README.md

---
base_model: meta-llama/Meta-Llama-3-8B-Instruct
inference: false
model_creator: astronomer-io
model_name: Meta-Llama-3-8B-Instruct
model_type: llama
pipeline_tag: text-generation
prompt_template: >-
  {% set loop_messages = messages %}{% for message in loop_messages %}{% set
  content = '<|start_header_id|>' + message['role'] + '<|end_header_id|>


  '+ message['content'] | trim + '<|eot_id|>' %}{% if loop.index0 == 0 %}{% set
  content = bos_token + content %}{% endif %}{{ content }}{% endfor %}{% if
  add_generation_prompt %}{{ '<|start_header_id|>assistant<|end_header_id|>


  ' }}{% endif %}
quantized_by: davidxmle
license: other
license_name: llama-3-community-license
license_link: https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct/blob/main/LICENSE
tags:
- llama
- llama-3
- facebook
- meta
- astronomer
- gptq
- pretrained
datasets:
- wikitext
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<img src="https://www.astronomer.io/logo/astronomer-logo-RGB-standard-1200px.png" alt="Astronomer" style="width: 60%; min-width: 400px; display: block; margin: auto;">
</div>
<div style="margin-top: 1.0em; margin-bottom: 1.0em;"></div>

<div style="text-align:center; margin-top: 0em; margin-bottom: 0em"><p style="margin-top: 0.25em; margin-bottom: 0em;">This model is generously created and made open source by <a href="https://astronomer.io">Astronomer</a>.</p></div>
<div style="text-align:center; margin-top: 0em; margin-bottom: 0em"><p style="margin-top: 0.25em; margin-bottom: 0em;">Astronomer is the de facto company for <a href="https://airflow.apache.org/">Apache Airflow</a>, the most trusted open-source framework for data orchestration and MLOps.</p></div>
<hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
<!-- header end -->
  
# Important Note Regarding a Known Bug in Llama 3
- Two files are modified to address a current issue regarding Llama 3 models keep on generating additional tokens non-stop until hitting max token limit.
- `generation_config.json`'s `eos_token_id` have been modified to add the other EOS token that Llama-3 uses.
- `tokenizer_config.json`'s `chat_template` has been modified to only add start generation token at the end of a prompt if `add_generation_prompt` is selected.
- For loading this model onto vLLM, make sure all requests have `"stop_token_ids":[128001, 128009]` to temporarily address the non-stop generation issue.
   - vLLM does not yet respect `generation_config.json`.
   - vLLM team is working on a a fix for this https://github.com/vllm-project/vllm/issues/4180

# Llama-3-8B-Instruct-GPTQ-8-Bit
- Original Model creator: [Meta Llama from Meta](https://huggingface.co/meta-llama)
- Original model: [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct)
- Built with Meta Llama 3
- Quantized by [Astronomer](https://astronomer.io)

<!-- description start -->
## Description

This repo contains 8 Bit quantized GPTQ model files for [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct).

This model can be loaded with just over 10GB of VRAM and can be served lightning fast with the cheapest Nvidia GPUs possible (Nvidia T4, Nvidia K80, RTX 4070, etc).

The 8 bit GPTQ quant has minimum quality degradation from the original `bfloat16` model due to its higher bitrate.

<!-- description end -->

## GPTQ Quantization Method
- This model is quantized by utilizing the AutoGPTQ library, following best practices noted by [GPTQ paper](https://arxiv.org/abs/2210.17323)
- Quantization is calibrated and aligned with random samples from the specified dataset (wikitext for now) for minimum accuracy loss.

| Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | VRAM Size | ExLlama | Desc |
| ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
| [main](https://huggingface.co/astronomer-io/Llama-3-8B-Instruct-GPTQ-8-Bit/tree/main) | 8 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 9.09 GB | No | 8-bit, with Act Order and group size 32g. Minimum accuracy loss with decent VRAM usage reduction. | 
| More variants to come | TBD | TBD | TBD | TBD | TBD | TBD | TBD | TBD | May upload additional variants of GPTQ 8 bit models in the future using different parameters such as 128g group size and etc. |

## Serving this GPTQ model using vLLM
Tested serving this model via vLLM using an Nvidia T4 (16GB VRAM).

Tested with the below command
```
python -m vllm.entrypoints.openai.api_server --model Llama-3-8B-Instruct-GPTQ-8-Bit --port 8123 --max-model-len 8192 --dtype float16
```
For the non-stop token generation bug, make sure to send requests with `stop_token_ids":[128001, 128009]` to vLLM endpoint
Example:
```
{
    "model": "Llama-3-8B-Instruct-GPTQ-8-Bit",
    "messages": [
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Who created Llama 3?"}
        ],
    "max_tokens": 2000,
    "stop_token_ids":[128001,128009]
}
```

### Contributors
- Quantized by [David Xue, Machine Learning Engineer from Astronomer](https://www.linkedin.com/in/david-xue-uva/)