Instructions to use lthn/lemer-hf-bf16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lthn/lemer-hf-bf16 with Transformers:

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

pipe = pipeline("text-generation", model="lthn/lemer-hf-bf16")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("lthn/lemer-hf-bf16")
model = AutoModelForImageTextToText.from_pretrained("lthn/lemer-hf-bf16")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use lthn/lemer-hf-bf16 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lthn/lemer-hf-bf16"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lthn/lemer-hf-bf16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/lthn/lemer-hf-bf16

SGLang

How to use lthn/lemer-hf-bf16 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 "lthn/lemer-hf-bf16" \
    --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": "lthn/lemer-hf-bf16",
		"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 "lthn/lemer-hf-bf16" \
        --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": "lthn/lemer-hf-bf16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use lthn/lemer-hf-bf16 with Docker Model Runner:
```
docker model run hf.co/lthn/lemer-hf-bf16
```

Lemer-HF-bf16 — HuggingFace safetensors format

LEK-aligned Gemma 4 E2B, bf16 reference weights, in HuggingFace safetensors layout.

Converted from lthn/lemer-mlx-bf16 for use on non-MLX platforms (NVIDIA/AMD GPU, Kaggle TPU, vanilla transformers).

Source	This repo
`lthn/lemer-mlx-bf16` (MLX format)	`lthn/lemer-hf-bf16` (HF safetensors)
`language_model.model.*` keys	`model.language_model.*` keys
MLX conv layout (C,K,I) / (O,H,W,I)	PyTorch layout (C,I,K) / (O,I,H,W)

Weights are byte-equivalent to lemer-mlx-bf16 after the key rename + conv permutation — identical LEK alignment, identical behaviour.

Loading

from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("lthn/lemer-hf-bf16")
model = AutoModelForCausalLM.from_pretrained(
    "lthn/lemer-hf-bf16",
    dtype="bfloat16",
    device_map="auto",
)