Instructions to use trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning with Transformers:

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

pipe = pipeline("text-generation", model="trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning")
model = AutoModelForCausalLM.from_pretrained("trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning", device_map="auto")
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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning

SGLang

How to use trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning 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 "trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning" \
    --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": "trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning",
		"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 "trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning" \
        --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": "trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning with Docker Model Runner:
```
docker model run hf.co/trl-internal-testing/tiny-NemotronHForCausalLM-3.5-lightning
```

albertvillanova HF Staff commited on 9 days ago

Commit

a6235a3

verified ·

1 Parent(s): 650f5b8

Upload NemotronHForCausalLM

Browse files

Files changed (2) hide show

config.json +7 -4
model.safetensors +1 -1

config.json CHANGED Viewed

@@ -26,7 +26,7 @@
   "mamba_num_heads": 8,
   "mamba_proj_bias": false,
   "mamba_ssm_cache_dtype": "float32",
-  "max_position_embeddings": 4096,
   "mlp_bias": false,
   "mlp_hidden_act": "relu2",
   "model_type": "nemotron_h",
@@ -42,16 +42,19 @@
   "n_groups": 1,
   "n_routed_experts": 4,
   "n_shared_experts": 1,
   "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,
   "rescale_prenorm_residual": true,
   "residual_in_fp32": false,
-  "routed_scaling_factor": 1.0,
   "sliding_window": null,
   "ssm_state_size": 16,
   "tie_word_embeddings": false,
@@ -69,6 +72,6 @@
   "use_bias": false,
   "use_cache": true,
   "use_conv_bias": true,
-  "use_mamba_kernels": false,
   "vocab_size": 131072
 }

   "mamba_num_heads": 8,
   "mamba_proj_bias": false,
   "mamba_ssm_cache_dtype": "float32",
+  "max_position_embeddings": 262144,
   "mlp_bias": false,
   "mlp_hidden_act": "relu2",
   "model_type": "nemotron_h",
   "n_groups": 1,
   "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": 1,
   "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,
   "use_bias": false,
   "use_cache": true,
   "use_conv_bias": true,
+  "use_mamba_kernels": true,
   "vocab_size": 131072
 }

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4fbfb8f5b85606889977cef4d24842c831e8e7343dac5a3b2ae5c10b392bc30d
 size 8456608

 version https://git-lfs.github.com/spec/v1
+oid sha256:7345b9e98a09a15c88cff7a684de680348da8ac8aa5f12337f48b8255025fbf4
 size 8456608