Instructions to use onnx-internal-testing/tiny-random-NemotronHForCausalLM with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use onnx-internal-testing/tiny-random-NemotronHForCausalLM with Transformers:

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

pipe = pipeline("text-generation", model="onnx-internal-testing/tiny-random-NemotronHForCausalLM", 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("onnx-internal-testing/tiny-random-NemotronHForCausalLM", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("onnx-internal-testing/tiny-random-NemotronHForCausalLM", 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 onnx-internal-testing/tiny-random-NemotronHForCausalLM with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/onnx-internal-testing/tiny-random-NemotronHForCausalLM

SGLang

How to use onnx-internal-testing/tiny-random-NemotronHForCausalLM 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 "onnx-internal-testing/tiny-random-NemotronHForCausalLM" \
    --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": "onnx-internal-testing/tiny-random-NemotronHForCausalLM",
		"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 "onnx-internal-testing/tiny-random-NemotronHForCausalLM" \
        --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": "onnx-internal-testing/tiny-random-NemotronHForCausalLM",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use onnx-internal-testing/tiny-random-NemotronHForCausalLM with Docker Model Runner:
```
docker model run hf.co/onnx-internal-testing/tiny-random-NemotronHForCausalLM
```

tiny-random-NemotronHForCausalLM / config.json

Xenova HF Staff

Upload optimized ONNX model (#1)

c6c9c3d 2 months ago

raw

history blame contribute delete

2.2 kB

	{
	"architectures": [
	"NemotronHForCausalLM"
	],
	"attention_bias": false,
	"attention_dropout": 0.0,
	"auto_map": {
	"AutoConfig": "configuration_nemotron_h.NemotronHConfig",
	"AutoModel": "modeling_nemotron_h.NemotronHForCausalLM",
	"AutoModelForCausalLM": "modeling_nemotron_h.NemotronHForCausalLM"
	},
	"bos_token_id": 1,
	"chunk_size": 128,
	"conv_kernel": 4,
	"dtype": "float32",
	"eos_token_id": 2,
	"expand": 2,
	"head_dim": 16,
	"hidden_dropout": 0.0,
	"hidden_size": 64,
	"initializer_range": 0.02,
	"intermediate_size": 32,
	"layer_norm_epsilon": 1e-05,
	"layers_block_type": [
	"mamba",
	"attention",
	"moe"
	],
	"mamba_head_dim": 16,
	"mamba_hidden_act": "silu",
	"mamba_num_heads": 4,
	"mamba_proj_bias": false,
	"mamba_ssm_cache_dtype": "float32",
	"max_position_embeddings": 262144,
	"mlp_bias": false,
	"mlp_hidden_act": "relu2",
	"model_type": "nemotron_h",
	"moe_intermediate_size": 16,
	"moe_latent_size": null,
	"moe_shared_expert_intermediate_size": 16,
	"moe_shared_expert_overlap": true,
	"mtp_layers_block_type": [
	"attention",
	"moe"
	],
	"n_group": 2,
	"n_groups": 2,
	"n_routed_experts": 4,
	"n_shared_experts": 1,
	"norm_eps": 1e-05,
	"norm_topk_prob": true,
	"num_attention_heads": 4,
	"num_experts_per_tok": 2,
	"num_key_value_heads": 2,
	"num_logits_to_keep": 1,
	"num_nextn_predict_layers": 0,
	"pad_token_id": 0,
	"partial_rotary_factor": 1.0,
	"rescale_prenorm_residual": true,
	"residual_in_fp32": false,
	"rope_theta": 10000,
	"routed_scaling_factor": 2.5,
	"sliding_window": null,
	"ssm_state_size": 16,
	"tie_word_embeddings": false,
	"time_step_floor": 0.0001,
	"time_step_limit": [
	0.0,
	{
	"__float__": "Infinity"
	}
	],
	"time_step_max": 0.1,
	"time_step_min": 0.001,
	"topk_group": 1,
	"transformers_version": "5.3.0.dev0",
	"use_bias": false,
	"use_cache": true,
	"use_conv_bias": true,
	"use_mamba_kernels": true,
	"vocab_size": 131072,
	"transformers.js_config": {
	"use_external_data_format": {
	"model.onnx": 1,
	"model_fp16.onnx": 1
	},
	"kv_cache_dtype": {
	"q4f16": "float16",
	"fp16": "float16"
	}
	}
	}