Upload 2 files

app.py

@@ -0,0 +1,249 @@
+import os
+import logging
+from typing import List, Optional, Tuple, Dict, Any
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel
+from transformers import AutoModelForCausalLM, GemmaTokenizerFast
+import torch
+
+# Configure logging
+logging.basicConfig(
+    level=logging.DEBUG,  # Set to DEBUG for detailed logs
+    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
+    handlers=[
+        logging.FileHandler("app.log"),  # Log to file
+        logging.StreamHandler()  # Log to console
+    ]
+)
+
+logger = logging.getLogger(__name__)
+
+# Initialize FastAPI app
+app = FastAPI()
+
+# Environment configuration
+device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+model_id = "google/gemma-2-2b-it"
+tokenizer = GemmaTokenizerFast.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    device_map="auto",
+    torch_dtype=torch.bfloat16
+)
+model.config.sliding_window = 4096
+model.eval()
+
+# Constants
+MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
+MAX_MAX_NEW_TOKENS = 2048
+DEFAULT_MAX_NEW_TOKENS = 1024
+
+# Data model for API request
+class Item(BaseModel):
+    input: Optional[str] = None
+    system_prompt: Optional[str] = None
+    system_output: Optional[str] = None
+    history: Optional[List[Tuple[str, str]]] = None
+    templates: Optional[List[str]] = None
+    temperature: float = 0.6
+    max_new_tokens: int = DEFAULT_MAX_NEW_TOKENS
+    top_p: float = 0.9
+    repetition_penalty: float = 1.2
+
+# Function to generate the response
+def generate_response(item: Item) -> Dict[str, Any]:
+    logger.debug(f"Received request: {item}")
+
+    conversation = []
+    if item.history:
+        for user, assistant in item.history:
+            conversation.extend([
+                {"role": "user", "content": user},
+                {"role": "assistant", "content": assistant},
+            ])
+    conversation.append({"role": "user", "content": item.input})
+
+    input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt")
+    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
+        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
+    
+    input_ids = input_ids.to(device)
+    generate_kwargs = dict(
+        input_ids=input_ids,
+        max_new_tokens=item.max_new_tokens,
+        do_sample=True,
+        top_p=item.top_p,
+        temperature=item.temperature,
+        num_beams=1,
+        repetition_penalty=item.repetition_penalty,
+    )
+    
+    try:
+        logger.debug("Starting text generation")
+        output = model.generate(**generate_kwargs, return_dict_in_generate=True)
+        decoded_output = tokenizer.decode(output.sequences[0], skip_special_tokens=True)
+        logger.debug("Text generation successful")
+        return {"output": decoded_output}
+    except Exception as e:
+        logger.error(f"Error during text generation: {str(e)}", exc_info=True)
+        raise HTTPException(status_code=500, detail=str(e))
+
+# Endpoint for generating text
+@app.post("/")
+async def generate_text(item: Item):
+    logger.info("Processing request")
+    if item.input is None and item.system_prompt is None:
+        logger.warning("Missing required parameters")
+        raise HTTPException(status_code=400, detail="Parameter `input` or `system prompt` is required.")
+    
+    response = generate_response(item)
+    logger.info("Request processed successfully")
+    return response
+
+# Run the app
+if __name__ == "__main__":
+    import uvicorn
+    logger.info("Starting server")
+    uvicorn.run(app, host="0.0.0.0", port=8000)
+
+'''import os
+from threading import Thread
+from typing import Iterator
+
+import gradio as gr
+import spaces
+import torch
+from transformers import AutoModelForCausalLM, GemmaTokenizerFast, TextIteratorStreamer
+
+DESCRIPTION = """\
+# Gemma 2 2B IT
+
+Gemma 2 is Google's latest iteration of open LLMs.
+This is a demo of [`google/gemma-2-2b-it`](https://huggingface.co/google/gemma-2-2b-it), fine-tuned for instruction following.
