Instructions to use WaveCut/YuE2-3B-OrbitQuant-W4A4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use WaveCut/YuE2-3B-OrbitQuant-W4A4 with Transformers:

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

pipe = pipeline("text-generation", model="WaveCut/YuE2-3B-OrbitQuant-W4A4")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("WaveCut/YuE2-3B-OrbitQuant-W4A4", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use WaveCut/YuE2-3B-OrbitQuant-W4A4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "WaveCut/YuE2-3B-OrbitQuant-W4A4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "WaveCut/YuE2-3B-OrbitQuant-W4A4",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/WaveCut/YuE2-3B-OrbitQuant-W4A4

SGLang

How to use WaveCut/YuE2-3B-OrbitQuant-W4A4 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 "WaveCut/YuE2-3B-OrbitQuant-W4A4" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "WaveCut/YuE2-3B-OrbitQuant-W4A4",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "WaveCut/YuE2-3B-OrbitQuant-W4A4" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "WaveCut/YuE2-3B-OrbitQuant-W4A4",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use WaveCut/YuE2-3B-OrbitQuant-W4A4 with Docker Model Runner:
```
docker model run hf.co/WaveCut/YuE2-3B-OrbitQuant-W4A4
```

YuE2-3B-OrbitQuant-W4A4 / src /kernel-source /orbitquant-native /torch-ext /torch_binding.h

WaveCut

Update packed inference to OrbitQuant0.9.6 and native1.0.3

f0c91ec verified 29 days ago

Raw History Blame Contribute Delete

2.45 kB

	#pragma once

	#if defined(CPU_KERNEL)
	#include <torch/csrc/stable/tensor.h>
	using OrbitQuantTensor = torch::stable::Tensor;
	#else
	#include <torch/torch.h>
	using OrbitQuantTensor = torch::Tensor;
	#endif

	void matmul_packed_weight(
	OrbitQuantTensor &out,
	OrbitQuantTensor const &x,
	OrbitQuantTensor const &packed_weight_indices,
	OrbitQuantTensor const &row_norms,
	OrbitQuantTensor const &centroids,
	OrbitQuantTensor const &bias,
	bool has_bias,
	int64_t bits,
	int64_t out_features,
	int64_t in_features,
	int64_t block_m,
	int64_t block_n,
	int64_t block_k);

	#if defined(CPU_KERNEL)
	void quantize_activations_cpu(
	OrbitQuantTensor &out,
	OrbitQuantTensor const &x,
	OrbitQuantTensor const &permutation,
	OrbitQuantTensor const &signs,
	OrbitQuantTensor const &centroids,
	OrbitQuantTensor const &boundaries,
	double eps,
	double inv_sqrt_block,
	int64_t block_size);

	void matmul_packed_adaln_int4_cpu(
	OrbitQuantTensor &out,
	OrbitQuantTensor const &x,
	OrbitQuantTensor const &packed_weight,
	OrbitQuantTensor const &scales,
	OrbitQuantTensor const &bias,
	bool has_bias,
	int64_t out_features,
	int64_t in_features,
	int64_t group_size);
	#endif

	#if defined(CUDA_KERNEL)
	void matmul_packed_w4a4_int8(
	torch::Tensor &out,
	torch::Tensor const &packed_activations,
	torch::Tensor const &packed_weight_indices,
	torch::Tensor const &token_norms,
	torch::Tensor const &row_norms,
	torch::Tensor const &activation_codes,
	torch::Tensor const &weight_codes,
	torch::Tensor const &bias,
	bool has_bias,
	double activation_scale,
	double weight_scale,
	int64_t out_features,
	int64_t in_features,
	int64_t tile_m,
	int64_t tile_n,
	bool async_packed,
	bool weight_k_major);

	void quantize_activations_packed_w4(
	torch::Tensor &packed_out,
	torch::Tensor &norms_out,
	torch::Tensor const &x,
	torch::Tensor const &permutation,
	torch::Tensor const &signs,
	torch::Tensor const &boundaries,
	double eps,
	double inv_sqrt_block,
	int64_t threads);

	void quantize_activations_int8(
	torch::Tensor &int8_out,
	torch::Tensor &norms_out,
	torch::Tensor const &x,
	torch::Tensor const &permutation,
	torch::Tensor const &signs,
	torch::Tensor const &boundaries,
	torch::Tensor const &codes,
	double eps,
	double inv_sqrt_block,
	int64_t threads);
	#endif