Instructions to use kodelow/Hy3-NVFP4-W4A16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kodelow/Hy3-NVFP4-W4A16 with Transformers:

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

pipe = pipeline("text-generation", model="kodelow/Hy3-NVFP4-W4A16")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("kodelow/Hy3-NVFP4-W4A16")
model = AutoModelForCausalLM.from_pretrained("kodelow/Hy3-NVFP4-W4A16")
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 Settings

vLLM

How to use kodelow/Hy3-NVFP4-W4A16 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kodelow/Hy3-NVFP4-W4A16"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kodelow/Hy3-NVFP4-W4A16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/kodelow/Hy3-NVFP4-W4A16

SGLang

How to use kodelow/Hy3-NVFP4-W4A16 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 "kodelow/Hy3-NVFP4-W4A16" \
    --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": "kodelow/Hy3-NVFP4-W4A16",
		"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 "kodelow/Hy3-NVFP4-W4A16" \
        --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": "kodelow/Hy3-NVFP4-W4A16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use kodelow/Hy3-NVFP4-W4A16 with Docker Model Runner:
```
docker model run hf.co/kodelow/Hy3-NVFP4-W4A16
```

olka-fi commited on 7 days ago

Commit

01af353

verified ·

1 Parent(s): ac62ce4

Upload folder using huggingface_hub

Browse files

Files changed (2) hide show

README.md +5 -2
model.safetensors.index.json +1 -1

README.md CHANGED Viewed

@@ -35,10 +35,13 @@ are quantized to 4-bit; everything quality-sensitive stays BF16.
 | | |
 |---|---|
-| **Size** | **169 GB** (down from ~590 GB BF16 source, ~29%) |
 | **Format** | compressed-tensors `nvfp4-pack-quantized`, **weight-only (W4A16)** — E2M1 4-bit weights + FP8-E4M3 group-16 scales + FP32 per-tensor global scale |
 | **Base** | [tencent/Hy3](https://huggingface.co/tencent/Hy3) — 295B MoE (21B active, 3.8B MTP), 80 layers + 1 MTP layer, 192 experts top-8 + 1 shared, 256K context |
 ### What is quantized to what
 | Component | Precision | Why |
@@ -65,7 +68,7 @@ Functional evals run **end-to-end on a single NVIDIA B300 (275 GB)** with the
 - **MTP preserved:** the MTP draft layer (layer 80) is served with
   `--speculative-config '{"method":"mtp","num_speculative_tokens":1}'` and reaches **83.4%**
   draft-token acceptance — lossless (GSM8K unchanged with it on).
-- **Footprint:** ~169 GB — fits a single large-VRAM GPU (verified on one **B300, 275 GB**)
   or multi-GPU tensor parallelism. `fastsafetensors` loads it in ~90 s.
 - **Provenance:** built with [qstream](https://github.com/olka/qstream), experts-only,
   NVFP4 with MSE-optimal FP8-E4M3 group-scale selection (group_size 16).

 | | |
 |---|---|
+| **Size** | **181 GB** (from ~598 GB BF16 source, ~30%) |
 | **Format** | compressed-tensors `nvfp4-pack-quantized`, **weight-only (W4A16)** — E2M1 4-bit weights + FP8-E4M3 group-16 scales + FP32 per-tensor global scale |
 | **Base** | [tencent/Hy3](https://huggingface.co/tencent/Hy3) — 295B MoE (21B active, 3.8B MTP), 80 layers + 1 MTP layer, 192 experts top-8 + 1 shared, 256K context |
+> **Note on HF's param count.** HF reports fewer "params" than the base's 295B because it counts *packed* 4-bit storage elements (each `U8` byte holds two FP4 weights) plus FP8 scales — not logical parameters. The logical model is unchanged: **295B total, 21B active**.
 ### What is quantized to what
 | Component | Precision | Why |
 - **MTP preserved:** the MTP draft layer (layer 80) is served with
   `--speculative-config '{"method":"mtp","num_speculative_tokens":1}'` and reaches **83.4%**
   draft-token acceptance — lossless (GSM8K unchanged with it on).
+- **Footprint:** ~181 GB — fits a single large-VRAM GPU (verified on one **B300, 275 GB**)
   or multi-GPU tensor parallelism. `fastsafetensors` loads it in ~90 s.
 - **Provenance:** built with [qstream](https://github.com/olka/qstream), experts-only,
   NVFP4 with MSE-optimal FP8-E4M3 group-scale selection (group_size 16).

model.safetensors.index.json CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:52945e20b5af6b33476d4f468969855b091afada8edc7463a262f11e483d13bb
 size 13818513

 version https://git-lfs.github.com/spec/v1
+oid sha256:3976e5f5f991872da5b15ac1cbf7656c33ca7ce25a9643be9e5effb5f74adc29
 size 13818513