xiguadong

1. update new dispatchers for quantizing all ops in quantable ops, this is used for npu backend 2. update prelu

Why the shape between pretrianed weights and config is different?

3

the output info as belows: Use load_from_local loader The model and loaded state dict do not match exactly size mismatch for keypoint_head.final_layer.weight: copying a param with shape torch.Size([16, 30, 1,...

hello，thanks for your wonderful work.But I have a question is that Why the new branch derives from the "upper" branch only in code? In paper ,the dervies structure is from...

求增加 插入自定义算子 / 自由编辑算子 attribute的功能

5

希望大佬能够支持，祝大佬bug少少，工资多多

您好，观察到 imagenet的top1 acc 精度较低，请问下是如何测试的呢？ 在 https://github.com/openai/CLIP/blob/main/notebooks/Prompt_Engineering_for_ImageNet.ipynb 这里提到，text_encoder编码时有加了80组的promt template，最后能够达到 76.2% top1 acc。想请问下cn-clip在测试时有采用相同的 trick嘛？

请问下 2.5什么时候开源呀？

1

嗷嗷待开源

可以提供一个 requirements.txt吗？

24

如题，装环境装的头疼

[Question] What's the groupsize of w4a16 + w8a16

1

Hello , in the https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GPTQ-Int4/blob/c34a4a91629f09f73a285f32dbd26106b033c654/config.json#L29 has mentioned the groupsize is 128 for 4bit or 8bit. So could you tell me the groupsize for this model? And If I want to...

in https://github.com/onnx/onnx/blob/main/docs/Operators.md#Pad , example mode=reflect as belows ``` data = [ [1.0, 1.2], [2.3, 3.4], [4.5, 5.7], ] pads = [0, 2, 0, 0] mode = 'reflect' output = [...