Instructions to use syzymon/long_llama_code_7b_instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use syzymon/long_llama_code_7b_instruct with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="syzymon/long_llama_code_7b_instruct", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("syzymon/long_llama_code_7b_instruct", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use syzymon/long_llama_code_7b_instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "syzymon/long_llama_code_7b_instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "syzymon/long_llama_code_7b_instruct",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/syzymon/long_llama_code_7b_instruct

SGLang

How to use syzymon/long_llama_code_7b_instruct with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "syzymon/long_llama_code_7b_instruct" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "syzymon/long_llama_code_7b_instruct",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "syzymon/long_llama_code_7b_instruct" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "syzymon/long_llama_code_7b_instruct",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use syzymon/long_llama_code_7b_instruct with Docker Model Runner:
```
docker model run hf.co/syzymon/long_llama_code_7b_instruct
```

long_llama_code_7b_instruct / README.md

syzymon

Update README.md

67ca5cc almost 3 years ago

preview code

Raw

History Blame Contribute Delete

4.31 kB

	---
	license: llama2
	---
	# LongLLaMA-Code 7B Instruct


	<div align="center">

	<table>
	<tr>
	<th style="font-size: 120%"> >_ 🎓 <a href="https://huggingface.co/syzymon/long_llama_code_7b_instruct">LongLLaMA-Code 7B Instruct</a> 📑🗨 </th>
	</tr>
	<tr>
	<td align="center">
	<a href="https://colab.research.google.com/github/CStanKonrad/long_llama/blob/main/long_llama_code_instruct_colab.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg"></a>
	</td>

	</tr>
	</table>

	</div>


	## TLDR
	[LongLLaMA-Code 7B Instruct](https://huggingface.co/syzymon/long_llama_code_7b_instruct) is [LongLLaMA-Code 7B](https://huggingface.co/syzymon/long_llama_code_7b) tuned on [TIGER-Lab/MathInstruct](https://huggingface.co/datasets/TIGER-Lab/MathInstruct), [OpenOrca](https://huggingface.co/datasets/Open-Orca/OpenOrca), and [ShareGPT-Processed](https://huggingface.co/datasets/zetavg/ShareGPT-Processed) datasets. It can answer basic questions about research papers and code. It can also perform a simple code refactoring. You can try the quantized version of the model using a free GPU in [Google Colab](https://colab.research.google.com/github/CStanKonrad/long_llama/blob/main/long_llama_code_instruct_colab.ipynb).

	## Tuning

	### Code
	The model was tuned on a TPU v3-128 pod with 128 batch size.
	For tuning, we have used the data preparation pipeline available in instruction_fine_tuning.
	However, we have replaced the Hugging Face Trainer with a modification of FoT continued pretraining code. This modification boils down to propagating the memory cache throughout the model (basically reproducing the Pytorch inference code functionality in JAX).

	### Training
	Here, we present the basic information about how the model was tuned. For more details, see the [GitHub repo](https://github.com/CStanKonrad/long_llama/tree/main/instruction_fine_tuning/misc).


	All inputs were truncated and randomly padded (left/right) to 3072 tokens.
	The last context length was set to 1536.
	The model was trained for 9k steps, started with a learning rate of 1.2e-5, 700 steps of warmup, and finished with a learning rate of 0.
	The optimizer was adamw.

	The question prompt (`pre_question_text`) was:
	```
	You are an AI assistant. User will you give you a task. Your goal is to complete the task as faithfully as you can.\n\n
	```

	To trigger the model answer one can use:
	```
	\nAnswer:
	```

	The chat prompt was:
	```
	A chat between a user (denoted as USER:) and an artificial intelligence assistant (denoted as ASSISTANT:). The assistant gives helpful, detailed, and polite answers to the user's questions.\n\n
	```

	To denote the assistant one can write:
	```
	\nASSISTANT:
	```

	To denote the user one can write:
	```
	\nUSER:
	```

	### Datasets and sampling probability
	* 0.71 - [TIGER-Lab/MathInstruct](https://huggingface.co/datasets/TIGER-Lab/MathInstruct)
	* 0.16, - [Open-Orca/OpenOrca](https://huggingface.co/datasets/Open-Orca/OpenOrca) questions with less than 5k chars
	* 0.08, - [Open-Orca/OpenOrca](https://huggingface.co/datasets/Open-Orca/OpenOrca) questions above 5k chars but below 12k chars
	* 0.02 - [zetavg/ShareGPT-Processed](https://huggingface.co/datasets/zetavg/ShareGPT-Processed) conversations below 6k chars
	* 0.01 - [zetavg/ShareGPT-Processed](https://huggingface.co/datasets/zetavg/ShareGPT-Processed) conversations above 6k chars but below 12k chars

	To improve the quality of the data, the datasets were filtered using regular expressions.



	## License
	The instruction/chat-tuned models are for research purposes only.
	[LongLLaMA-Code 7B Instruct](https://huggingface.co/syzymon/long_llama_code_7b_instruct) is [LongLLaMA-Code 7B](https://huggingface.co/syzymon/long_llama_code_7b) tuned on [TIGER-Lab/MathInstruct](https://huggingface.co/datasets/TIGER-Lab/MathInstruct), [OpenOrca](https://huggingface.co/datasets/Open-Orca/OpenOrca), and [ShareGPT-Processed](https://huggingface.co/datasets/zetavg/ShareGPT-Processed) datasets. Note that those datasets contain outputs from ChatGPT. See also the [codellama/CodeLlama-7b-hf](https://huggingface.co/codellama/CodeLlama-7b-hf) license.

	## Acknowledgements
	We gratefully acknowledge the TPU Research Cloud program, which was instrumental to our research by providing significant computational resources.