Instructions to use ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k with Transformers:

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

pipe = pipeline("text-generation", model="ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k")
model = AutoModelForCausalLM.from_pretrained("ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k

SGLang

How to use ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k 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 "ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k" \
    --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": "ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k",
		"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 "ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k" \
        --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": "ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k with Docker Model Runner:
```
docker model run hf.co/ryota39/llm-jp-1b-sft-100k-LoRA-dpo-12k
```

How to finetune the model?

by Labmem009 - opened May 8, 2024

Discussion

Labmem009

May 8, 2024

•

edited May 8, 2024

What kind of model is this, GPT2? How can I finetune the model? Thanks a lot!

ryota39

Owner May 8, 2024

Hi, I used this model as base model.
After training 100k wikipedia-based QA via SFT manner, DPO was performed using dataset
Thanks for comment!

Labmem009

May 8, 2024

Hi, I used this model as base model.
After training 100k wikipedia-based QA via SFT manner, DPO was performed using dataset
Thanks for comment!

I'm interested in this jp SLM. How can I fine-tune this model, could you pls offer a script or tell me how to fine-tune it?
Thanks a lot!

ryota39

Owner May 12, 2024

•

edited May 12, 2024

Hi,
I forgot to answer your previous question. Model architecture is GPT2.
I release the code in this repository.
Sorry for that English version is not ready now, but you can use codes.
You should modify the part of prompts from Japanese to your target language.
Thanks!

Labmem009

May 14, 2024

Hi,
I forgot to answer your previous question. Model architecture is GPT2.
I release the code in this repository.
Sorry for that English version is not ready now, but you can use codes.
You should modify the part of prompts from Japanese to your target language.
Thanks!

Thanks a lot!

Labmem009 changed discussion status to closed May 14, 2024

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment