Update app.py

app.py

@@ -1,191 +1,149 @@
+# app.py
 import gradio as gr
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
 import os
-import sys
-import json 
-import requests
-import random
-
-
-MODEL = "gpt-4.1-mini"
-API_URL = os.getenv("API_URL")
-DISABLED = os.getenv("DISABLED") == 'True'
-OPENAI_API_KEYS = os.getenv("OPENAI_API_KEYS").split(',')
-print (API_URL)
-print (OPENAI_API_KEYS)
-NUM_THREADS = int(os.getenv("NUM_THREADS"))
-
-print (NUM_THREADS)
-
-def exception_handler(exception_type, exception, traceback):
-    print("%s: %s" % (exception_type.__name__, exception))
-sys.excepthook = exception_handler
-sys.tracebacklimit = 0
-    
-def predict(inputs, top_p, temperature, chat_counter, chatbot, history, request:gr.Request):
-    payload = {
-        "model": MODEL,
-        "messages": [{"role": "user", "content": f"{inputs}"}],
-        "temperature": temperature,
-        "top_p": top_p,
-        "n" : 1,
-        "stream": True,
-        "presence_penalty":0,
-        "frequency_penalty":0,
-    }
-    OPENAI_API_KEY = random.choice(OPENAI_API_KEYS)
-    print (OPENAI_API_KEY)
-
-    headers_dict = {key.decode('utf-8'): value.decode('utf-8') for key, value in request.headers.raw}
-
-    headers = {
-        "Content-Type": "application/json",
-        "Authorization": f"Bearer {OPENAI_API_KEY}",
-        "Headers": f"{headers_dict}"
-    }
-
-    # print(f"chat_counter - {chat_counter}")
-    if chat_counter != 0 :
-        messages = []
-        for i, data in enumerate(history):
-            if i % 2 == 0:
-                role = 'user'
-            else:
-                role = 'assistant'
-            message = {}
-            message["role"] = role
-            message["content"] = data
-            messages.append(message)
-        
-        message = {}
-        message["role"] = "user" 
-        message["content"] = inputs
-        messages.append(message)
-        payload = {
-            "model": MODEL,
-            "messages": messages,
-            "temperature" : temperature,
-            "top_p": top_p,
-            "n" : 1,
-            "stream": True,
-            "presence_penalty":0,
-            "frequency_penalty":0,
-        }
-
-    chat_counter += 1
-
-    history.append(inputs)
-    token_counter = 0 
-    partial_words = "" 
-    counter = 0
 
+# --- Configuration ---
+# <<< --- Updated Model ID --- >>>
+MODEL_ID = "Alfaxad/gemma2-2b-swahili-it"
+# <<< ---------------------- >>>
+
+# Use bf16 for performance on compatible GPUs, otherwise fp32
+TORCH_DTYPE = torch.bfloat16 if torch.cuda.is_bf16_supported() else torch.float32
+
+# device_map="auto" will handle placing the model on GPU if available, otherwise CPU.
+# For a 2B model, this works perfectly and is simpler than manual device handling.
+device_map = "auto"
+print(f"--- Using model: {MODEL_ID} ---")
+print(f"--- Using dtype: {TORCH_DTYPE} ---")
+print(f"--- Using device_map: {device_map} ---")
+
+# --- Load Model and Tokenizer ---
+# Use HF_TOKEN environment variable in Space secrets if the model requires authentication
+# access_token = os.getenv("HF_TOKEN")
+
+try:
+    print("Loading tokenizer...")
+    # trust_remote_code=True might be needed depending on the model's specifics
+    tokenizer = AutoTokenizer.from_pretrained(MODEL_ID) #, token=access_token)
+
+    print("Loading model...")
+    # device_map="auto" handles GPU/CPU placement
+    # torch_dtype optimizes for performance/memory on GPU
+    # Quantization options (load_in_8bit/4bit) are removed as they are unnecessary for a 2B model on A10G/CPU
+    model = AutoModelForCausalLM.from_pretrained(
+        MODEL_ID,
+        torch_dtype=TORCH_DTYPE,
+        device_map=device_map,
+        # token=access_token, # Uncomment if needed
+        # trust_remote_code=True # Uncomment if required by the model
+    )
+    print("Model loaded successfully.")
+
+    # <<< --- Set model to evaluation mode --- >>>
+    model.eval()
+    print("Model set to evaluation mode.")
+    # <<< ----------------------------------- >>>
+
+except Exception as e:
+    print(f"Error loading model: {e}")
+    # Use the updated MODEL_ID in the error message
+    raise gr.Error(f"Failed to load model {MODEL_ID}. Error: {e}")
+
+# --- Define Inference Function ---
+# Use inference mode decorator for efficiency
+@torch.inference_mode()
+def generate_swahili_text(prompt, max_new_tokens=150, temperature=0.7, top_p=0.9):
+    """
+    Generates text continuation in Swahili using the loaded model.
