Instructions to use open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k with Transformers:

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

pipe = pipeline("text-generation", model="open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k

SGLang

How to use open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k 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 "open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k" \
    --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": "open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k",
		"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 "open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k" \
        --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": "open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k with Docker Model Runner:
```
docker model run hf.co/open-sci/open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k
```

open-sci-ref-v0.02-1.7b-dclm-1T-4096-rope_theta-100k

1.7B open-sci-ref model trained on DCLM for 1T tokens (sequence length 4096, RoPE theta = 100000).

The main branch holds the final checkpoint (iter 238419). Intermediate checkpoints (iters 58000-238000, every 2000) are available as branches named iter_XXXXXXX.

Evaluation

Final checkpoint on the open-sci-0.01 suite (lm-eval-harness). Metrics collected with oellm collect-results.

Task	n-shot	Metric	Score
arc_challenge	10	acc_norm	0.4761
arc_easy	10	acc_norm	0.7689
boolq	10	acc	0.7664
commonsense_qa	10	acc	0.4676
copa	0	acc	0.8200
hellaswag	10	acc_norm	0.7367
lambada_openai	0	acc	0.7019
mmlu	5	acc	0.4002
openbookqa	0	acc_norm	0.4080
piqa	10	acc_norm	0.7845
social_iqa	0	acc	0.4463
winogrande	0	acc	0.6606
average			0.6198

Downloads last month: 836

Safetensors

Model size

2B params

Tensor type

BF16