Extras -> Single Image Upscale doesn't work for some models

[GooeyAvocado]

Hi Team,

I'm not sure if this is just something wrong with my models but I thought to reach out. It appears the Upscale function doesn't work (Either through the UI or through API)

Steps to Reproduce

1. Navigate to Extras -> