Instructions to use sapidlabs/Hy3-REAP-48e with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use sapidlabs/Hy3-REAP-48e with Transformers:

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

pipe = pipeline("text-generation", model="sapidlabs/Hy3-REAP-48e")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("sapidlabs/Hy3-REAP-48e")
model = AutoModelForCausalLM.from_pretrained("sapidlabs/Hy3-REAP-48e")
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 sapidlabs/Hy3-REAP-48e with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "sapidlabs/Hy3-REAP-48e"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "sapidlabs/Hy3-REAP-48e",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/sapidlabs/Hy3-REAP-48e

SGLang

How to use sapidlabs/Hy3-REAP-48e 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 "sapidlabs/Hy3-REAP-48e" \
    --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": "sapidlabs/Hy3-REAP-48e",
		"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 "sapidlabs/Hy3-REAP-48e" \
        --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": "sapidlabs/Hy3-REAP-48e",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use sapidlabs/Hy3-REAP-48e with Docker Model Runner:
```
docker model run hf.co/sapidlabs/Hy3-REAP-48e
```

Hy3-REAP-48e — REAP-pruned Hunyuan Hy3 (experimental)

An expert-pruned version of tencent/Hy3, produced with REAP (Router-weighted Expert Activation Pruning). Each MoE layer's 192 routed experts are reduced to the 48 most salient, shrinking the checkpoint from 557 GB to 157 GB (BF16).

⚠️ Experimental research artifact. This is an aggressive prune (75% of experts removed) calibrated on a small, code-only set. It generates fluent text and is genuinely strong at code, but general/factual accuracy is degraded (see below). Not intended for production. A better-calibrated version is planned — this repo will be updated in place.

	Base	This model
Routed experts / MoE layer	192	48
Experts per token	8	8 (unchanged)
Size (BF16)	~557 GB	~157 GB
Architecture	`hy_v3` (`HYV3ForCausalLM`)	same

How it was made

The entire pipeline ran on a single NVIDIA DGX Spark (128 GB) — a model far larger than RAM — using per-layer disk streaming:

Method: REAP, one-shot (no retraining). Each expert is scored by the mean of router_weight × activation_norm over a calibration set; the least-salient are dropped.
Compression: compression_ratio 0.75 → 192 → 48 experts kept per MoE layer.
Calibration: 64 samples from theblackcat102/evol-codealpaca-v1 (code).
Streaming: the 557 GB model was built on meta and each of the 80 layers streamed from the on-disk safetensors shards (~7 GB resident at a time), for both calibration and the prune. See the How To Spark training notes for the recipe.

Quality profile (sanity check)

Fluent everywhere; quality tracks the (code-only) calibration domain:

Prompt	Output
Write an nth-Fibonacci function	Correct, well-documented Python ✅
"The sky is…"	"Blue." ✅
Capital of France?	"France does not have a single capital city…" ❌ (fluent but wrong)

The pattern is expected: code experts (the calibration domain) survived; experts carrying broad world-knowledge scored low on code data and were pruned. A larger, more diverse calibration set and/or a gentler prune ratio would recover accuracy.

Running it

At 157 GB BF16 this exceeds a single 128 GB device. Options:

Two GPUs / two nodes — e.g. vllm serve <repo> --pipeline-parallel-size 2 (vLLM supports HYV3ForCausalLM). This model was validated running pipeline-parallel across two DGX Sparks.
4-bit quantization (~40 GB) to fit a single device (tooling for hy_v3 quantization is still maturing).

License

Inherits the base model's license (Apache-2.0). Derivative of tencent/Hy3.

Downloads last month: 184

Safetensors

Model size

84B params

Tensor type

BF16

F32

Model tree for sapidlabs/Hy3-REAP-48e

Base model

tencent/Hy3

Finetuned

(1)

this model