Instructions to use nvidia/Nemotron-H-8B-Reasoning-128K-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nvidia/Nemotron-H-8B-Reasoning-128K-FP8 with Transformers:

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

pipe = pipeline("text-generation", model="nvidia/Nemotron-H-8B-Reasoning-128K-FP8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("nvidia/Nemotron-H-8B-Reasoning-128K-FP8", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use nvidia/Nemotron-H-8B-Reasoning-128K-FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nvidia/Nemotron-H-8B-Reasoning-128K-FP8"
# 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-H-8B-Reasoning-128K-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/nvidia/Nemotron-H-8B-Reasoning-128K-FP8

SGLang

How to use nvidia/Nemotron-H-8B-Reasoning-128K-FP8 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-H-8B-Reasoning-128K-FP8" \
    --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-H-8B-Reasoning-128K-FP8",
		"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-H-8B-Reasoning-128K-FP8" \
        --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-H-8B-Reasoning-128K-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use nvidia/Nemotron-H-8B-Reasoning-128K-FP8 with Docker Model Runner:
```
docker model run hf.co/nvidia/Nemotron-H-8B-Reasoning-128K-FP8
```

Nemotron-H-8B-Reasoning-128K-FP8 / config.json

bkartal

Upload folder using huggingface_hub

c2d8a6a verified about 1 year ago

Raw

History Blame Contribute Delete

3.57 kB

Invalid JSON:Unexpected token 'I', ...", Infinity "... is not valid JSON

	{
	"architectures": [
	"NemotronHForCausalLM"
	],
	"attention_bias": false,
	"attention_dropout": 0.0,
	"attention_head_dim": 128,
	"bos_token_id": 1,
	"chunk_size": 256,
	"conv_kernel": 4,
	"eos_token_id": 2,
	"expand": 2,
	"hidden_dropout": 0.0,
	"hidden_size": 4096,
	"hybrid_override_pattern": "M-M-M-M-M-M-M-M-M-M-M-M-M-M-M-M-M-M-M-M-M-M-M-M-",
	"initializer_range": 0.02,
	"intermediate_size": 21504,
	"layer_norm_epsilon": 1e-05,
	"mamba_head_dim": 64,
	"mamba_hidden_act": "silu",
	"mamba_num_heads": 128,
	"mamba_proj_bias": false,
	"max_position_embeddings": 131072,
	"mlp_bias": false,
	"mlp_hidden_act": "relu2",
	"model_type": "nemotron_h",
	"n_groups": 8,
	"num_attention_heads": 32,
	"num_hidden_layers": 52,
	"num_key_value_heads": 8,
	"num_logits_to_keep": 1,
	"pad_token_id": 0,
	"rescale_prenorm_residual": true,
	"residual_in_fp32": false,
	"rms_norm_eps": 1e-05,
	"sliding_window": null,
	"ssm_state_size": 128,
	"tie_word_embeddings": false,
	"time_step_floor": 0.0001,
	"time_step_limit": [
	0.0,
	Infinity
	],
	"time_step_max": 0.1,
	"time_step_min": 0.001,
	"time_step_rank": 256,
	"torch_dtype": "bfloat16",
	"transformers_version": "4.51.3",
	"use_bias": false,
	"use_cache": true,
	"use_conv_bias": true,
	"use_mamba_kernels": true,
	"vocab_size": 131072,
	"quantization_config": {
	"config_groups": {
	"group_0": {
	"input_activations": {
	"dynamic": false,
	"num_bits": 8,
	"type": "float"
	},
	"weights": {
	"dynamic": false,
	"num_bits": 8,
	"type": "float"
	}
	}
	},
	"ignore": [
	"model.layers.backbone.layers.0.mixer.conv1d",
	"model.layers.backbone.layers.11.mixer.conv1d",
	"model.layers.backbone.layers.13.mixer.conv1d",
	"model.layers.backbone.layers.15.mixer.conv1d",
	"model.layers.backbone.layers.17.mixer.conv1d",
	"model.layers.backbone.layers.2.mixer.conv1d",
	"model.layers.backbone.layers.20.mixer.conv1d",
	"model.layers.backbone.layers.22.mixer.conv1d",
	"model.layers.backbone.layers.24.mixer.conv1d",
	"model.layers.backbone.layers.26.mixer.conv1d",
	"model.layers.backbone.layers.28.mixer.conv1d",
	"model.layers.backbone.layers.31.mixer.conv1d",
	"model.layers.backbone.layers.33.mixer.conv1d",
	"model.layers.backbone.layers.35.mixer.conv1d",
	"model.layers.backbone.layers.37.mixer.conv1d",
	"model.layers.backbone.layers.39.mixer.conv1d",
	"model.layers.backbone.layers.4.mixer.conv1d",
	"model.layers.backbone.layers.42.mixer.conv1d",
	"model.layers.backbone.layers.44.mixer.conv1d",
	"model.layers.backbone.layers.46.mixer.conv1d",
	"model.layers.backbone.layers.48.mixer.conv1d",
	"model.layers.backbone.layers.50.mixer.conv1d",
	"model.layers.backbone.layers.6.mixer.conv1d",
	"model.layers.backbone.layers.9.mixer.conv1d",
	"model.layers.lm_head"
	],
	"quant_algo": "FP8",
	"kv_cache_scheme": "FP8",
	"producer": {
	"name": "modelopt",
	"version": "0.30.1.dev44+gd6269109"
	}
	}
	}