Instructions to use nvidia/Nemotron-Flash-3B-Instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nvidia/Nemotron-Flash-3B-Instruct with Transformers:

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

pipe = pipeline("text-generation", model="nvidia/Nemotron-Flash-3B-Instruct", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("nvidia/Nemotron-Flash-3B-Instruct", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use nvidia/Nemotron-Flash-3B-Instruct with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/nvidia/Nemotron-Flash-3B-Instruct

SGLang

How to use nvidia/Nemotron-Flash-3B-Instruct 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-Flash-3B-Instruct" \
    --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": "nvidia/Nemotron-Flash-3B-Instruct",
		"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 "nvidia/Nemotron-Flash-3B-Instruct" \
        --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": "nvidia/Nemotron-Flash-3B-Instruct",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use nvidia/Nemotron-Flash-3B-Instruct with Docker Model Runner:
```
docker model run hf.co/nvidia/Nemotron-Flash-3B-Instruct
```

Nemotron-Flash-3B-Instruct / config.json

YongganFu

Upload NemotronFlashForCausalLM

81a7d5f verified 6 months ago

raw

history blame contribute delete

2.35 kB

	{
	"architectures": [
	"NemotronFlashForCausalLM"
	],
	"attention_dropout": 0.0,
	"attn_hidden_size": -1,
	"attn_implementation": "fused_mha",
	"attn_implementation_new": "fused_mha",
	"auto_map": {
	"AutoConfig": "configuration_nemotron_flash.NemotronFlashConfig",
	"AutoModelForCausalLM": "modeling_nemotron_flash.NemotronFlashForCausalLM"
	},
	"bos_token_id": 1,
	"calc_logits_for_entire_prompt": false,
	"d_conv": 4,
	"dtype": "bfloat16",
	"eos_token_id": 2,
	"ffn_expand_ratio": 3,
	"global_attn_idx": [],
	"hidden_act": "silu",
	"hidden_size": 3072,
	"hybrid_decoder_layer": "mamba",
	"initializer_range": 0.02,
	"intermediate_size": 0,
	"kq_head_dim": -1,
	"kq_norm": "none",
	"layer_type": [
	"m",
	"a",
	"m",
	"a",
	"a",
	"a",
	"m",
	"a",
	"m",
	"a",
	"m",
	"a",
	"a",
	"a",
	"m",
	"a",
	"m",
	"a",
	"m",
	"a",
	"a",
	"a",
	"m",
	"a",
	"m",
	"a",
	"m",
	"a",
	"m",
	"a",
	"m",
	"a",
	"m",
	"a",
	"m",
	"a"
	],
	"layer_types": [
	"deltanet",
	"f",
	"m2",
	"f",
	"a",
	"f",
	"m2",
	"f",
	"deltanet",
	"f",
	"m2",
	"f",
	"a",
	"f",
	"m2",
	"f",
	"deltanet",
	"f",
	"m2",
	"f",
	"a",
	"f",
	"m2",
	"f",
	"deltanet",
	"f",
	"m2",
	"f",
	"deltanet",
	"f",
	"m2",
	"f",
	"deltanet",
	"f",
	"m2",
	"f"
	],
	"mamba2_headdim": 64,
	"mamba_conv_bias": true,
	"mamba_d_conv": 4,
	"mamba_d_state": 128,
	"mamba_dt_rank": 192,
	"mamba_expand": 2,
	"mamba_inner_layernorms": true,
	"mamba_proj_bias": false,
	"max_position_embeddings": 29000,
	"mlp_hidden_act": "silu",
	"model_type": "nemotron_flash",
	"new_seq_length": 2048,
	"num_attention_heads": 24,
	"num_experts": 1,
	"num_experts_per_tok": 1,
	"num_hidden_layers": 36,
	"num_key_value_heads": 6,
	"num_memory_tokens": 256,
	"orig_max_position_embeddings": 4096,
	"output_router_logits": false,
	"pad_token_id": 0,
	"rms_norm_eps": 1e-06,
	"rope": true,
	"rope_theta": 10000.0,
	"rope_type": "ntk",
	"router_aux_loss_coef": 0.001,
	"sliding_window": null,
	"tie_word_embeddings": true,
	"transformers_version": "4.56.2",
	"use_cache": false,
	"use_mamba_kernels": true,
	"v_head_dim": -1,
	"vocab_size": 131075
	}