app.py

!wget -q https://github.com/PanQiWei/AutoGPTQ/releases/download/v0.4.1/auto_gptq-0.4.1+cu118-cp310-cp310-linux_x86_64.whl
!BUILD_CUDA_EXT=0 pip install -qqq auto_gptq-0.4.1+cu118-cp310-cp310-linux_x86_64.whl --progress-bar off

import gradio as gr
from auto_gptq import AutoGPTQForCausalLM
from transformers import AutoTokenizer, TextStreamer

title = "Npradhaph"

examples = [
    ["The tower is 324 metres (1,063 ft) tall,"],
    ["The Moon's orbit around Earth has"],
    ["The smooth Borealis basin in the Northern Hemisphere covers 40%"],
]


# Load the trained model
model_path = "huggingface/pradhaph/medical-falcon-7b"
model = AutoGPTQForCausalLM.from_quantized(
    model_path,
    revision="main",
    # revision="gptq-8bit-128g-actorder_True",
    model_basename="model",
    use_safetensors=True,
    trust_remote_code=True,
    inject_fused_attention=False,
    device_map="cuda",
    quantize_config=None,
)
tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=True)

# Define the input and output interfaces
def answer_question(context):
    # Generate an answer based on the context
    inputs = tokenizer(context, return_tensors="pt", max_length=512, truncation=True)
    outputs = model.generate(**inputs, max_length=200, num_return_sequences=1)
    answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return answer

# Run the interface
iface = gr.Interface(
    fn=answer_question,
    inputs="text",
    outputs="text",
    title="Question Answering with GPT",
    description="Enter a context to get an answer."
)

# demo = gr.load(
#     "huggingface/pradhaph/medical-falcon-7b",
#     inputs=gr.Textbox(lines=5, max_lines=6, label="Input Text"),
#     title=title,
#     examples=examples,
#     trust_remote_code=True,
# )

if __name__ == "__main__":
    iface.launch()
    # demo.launch()