Pinned Loading

1. Qwen-Image-2.1-Accel Qwen-Image-2.1-Accel Public

   Accelerate Qwen-Image-2.1 text-to-image generation and image editing on NVIDIA RTX 5090. BF16 Triton kernels, optional MXFP8 + SageAttention2, Python API, CLI, and reproducible benchmarks.

   Python

2. Qwen-Image-2.1-Coreml Qwen-Image-2.1-Coreml Public

   Run Qwen-Image-2.1 locally on Apple silicon Macs with Core ML. 1024x1024 text-to-image, FP16 models and Python inference. Measured 2.4-2.6x faster denoising steps vs PyTorch bf16/MPS on M5 (32 GB).…

   Python

3. DeepSeek-V4.1-Flash-Accel DeepSeek-V4.1-Flash-Accel Public

   Run DeepSeek-V4.1-Flash on 8x RTX 5090 + 503 GiB RAM. 5.6x output throughput vs patched eager mode, with vLLM fixes, CPU offload & reproducible benchmarks.

   Python 1

4. onnx2coreml onnx2coreml Public

   Convert ONNX models to Apple Core ML (.mlpackage / .mlmodel) with numerical-parity verification against ONNX Runtime.

   Python 153

5. coreai-onnx coreai-onnx Public

   Convert ONNX models directly to Apple's Core AI (.aimodel) format, the next-generation successor to Core ML.

   Python 34

6. Depth-Anything-3-Onnx Depth-Anything-3-Onnx Public

   Depth-Anything-3-Onnx

   Python 23