---
base_model: winglian/Llama-3-8b-64k-PoSE
library_name: transformers
tags:
- axolotl
- finetune
- dpo
- facebook
- meta
- pytorch
- llama
- llama-3
- 64k
- pose
language:
- en
pipeline_tag: text-generation
license: llama3
license_name: llama3
license_link: LICENSE
inference: false
model_creator: MaziyarPanahi
model_name: Llama-3-8B-Instruct-64k
quantized_by: MaziyarPanahi
datasets:
- Intel/orca_dpo_pairs
---

<img src="./llama-3-merges.webp" alt="Llama-3 DPO Logo" width="500" style="margin-left:'auto' margin-right:'auto' display:'block'"/>


# MaziyarPanahi/Llama-3-8B-Instruct-64k

This model has been made based on a great of [@winglian](https://huggingface.co/winglian/) with his latest model [winglian/Llama-3-8b-64k-PoSE](https://huggingface.co/winglian/Llama-3-8b-64k-PoSE/)

> This model uses [PoSE](https://huggingface.co/papers/2309.10400) to extend Llama's context length from 8k to 64k @ rope_theta: 500000.0. 
> We used PoSE with continued pretraining on 300M tokens from the RedPajama V1 dataset using data between 6k-8k tokens.
> We have further set rope_theta to 2M after continued pre-training to potentially further extend the context past 64k. 
> This was trained on a subset of the RedPajama v1 dataset with text between 6k-8k context. We trained a rank stabilized LoRA of rank 256. [WandB](https://wandb.ai/oaaic/llama-3-64k/runs/tkcyjt37)

# Quantized GGUF

All GGUF models come with context length of `64000`: [MaziyarPanahi/Llama-3-8B-Instruct-64k-GGUF](https://huggingface.co/MaziyarPanahi/Llama-3-8B-Instruct-64k-GGUF)


# How to use

You can use this model by using `MaziyarPanahi/Llama-3-8B-Instruct-DPO-v0.3` as the model name in Hugging Face's
transformers library.

```python
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
from transformers import pipeline
import torch

model_id = "MaziyarPanahi/Llama-3-8B-Instruct-64k"

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True,
    # attn_implementation="flash_attention_2"
)

tokenizer = AutoTokenizer.from_pretrained(
    model_id,
    trust_remote_code=True
)

streamer = TextStreamer(tokenizer)

pipeline = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    model_kwargs={"torch_dtype": torch.bfloat16},
    streamer=streamer
)

# Then you can use the pipeline to generate text.

messages = [
    {"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"},
    {"role": "user", "content": "Who are you?"},
]

prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

terminators = [
    tokenizer.eos_token_id,
    tokenizer.convert_tokens_to_ids("<|im_end|>")
]

outputs = pipeline(
    prompt,
    max_new_tokens=8192,
    eos_token_id=terminators,
    do_sample=True,
    temperature=0.6,
    top_p=0.95,
)
print(outputs[0]["generated_text"][len(prompt):])
```