+For more details, please check [our post](https://huggingface.co/blog/gemma2).
+
+👉 Looking for a larger and more powerful version? Try the 27B version in [HuggingChat](https://huggingface.co/chat/models/google/gemma-2-27b-it) and the 9B version in [this Space](https://huggingface.co/spaces/huggingface-projects/gemma-2-9b-it).
+"""
+
+MAX_MAX_NEW_TOKENS = 2048
+DEFAULT_MAX_NEW_TOKENS = 1024
+MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
+
+device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+
+model_id = "google/gemma-2-2b-it"
+tokenizer = GemmaTokenizerFast.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    device_map="auto",
+    torch_dtype=torch.bfloat16,
+)
+model.config.sliding_window = 4096
+model.eval()
+
+
+@spaces.GPU(duration=90)
+def generate(
+    message: str,
+    chat_history: list[tuple[str, str]],
+    max_new_tokens: int = 1024,
+    temperature: float = 0.6,
+    top_p: float = 0.9,
+    top_k: int = 50,
+    repetition_penalty: float = 1.2,
+) -> Iterator[str]:
+    conversation = []
+    for user, assistant in chat_history:
+        conversation.extend(
+            [
+                {"role": "user", "content": user},
+                {"role": "assistant", "content": assistant},
+            ]
+        )
+    conversation.append({"role": "user", "content": message})
+
+    input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt")
+    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
+        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
+        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
+    input_ids = input_ids.to(model.device)
+
+    streamer = TextIteratorStreamer(tokenizer, timeout=20.0, skip_prompt=True, skip_special_tokens=True)
+    generate_kwargs = dict(
+        {"input_ids": input_ids},
+        streamer=streamer,
+        max_new_tokens=max_new_tokens,
+        do_sample=True,
+        top_p=top_p,
+        top_k=top_k,
+        temperature=temperature,
+        num_beams=1,
+        repetition_penalty=repetition_penalty,
+    )
+    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    t.start()
+
+    outputs = []
+    for text in streamer:
+        outputs.append(text)
+        yield "".join(outputs)
+
+
+chat_interface = gr.ChatInterface(
+    fn=generate,
+    additional_inputs=[
+        gr.Slider(
+            label="Max new tokens",
+            minimum=1,
+            maximum=MAX_MAX_NEW_TOKENS,
+            step=1,
+            value=DEFAULT_MAX_NEW_TOKENS,
+        ),
+        gr.Slider(
+            label="Temperature",
+            minimum=0.1,
+            maximum=4.0,
+            step=0.1,
+            value=0.6,
+        ),
+        gr.Slider(
+            label="Top-p (nucleus sampling)",
+            minimum=0.05,
+            maximum=1.0,
+            step=0.05,
+            value=0.9,
+        ),
+        gr.Slider(
+            label="Top-k",
+            minimum=1,
+            maximum=1000,
+            step=1,
+            value=50,
+        ),
+        gr.Slider(
+            label="Repetition penalty",
+            minimum=1.0,
+            maximum=2.0,
+            step=0.05,
+            value=1.2,
+        ),
+    ],
+    stop_btn=None,
+    examples=[
+        ["Hello there! How are you doing?"],
+        ["Can you explain briefly to me what is the Python programming language?"],
+        ["Explain the plot of Cinderella in a sentence."],
+        ["How many hours does it take a man to eat a Helicopter?"],
+        ["Write a 100-word article on 'Benefits of Open-Source in AI research'"],
+    ],
+    cache_examples=False,
+)
+
+with gr.Blocks(css="style.css", fill_height=True) as demo:
+    gr.Markdown(DESCRIPTION)
+    gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
+    chat_interface.render()
+
+if __name__ == "__main__":
+    demo.queue(max_size=20).launch()
+'''

requirements.txt

@@ -1,7 +1,6 @@
-fastapi
-uvicorn
-huggingface_hub
-pydantic
-google-cloud-aiplatform
-requests
-transformers
+accelerate==0.33.0
+bitsandbytes==0.43.2
+gradio==4.39.0
+spaces==0.29.2
+torch==2.2.0
+transformers==4.43.3