Text Generation
Transformers
Safetensors
English
nemotron_h
nvidia
nemotron-cascade-2
reasoning
general-purpose
SFT
RL
conversational
custom_code
Eval Results
Instructions to use nvidia/Nemotron-Cascade-2-30B-A3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use nvidia/Nemotron-Cascade-2-30B-A3B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="nvidia/Nemotron-Cascade-2-30B-A3B", trust_remote_code=True) messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("nvidia/Nemotron-Cascade-2-30B-A3B", trust_remote_code=True) model = AutoModelForCausalLM.from_pretrained("nvidia/Nemotron-Cascade-2-30B-A3B", trust_remote_code=True) 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
- vLLM
How to use nvidia/Nemotron-Cascade-2-30B-A3B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "nvidia/Nemotron-Cascade-2-30B-A3B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "nvidia/Nemotron-Cascade-2-30B-A3B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/nvidia/Nemotron-Cascade-2-30B-A3B
- SGLang
How to use nvidia/Nemotron-Cascade-2-30B-A3B 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 "nvidia/Nemotron-Cascade-2-30B-A3B" \ --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": "nvidia/Nemotron-Cascade-2-30B-A3B", "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 "nvidia/Nemotron-Cascade-2-30B-A3B" \ --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": "nvidia/Nemotron-Cascade-2-30B-A3B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use nvidia/Nemotron-Cascade-2-30B-A3B with Docker Model Runner:
docker model run hf.co/nvidia/Nemotron-Cascade-2-30B-A3B
Cascade-2 cheating attempts on Math is... cute
👍🤯 4
2
#25 opened about 1 month ago
by
chankhavu
use system prompt for reasoning benchmarks reproduction (IMO-AnswerBench,..)
👍 2
#24 opened about 2 months ago
by
ychenNLP
Nemotron Cascade mutates nested path literals in tool arguments even with explicit literal-preservation instructions and temperature=0.
👍 3
#23 opened about 2 months ago
by
joonsoo-me
MoE efficiency for agentic workflows: thinking mode trade-offs
#22 opened about 2 months ago
by
O96a
useless but fast!
👀👍 9
1
#21 opened about 2 months ago
by
Kosh69
187 tok/s on RTX 3090, 625K Context, Agent Coding (IQ4_XS + Hermes Agent)
🚀🔥 3
7
#20 opened about 2 months ago
by
ychenNLP
Cascade reasoning latency in production
🚀 2
#19 opened about 2 months ago
by
O96a
Integration Proposal: Aligning with the AURA Sovereign Protocol for Human Enlightenment
#18 opened about 2 months ago
by
Swiatel
COMMUNITY DISCUSSION
#17 opened about 2 months ago
by
nairobiansmasher
pruned version
👀🔥 1
2
#16 opened about 2 months ago
by
pirola
The model's self-perception is unstable within a Chinese context.
👀 1
1
#15 opened about 2 months ago
by
Jianqiao1
TEST DISCUSSION
1
#14 opened about 2 months ago
by
DiakonFrost
modeling_nemotron_h.py: Multiple bugs in HybridMambaAttentionDynamicCache break generation with CUDA fast path
2
#13 opened about 2 months ago
by
trohrbaugh
no quants working
2
#12 opened about 2 months ago
by
audioedge
Tool call format degrading at higher context
2
#10 opened about 2 months ago
by
ilintar
Official quantizations?
👍 2
6
#9 opened about 2 months ago
by
wijjjj
Add documentation on how to use with vLLM to README.md
🤝👍 6
1
#7 opened about 2 months ago
by
stelterlab
Installation Video and Testing - Step by Step
🤯 4
#5 opened 2 months ago
by
fahdmirzac
well cooked !! rtx 3060 loves it
❤️ 5
#4 opened 2 months ago
by
gopi87
Tool calling ability
🔥 2
1
#3 opened 2 months ago
by
spanspek
Add eval results
#2 opened 2 months ago
by
merve