Instructions to use ModalityDance/slpo-codi-gpt2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ModalityDance/slpo-codi-gpt2 with Transformers:

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

pipe = pipeline("text-generation", model="ModalityDance/slpo-codi-gpt2")

# Load model directly
from transformers import AutoTokenizer, LatentCODIGPT2

tokenizer = AutoTokenizer.from_pretrained("ModalityDance/slpo-codi-gpt2")
model = LatentCODIGPT2.from_pretrained("ModalityDance/slpo-codi-gpt2", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ModalityDance/slpo-codi-gpt2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ModalityDance/slpo-codi-gpt2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ModalityDance/slpo-codi-gpt2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/ModalityDance/slpo-codi-gpt2

SGLang

How to use ModalityDance/slpo-codi-gpt2 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 "ModalityDance/slpo-codi-gpt2" \
    --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": "ModalityDance/slpo-codi-gpt2",
		"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 "ModalityDance/slpo-codi-gpt2" \
        --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": "ModalityDance/slpo-codi-gpt2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use ModalityDance/slpo-codi-gpt2 with Docker Model Runner:
```
docker model run hf.co/ModalityDance/slpo-codi-gpt2
```

dd101bb commited on 7 days ago

Commit

252a35d

verified ·

1 Parent(s): 1dd7ec4

Update model card with arXiv 2607.19691

Browse files

Files changed (1) hide show

README.md +8 -5

README.md CHANGED Viewed

@@ -14,7 +14,7 @@ base_model: ModalityDance/latent-tts-codi
 # CODI + SLPO (GPT-2)
 Surrogate Latent Policy Optimization (**SLPO**) checkpoint on top of [CODI](https://huggingface.co/ModalityDance/latent-tts-codi) (GPT-2 124M).
-This is the **CODI+SLPO** model reported in the paper *SLPO: Scaling Latent Reasoning with Surrogate Policy Optimization*.
 ## Model Details
@@ -36,8 +36,8 @@ Deterministic accuracy with dropout disabled and learned stop gate:
 ## Related
-- Paper (arXiv): *coming soon* — placeholder `https://arxiv.org/abs/XXXX.XXXXX`
-- Hugging Face Paper: *coming soon* — placeholder `https://huggingface.co/papers/XXXX.XXXXX`
 - Code: [ModalityDance/SLPO](https://github.com/ModalityDance/SLPO)
 - Project page: [modalitydance.github.io/SLPO](https://modalitydance.github.io/SLPO/)
 - Base model: [ModalityDance/latent-tts-codi](https://huggingface.co/ModalityDance/latent-tts-codi)
@@ -116,9 +116,12 @@ print(tokenizer.decode(sequences[0], skip_special_tokens=True))
 ```bibtex
 @misc{you2026slpo,
-  title   = {SLPO: Scaling Latent Reasoning with Surrogate Policy Optimization},
   author  = {You, Runyang and Liu, Zhiyuan and Li, Yongqi and Li, Wenjie},
   year    = {2026},
-  note    = {Code: https://github.com/ModalityDance/SLPO}
 }
 ```

 # CODI + SLPO (GPT-2)
 Surrogate Latent Policy Optimization (**SLPO**) checkpoint on top of [CODI](https://huggingface.co/ModalityDance/latent-tts-codi) (GPT-2 124M).
+This is the **CODI+SLPO** model reported in the paper *SLPO: Scaling Latent Reasoning via a Surrogate Policy*.
 ## Model Details
 ## Related
+- Paper (arXiv): [2607.19691](https://arxiv.org/abs/2607.19691)
+- Hugging Face Paper: [2607.19691](https://huggingface.co/papers/2607.19691)
 - Code: [ModalityDance/SLPO](https://github.com/ModalityDance/SLPO)
 - Project page: [modalitydance.github.io/SLPO](https://modalitydance.github.io/SLPO/)
 - Base model: [ModalityDance/latent-tts-codi](https://huggingface.co/ModalityDance/latent-tts-codi)
 ```bibtex
 @misc{you2026slpo,
+  title   = {SLPO: Scaling Latent Reasoning via a Surrogate Policy},
   author  = {You, Runyang and Liu, Zhiyuan and Li, Yongqi and Li, Wenjie},
   year    = {2026},
+  eprint  = {2607.19691},
+  archivePrefix = {arXiv},
+  primaryClass = {cs.CL},
+  url     = {https://arxiv.org/abs/2607.19691}
 }
 ```