+    """
+    print(f"\n--- Received Prompt: ---\n{prompt}\n-----------------------")
+
+    # Tokenize input - device placement handled by device_map="auto"
+    inputs = tokenizer(prompt, return_tensors="pt")
+
+    print("Generating response...")
     try:
-        # make a POST request to the API endpoint using the requests.post method, passing in stream=True
-        response = requests.post(API_URL, headers=headers, json=payload, stream=True)
-        response_code = f"{response}"
-        #if response_code.strip() != "<Response [200]>":
-        #    #print(f"response code - {response}")
-        #    raise Exception(f"Sorry, hitting rate limit. Please try again later. {response}")
-        
-        for chunk in response.iter_lines():
-            #print (chunk)
-            #sys.stdout.flush()
-            #Skipping first chunk
-            if counter == 0:
-                counter += 1
-                continue
-                #counter+=1
-            # check whether each line is non-empty
-            if chunk.decode() :
-                chunk = chunk.decode()
-                # decode each line as response data is in bytes
-                if len(chunk) > 12 and "content" in json.loads(chunk[6:])['choices'][0]['delta']:
-                    partial_words = partial_words + json.loads(chunk[6:])['choices'][0]["delta"]["content"]
-                    if token_counter == 0:
-                        history.append(" " + partial_words)
-                    else:
-                        history[-1] = partial_words
-                    token_counter += 1
-                    yield [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2) ], history, chat_counter, response, gr.update(interactive=False), gr.update(interactive=False)  # resembles {chatbot: chat, state: history}  
+        # Generate text
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=max_new_tokens,
+            temperature=temperature,
+            top_p=top_p,
+            do_sample=True,
+            pad_token_id=tokenizer.eos_token_id
+        )
+
+        # Decode the generated tokens
+        result = tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+        print(f"--- Generated Output: ---\n{result}\n------------------------")
+        return result
+
     except Exception as e:
-        print (f'error found: {e}')
-    yield [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2) ], history, chat_counter, response, gr.update(interactive=True), gr.update(interactive=True)
-    print(json.dumps({"chat_counter": chat_counter, "payload": payload, "partial_words": partial_words, "token_counter": token_counter, "counter": counter}))
-                   
-
-def reset_textbox():
-    return gr.update(value='', interactive=False), gr.update(interactive=False)
-
-title = """<h1 align="center">GPT-4.1 mini: Research Preview (Short-Term Availability)</h1>"""
-if DISABLED:
-    title = """<h1 align="center" style="color:red">This app has reached OpenAI's usage limit. Please check back tomorrow.</h1>"""
-description = """Language models can be conditioned to act like dialogue agents through a conversational prompt that typically takes the form:
-```
-User: <utterance>
-Assistant: <utterance>
-User: <utterance>
-Assistant: <utterance>
-...
-```
-In this app, you can explore the outputs of a gpt-4 turbo LLM.
-"""
-
-theme = gr.themes.Default(primary_hue="green")                
-
-with gr.Blocks(css = """#col_container { margin-left: auto; margin-right: auto;}
-                #chatbot {height: 520px; overflow: auto;}""",
-              theme=theme) as demo:
-    gr.HTML(title)
-    gr.HTML("""<h3 align="center">This app provides you full access to GPT-4.1 mini (1M token limit). You don't need any OPENAI API key.</h3>""")
-
-    #gr.HTML('''<center><a href="https://huggingface.co/spaces/ysharma/ChatGPT4?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>Duplicate the Space and run securely with your OpenAI API Key</center>''')
-    with gr.Column(elem_id = "col_container", visible=False) as main_block:
-        #GPT4 API Key is provided by Huggingface 
-        #openai_api_key = gr.Textbox(type='password', label="Enter only your GPT4 OpenAI API key here")
-        chatbot = gr.Chatbot(elem_id='chatbot') #c
-        inputs = gr.Textbox(placeholder= "Hi there!", label= "Type an input and press Enter") #t
-        state = gr.State([]) #s
-        with gr.Row():
-            with gr.Column(scale=7):
-                b1 = gr.Button(visible=not DISABLED) #.style(full_width=True)
-            with gr.Column(scale=3):
-                server_status_code = gr.Textbox(label="Status code from OpenAI server", )
-    
-        #inputs, top_p, temperature, top_k, repetition_penalty
-        with gr.Accordion("Parameters", open=False):
-            top_p = gr.Slider( minimum=-0, maximum=1.0, value=1.0, step=0.05, interactive=True, label="Top-p (nucleus sampling)",)
-            temperature = gr.Slider( minimum=-0, maximum=5.0, value=1.0, step=0.1, interactive=True, label="Temperature",)
-            #top_k = gr.Slider( minimum=1, maximum=50, value=4, step=1, interactive=True, label="Top-k",)
-            #repetition_penalty = gr.Slider( minimum=0.1, maximum=3.0, value=1.03, step=0.01, interactive=True, label="Repetition Penalty", )
-            chat_counter = gr.Number(value=0, visible=False, precision=0)
-    
-    with gr.Column(elem_id = "user_consent_container") as user_consent_block:
-        # Get user consent
-        accept_checkbox = gr.Checkbox(visible=False)
-        js = "(x) => confirm('By clicking \"OK\", I agree that my data may be published or shared.')"
