akshay6893

When you put `--listen` the process is launched on `0.0.0.0` which means any one can access it and it listens to all. But without `--listen` process is launched on `127.0.0.1`...

`docker run -it -p 7861:7860` - this will launch on public network 0.0.0.0:7861. TO launch on local host - `docker run -it -p 127.0.0.1:7861:7860`

There is [cococo](https://cococozibojia.github.io/), which does the same in-painting and they claim that they are better than AVID or at least equal to it.

I also have similar problem. With gpu, the model (in detection.py) - `with torch.inference_mode(): pred = model(pixel_values=batch)` ouputs a tensor with nan values. This in turn results in `list out...

It turns out that the layer - `hidden_states = self.sr(hidden_states)` where, ``` self.sr = nn.Conv2d( hidden_size, hidden_size, kernel_size=sequence_reduction_ratio, stride=sequence_reduction_ratio ) ``` in the class `SegformerEfficientSelfAttention` was the problem. It is...