-
Hi
I was looking at the various versions you have in the `pypi_packages` folder and noticed that the order of computation of weight decay (which for some options modifies `grad`) and of `grad_residua…
-
I encountered an issue when trying to export the facebook/m2m100_418M model using the optimum-cli tool. The error message indicates that the m2m-100-encoder is not supported, despite m2m-100 being lis…
-
Testing the inference numbers
-
## ❓ Question
I created a script to compare inference runtimes with `torch`, `torch.compile` and `torch_tensorrt.compile` for any timm model, input shape and dtype and some runtimes are worse using…
-
### SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers
基于transformer的语义分割模型
[论文地址](https://arxiv.org/abs/2105.15203)
 cyx@master:/cluster/data7a/cyx/jx/ClearCLIP$ python3 eval_all.py 2>&1 | tee reproduce.log
/cluster/home2/cyx/miniconda3/envs/Cle…
-
I have clean new comfy ui install for LVCD.
But i cant run sample LVCD workflow. I get erorr on loading module. I tried to fix node.py usng chatGPT fixed some parts created new problems....
Loa…
-
### Description of the bug | 错误描述
处理表格报错，参考issue#659处理问题仍然无法成功
2024-10-10 06:01:02.508 | INFO | magic_pdf.libs.pdf_check:detect_invalid_chars:57 - cid_count: 0, text_len: 1303, cid_chars_radio: …
-
When using Flash-Attention version 2.6.3, there is an issue with the ONNX file saved using torch.onnx.export.
code:
import sys
import torch
qkv=torch.load("/home/qkv.pth")
from modeling_inte…