QingshuiL

I tried using smbclient to load PyTorch model weights from a server, but I encountered some issues. ```python file_path = r"\\x.x.x.x\home\pytorch-models\resnet18-f37072fd.pth" with smbclient.open_file(file_path, mode='rb', buffering=0) as fd: data = torch.load(fd,...

How can I make my metric depth estimation results more accurate?

2

I use the indoor and outdoor models for inference on my dataset, but the real depth estimates are not accurate. Now my dataset doesn't have GT and can't do finetune...

请问如何复现 InternVideo2s1-1B在 K400 K600 K700 数据集上的评估精度？

2

我在 single_modality 和 multi_modality 目录下没有找到对应的评测脚本，我应该如何复现 InternVideo2s1-1B 在 K400 K600 K700 三个数据集上的精度？

only support hopper GPU?

3

Very good work, but I have some questions to consult. When I tried to run the code, I encountered the following error. [rank0]: Traceback (most recent call last): [rank0]: File...

**When I perform the multi-GPU evaluation for DeepSeekv3, the following error message is displayed:** 2025-02-21 16:29:40.579 | INFO | llmc.eval.eval_base:__init__:21 - eval_cfg : {'eval_pos': ['pretrain', 'transformed', 'fake_quant'], 'name': 'wikitext2', 'download':...

请教pytorch版本的精度损失

4

想把paddleOCR模型部署，正头疼的是否看到了作者的这个库，感觉发现了宝藏。不知道作者复现paddleOCR的pytorch版本以后是否有测试下模型的精度，相比原版有没有精度损失，损失大不大？

我尝试调用 llm_exporter.export() 接口，遇到了 issues#3 同样的问题。 然后我添加如下代码： self.image_start_id = self.config.im_start_token self.blocks = torch.nn.ModuleList(self.blocks) 遇到了如下问题： ValueError: Attention weights should be of size (tensor(1), 16, tensor(3), tensor(19)), but is torch.Size([1, 16, 3, 3])...