Instructions to use ArchSpace-Collection/OLMo3-3B-stage3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ArchSpace-Collection/OLMo3-3B-stage3 with Transformers:

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

pipe = pipeline("text-generation", model="ArchSpace-Collection/OLMo3-3B-stage3")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ArchSpace-Collection/OLMo3-3B-stage3")
model = AutoModelForCausalLM.from_pretrained("ArchSpace-Collection/OLMo3-3B-stage3", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ArchSpace-Collection/OLMo3-3B-stage3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ArchSpace-Collection/OLMo3-3B-stage3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ArchSpace-Collection/OLMo3-3B-stage3",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/ArchSpace-Collection/OLMo3-3B-stage3

SGLang

How to use ArchSpace-Collection/OLMo3-3B-stage3 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 "ArchSpace-Collection/OLMo3-3B-stage3" \
    --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": "ArchSpace-Collection/OLMo3-3B-stage3",
		"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 "ArchSpace-Collection/OLMo3-3B-stage3" \
        --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": "ArchSpace-Collection/OLMo3-3B-stage3",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use ArchSpace-Collection/OLMo3-3B-stage3 with Docker Model Runner:
```
docker model run hf.co/ArchSpace-Collection/OLMo3-3B-stage3
```

OLMo3-3B-stage3

File size: 1,535 Bytes

1830ad1

---
library_name: transformers
pipeline_tag: text-generation
tags:
- olmo3
- safetensors
- sliding-window-attention
- baseline
---

# OLMo 3 3B Baseline — Stage 3 Long-context Training

This repository is the Hugging Face export of `o3b3b-s3-s65536-g64-m1-ga1-tp2-cp8-dp64-h64-b2-lr2p5e4-w200-save1000-1024npu-share-0906045054-s3v1` at
iteration `11921`. This is the matched pure OLMo 3 baseline. It uses Transformers' official `Olmo3ForCausalLM` implementation and does not require remote code.

- Training sequence length: 65,536
- Model context capacity: 65,536
- Sliding-window size: 4,096
- Attention pattern: `[SWA, SWA, SWA, Full]`
- Vocabulary: 100,278 real tokens; 74 Megatron padding-only rows removed

Stage 3/4 use the frozen 65,536-token configuration. YaRN applies to the Full
Attention layers; SWA layers retain their original RoPE and 4,096-token local
window.

## Loading

Use `transformers>=4.57.6,<5`.

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

repo_id = "ArchSpace-Collection/OLMo3-3B-stage3"
tokenizer = AutoTokenizer.from_pretrained(
    repo_id,
    use_fast=True,
    fix_mistral_regex=False,
)
model = AutoModelForCausalLM.from_pretrained(
    repo_id,
    dtype=torch.bfloat16,
    attn_implementation="sdpa",
)
```

`fix_mistral_regex=False` preserves the exact tokenizer behavior used during
training. Conversion provenance, per-tensor hashes, and CPU validation results
are included in `conversion_manifest.json`, `SHA256SUMS`, and
`hf_validation_report.json`.