wejoncy

**Description**: Describe your changes. Xnnpack don't expect an external constant initializer. We have a helper class to unpack initializer_tensor. **Motivation and Context** - Why is this change required? What problem...

**Description**: Describe your changes. **Motivation and Context** - Why is this change required? What problem does it solve? - If it fixes an open issue, please link to the issue...

[xnnpack] Add EP Doc

1

**Description**: Describe your changes. Xnnpack EP Preview: https://wejoncy.github.io/onnxruntime/ **Motivation and Context** - Why is this change required? What problem does it solve? - If it fixes an open issue, please...

### Description 1. Support quantized GPTQ weight in huggingface like [TheBloke/Llama-2-7B-Chat-GPTQ](https://huggingface.co/TheBloke/Llama-2-7B-Chat-GPTQ) 2. Support Act_order for GPTQ 3. Support [HQQ](https://mobiusml.github.io/hqq_blog/) algorithm to quantize matmul weight and add quant script ### Motivation...

- cast - argmax - gelu - cast - LayerNorm - GroupNorm - InstanceNorm ### Description ### Motivation and Context

`preload_module_classes` is lost in attach_execution_device_hook.

4

### System Info https://github.com/huggingface/accelerate/blob/3fcc9461c4fcb7228df5e5246809ba09cfbb232e/src/accelerate/hooks.py#L439 Should we pass preload_module_classes to the next calling? such as ``` attach_execution_device_hook(child, execution_device, tied_params_map=tied_params_map, preload_module_classes=preload_module_classes) ``` ### Information - [ ] The official example scripts -...

Resize in FP16

1

According to https://onnx.ai/onnx/operators/onnx__Resize.html#resize its input `scale` is force to float32. However, `convert_float_to_float16` doesn't handle it.

[Resize] which conversion mode should be taken in RESIZE output_dim computation, Floor or Round?

7

# Ask a Question output_dim computation is not unified on how to convert float to int. Is this a bug? ### Question In Resize-18, `keep_aspect_ratio_policy ` has three different options,...

https://github.com/wejoncy/sfllm Hi This can be worked on Windows/Linux/Macos or any torch compatiable OS. For Learning purpose but with impressive performance.