Instructions to use learning-unit/L1-16B-A3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use learning-unit/L1-16B-A3B with Transformers:

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

pipe = pipeline("text-generation", model="learning-unit/L1-16B-A3B", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("learning-unit/L1-16B-A3B", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use learning-unit/L1-16B-A3B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "learning-unit/L1-16B-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": "learning-unit/L1-16B-A3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/learning-unit/L1-16B-A3B

SGLang

How to use learning-unit/L1-16B-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 "learning-unit/L1-16B-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": "learning-unit/L1-16B-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 "learning-unit/L1-16B-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": "learning-unit/L1-16B-A3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use learning-unit/L1-16B-A3B with Docker Model Runner:
```
docker model run hf.co/learning-unit/L1-16B-A3B
```

jeffkang-lunit commited on Apr 8

Commit

ea1cb07

verified ·

1 Parent(s): dbb1dac

Upload README.md with huggingface_hub

Browse files

Files changed (1) hide show

README.md +69 -3

README.md CHANGED Viewed

@@ -1,3 +1,69 @@
----
-license: apache-2.0
----

+---
+license: apache-2.0
+language:
+  - en
+tags:
+  - medical
+  - clinical
+  - moe
+  - mixture-of-experts
+  - gravity-moe
+  - sft
+library_name: transformers
+pipeline_tag: text-generation
+---
+# L1
+L1 is a clinical language model built on the **GravityMoE** (Mixture-of-Experts) architecture, fine-tuned for medical and clinical decision support tasks.
+## Model Details
+| Property | Value |
+|---|---|
+| Architecture | GravityMoE (Mixture-of-Experts) |
+| Total Parameters | ~16B |
+| Active Parameters | ~4.5B per token |
+| Routed Experts | 64 |
+| Shared Experts | 1 |
+| Experts per Token | 8 |
+| Hidden Size | 2048 |
+| Layers | 28 |
+| Attention Heads | 16 |
+| KV LoRA Rank | 512 |
+| Max Context Length | 32,768 tokens |
+| Precision | bfloat16 |
+| Vocab Size | 151,552 |
+## Usage
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+model_name = "learning-unit/L1"
+tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    torch_dtype="auto",
+    device_map="auto",
+    trust_remote_code=True,
+)
+messages = [
+    {"role": "user", "content": "What are the diagnostic criteria for sepsis?"}
+]
+inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True)
+inputs = inputs.to(model.device)
+outputs = model.generate(inputs, max_new_tokens=512)
+print(tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True))
+```
+## Training
+- **Method**: Supervised Fine-Tuning (SFT)
+- **Epochs**: 3
+- **Final Training Loss**: 0.247
+## License
+Apache 2.0