philschmid/codegen-6B-mono-sharded-bnb

Sharded fork of Salesforce/codegen-6B-mono with a custom pipeline.py

This repository implements a custom pipeline task for text-generation for 🤗 Inference Endpoints for LLM inference using bitsandbytes quantization. The code for the customized pipeline is in the pipeline.py.

There is also a notebook included.

expected Request payload

{
    "inputs": "# load distilbert model and initialize text-classification pipeline\nmodel_id = 'distil",
    "parameters": {
        "top_k": 100,
        "max_length": 64,
        "early_stopping": true,
        "do_sample": true,
        "eos_token_id": 50256,
    }
}

below is an example on how to run a request using Python and requests.

Run Request

import json
from typing import List
import requests as r
import base64
ENDPOINT_URL = ""
HF_TOKEN = ""

parameters={
        "top_k": 100,
        "max_length": 64,
        "early_stopping": True,
        "do_sample": True,
        "eos_token_id": 50256,
    }

def predict(code_snippet:str=None):
    payload = {"inputs": code_snippet,"parameters": parameters}
    response = r.post(
        ENDPOINT_URL, headers={"Authorization": f"Bearer {HF_TOKEN}"}, json=payload
    )
    return response.json()
prediction = predict(
    code_snippet="# load distilbert model and initialize text-classification pipeline\nmodel_id = 'distil"
)

expected output

{'generated_text': "# load distilbert model and initialize text-classification pipeline\nmodel_id = 'distilbert-base-uncased'\nmodel_url = 'https://tfhub.dev/tensorflow/small_bert/1'\n\nmodel_dir = './distilBERT'"}