-        with gr.Accordion("User Consent for Data Collection, Use, and Sharing", open=True):
-            gr.HTML("""
-            <div>
-                <p>By using our app, which is powered by OpenAI's API, you acknowledge and agree to the following terms regarding the data you provide:</p>
-                <ol>
-                    <li><strong>Collection:</strong> We may collect information, including the inputs you type into our app, the outputs generated by OpenAI's API, and certain technical details about your device and connection (such as browser type, operating system, and IP address) provided by your device's request headers.</li>
-                    <li><strong>Use:</strong> We may use the collected data for research purposes, to improve our services, and to develop new products or services, including commercial applications, and for security purposes, such as protecting against unauthorized access and attacks.</li>
-                    <li><strong>Sharing and Publication:</strong> Your data, including the technical details collected from your device's request headers, may be published, shared with third parties, or used for analysis and reporting purposes.</li>
-                    <li><strong>Data Retention:</strong> We may retain your data, including the technical details collected from your device's request headers, for as long as necessary.</li>
-                </ol>
-                <p>By continuing to use our app, you provide your explicit consent to the collection, use, and potential sharing of your data as described above. If you do not agree with our data collection, use, and sharing practices, please do not use our app.</p>
-            </div>
-            """)
-            accept_button = gr.Button("I Agree")
-
-        def enable_inputs():
-            return gr.update(visible=False), gr.update(visible=True)
-
-    accept_button.click(None, None, accept_checkbox, js=js, queue=False)
-    accept_checkbox.change(fn=enable_inputs, inputs=[], outputs=[user_consent_block, main_block], queue=False)
-
-    inputs.submit(reset_textbox, [], [inputs, b1], queue=False)
-    inputs.submit(predict, [inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter, server_status_code, inputs, b1],)  #openai_api_key
-    b1.click(reset_textbox, [], [inputs, b1], queue=False)
-    b1.click(predict, [inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter, server_status_code, inputs, b1],)  #openai_api_key
-             
-    demo.queue(max_size=10, default_concurrency_limit=NUM_THREADS, api_open=False).launch(share=False)
+        print(f"Error during generation: {e}")
+        return f"Samahani, kosa limetokea wakati wa kutengeneza maandishi. (Sorry, an error occurred during text generation: {e})"
+
+# --- Create Gradio Interface ---
+theme = gr.themes.Default(primary_hue="blue", secondary_hue="neutral").set(
+    body_background_fill="#f0f0f0",
+)
+
+with gr.Blocks(theme=theme) as demo:
+    # <<< --- Updated Markdown Description --- >>>
+    gr.Markdown(f"""
+    # 🇹🇿🇰🇪🇺🇬 Gemma-2 **2B** Swahili IT Demo 🇰🇪🇺🇬🇹🇿
+
+    This Space runs the `Alfaxad/gemma2-2b-swahili-it` (2 Billion parameter) model.
+
+    Enter a prompt in Swahili, and the model will generate a continuation.
+    *This 2B model should load and respond relatively quickly.*
+    """)
+    # <<< ---------------------------------- >>>
+
+    with gr.Row():
+        with gr.Column(scale=2):
+            prompt_input = gr.Textbox(
+                label="Andika Maandishi Yako Hapa (Enter Your Prompt Here)",
+                placeholder="Mfano: Habari za asubuhi! Leo hali ya hewa ikoje Nairobi?",
+                lines=4
+            )
+            submit_button = gr.Button("Tengeneza Maandishi (Generate Text)", variant="primary")
+
+        with gr.Column(scale=3):
+             output_text = gr.Textbox(
+                 label="Matokeo (Output)",
+                 lines=10,
+                 interactive=False
+            )
+
+    with gr.Accordion("Advanced Settings", open=False):
+         max_tokens_slider = gr.Slider(minimum=10, maximum=500, value=150, step=10, label="Max New Tokens")
+         temperature_slider = gr.Slider(minimum=0.1, maximum=1.5, value=0.7, step=0.1, label="Temperature (Randomness)")
+         top_p_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.9, step=0.05, label="Top-P (Nucleus Sampling)")
+
+    # Link components to the function
+    submit_button.click(
+        fn=generate_swahili_text,
+        inputs=[prompt_input, max_tokens_slider, temperature_slider, top_p_slider],
+        outputs=output_text,
+        api_name="generate"
+    )
+
+    gr.Examples(
+        examples=[
+            ["Kenya ni nchi iliyoko wapi?", 100, 0.7, 0.9],
+            ["Niambie kuhusu historia ya Zanzibar.", 200, 0.8, 0.95],
+            ["Eleza maana ya methali 'Haraka haraka haina baraka'.", 150, 0.6, 0.9],
+        ],
+        inputs=[prompt_input, max_tokens_slider, temperature_slider, top_p_slider],
+        outputs=output_text,
+        fn=generate_swahili_text,
+        cache_examples=False
+    )
+
+# --- Launch the Gradio App ---
+print("Launching Gradio interface...")
+demo.queue().launch()
+print("Gradio interface launched.")