.Button("🚀 Start Training", variant="primary")
+            train_output = gr.Textbox(label="Training Output", lines=5, interactive=False)
+            
+            train_btn.click(
+                fn=train_model,
+                inputs=[train_dataset, train_epochs, train_lr],
+                outputs=train_output
+            )
+        
+        # Fine-tune Tab
+        with gr.TabItem("🔧 Fine-tune Model"):
+            gr.Markdown("### Fine-tune the existing model")
+            finetune_dataset = gr.Textbox(
+                label="Fine-tuning Dataset",
+                placeholder="Upload or paste your fine-tuning data...",
+                lines=5
+            )
+            with gr.Row():
+                finetune_epochs = gr.Number(label="Epochs", value=1, minimum=1)
+                finetune_lr = gr.Number(label="Learning Rate", value=5e-5, minimum=1e-6)
+            
+            finetune_btn = gr.Button("🔧 Start Fine-tuning", variant="primary")
+            finetune_output = gr.Textbox(label="Fine-tuning Output", lines=5, interactive=False)
+            
+            finetune_btn.click(
+                fn=finetune_model,
+                inputs=[finetune_dataset, finetune_epochs, finetune_lr],
+                outputs=finetune_output
+            )
+
+if __name__ == "__main__":
+    iface.launch()

requirements.txt

@@ -0,0 +1,6 @@
+gradio==4.36.1
+spaces
+torch
+transformers
+datasets
+accelerate