Instructions to use inclusionAI/LLaDA2.0-mini-CAP with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use inclusionAI/LLaDA2.0-mini-CAP with Transformers:

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

pipe = pipeline("text-generation", model="inclusionAI/LLaDA2.0-mini-CAP", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("inclusionAI/LLaDA2.0-mini-CAP", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use inclusionAI/LLaDA2.0-mini-CAP with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "inclusionAI/LLaDA2.0-mini-CAP"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "inclusionAI/LLaDA2.0-mini-CAP",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/inclusionAI/LLaDA2.0-mini-CAP

SGLang

How to use inclusionAI/LLaDA2.0-mini-CAP 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 "inclusionAI/LLaDA2.0-mini-CAP" \
    --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": "inclusionAI/LLaDA2.0-mini-CAP",
		"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 "inclusionAI/LLaDA2.0-mini-CAP" \
        --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": "inclusionAI/LLaDA2.0-mini-CAP",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use inclusionAI/LLaDA2.0-mini-CAP with Docker Model Runner:
```
docker model run hf.co/inclusionAI/LLaDA2.0-mini-CAP
```

luguoshan commited on Dec 11, 2025

Commit

3a5e99b

1 Parent(s): 8a40c50

update README.md

Browse files

Files changed (1) hide show

README.md +6 -4

README.md CHANGED Viewed

@@ -36,13 +36,15 @@ _Evaluated on 12 diverse benchmarks covering knowledge, reasoning, coding, and m
 ### Technical Overview
 The training objective combines two complementary losses:
-$ \mathcal{L}(\theta) = \mathcal{L}_{\text{SFT}}(\theta) + \lambda \mathcal{L}_{\text{conf}}(\theta) $
 Where:
-+ $ \mathcal{L}_{\text{SFT}} $: Supervised fine-tuning loss ensuring prediction correctness
-+ $ \mathcal{L}_{\text{conf}} $: Confidence loss that minimizes entropy only for correctly predicted tokens
-+ $ \lambda $: Hyperparameter balancing the two objectives
 ### Why CAP Works
 1. **Sharpens Correct Predictions**: While standard training ensures correctness, it provides diminishing incentive to increase confidence on already-correct tokens. CAP explicitly optimizes for high-confidence predictions.

 ### Technical Overview
 The training objective combines two complementary losses:
+```math
+L(θ) = L_SFT(θ) + λL_conf(θ)
+```
 Where:
++ **L_SFT**: Supervised fine-tuning loss ensuring prediction correctness
++ **L_conf**: Confidence loss that minimizes entropy only for correctly predicted tokens
++ **λ**: Hyperparameter balancing the two objectives
 ### Why CAP Works
 1. **Sharpens Correct Predictions**: While standard training ensures correctness, it provides diminishing incentive to increase confidence on already-correct tokens. CAP explicitly optimizes for high-confidence predictions.