Instructions to use skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo with Transformers:

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

pipe = pipeline("text-generation", model="skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo")
model = AutoModelForCausalLM.from_pretrained("skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo

SGLang

How to use skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo 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 "skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo with Docker Model Runner:
```
docker model run hf.co/skyai798/STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo
```

STAR-1_DeepSeek-R1-Distill-Llama-8B_sft-complete-dpo / trainer_log.jsonl

skyai798

Upload folder using huggingface_hub

261430f verified 11 months ago

Raw

History Blame Contribute Delete

2.23 kB

	{"current_steps": 2, "total_steps": 20, "loss": 0.6931, "accuracy": 0.0, "lr": 9.931806517013612e-06, "epoch": 0.19047619047619047, "percentage": 10.0, "elapsed_time": "0:00:09", "remaining_time": "0:01:25"}
	{"current_steps": 4, "total_steps": 20, "loss": 0.9007, "accuracy": 0.5625, "lr": 9.397368756032445e-06, "epoch": 0.38095238095238093, "percentage": 20.0, "elapsed_time": "0:00:16", "remaining_time": "0:01:07"}
	{"current_steps": 6, "total_steps": 20, "loss": 2.869, "accuracy": 0.25, "lr": 8.386407858128707e-06, "epoch": 0.5714285714285714, "percentage": 30.0, "elapsed_time": "0:00:24", "remaining_time": "0:00:57"}
	{"current_steps": 8, "total_steps": 20, "loss": 3.0671, "accuracy": 0.21875, "lr": 7.008477123264849e-06, "epoch": 0.7619047619047619, "percentage": 40.0, "elapsed_time": "0:00:32", "remaining_time": "0:00:48"}
	{"current_steps": 10, "total_steps": 20, "loss": 1.3723, "accuracy": 0.5, "lr": 5.412896727361663e-06, "epoch": 0.9523809523809523, "percentage": 50.0, "elapsed_time": "0:00:39", "remaining_time": "0:00:39"}
	{"current_steps": 12, "total_steps": 20, "loss": 1.9476, "accuracy": 0.875, "lr": 3.7725725642960047e-06, "epoch": 1.1428571428571428, "percentage": 60.0, "elapsed_time": "0:03:11", "remaining_time": "0:02:07"}
	{"current_steps": 14, "total_steps": 20, "loss": 0.5622, "accuracy": 0.90625, "lr": 2.265259209387867e-06, "epoch": 1.3333333333333333, "percentage": 70.0, "elapsed_time": "0:03:19", "remaining_time": "0:01:25"}
	{"current_steps": 16, "total_steps": 20, "loss": 0.2122, "accuracy": 0.90625, "lr": 1.0542974530180327e-06, "epoch": 1.5238095238095237, "percentage": 80.0, "elapsed_time": "0:03:26", "remaining_time": "0:00:51"}
	{"current_steps": 18, "total_steps": 20, "loss": 0.3109, "accuracy": 0.90625, "lr": 2.7091379149682683e-07, "epoch": 1.7142857142857144, "percentage": 90.0, "elapsed_time": "0:03:34", "remaining_time": "0:00:23"}
	{"current_steps": 20, "total_steps": 20, "loss": 0.8478, "accuracy": 0.875, "lr": 0.0, "epoch": 1.9047619047619047, "percentage": 100.0, "elapsed_time": "0:03:41", "remaining_time": "0:00:00"}
	{"current_steps": 20, "total_steps": 20, "epoch": 1.9047619047619047, "percentage": 100.0, "elapsed_time": "0:07:12", "remaining_time": "0:00:00"}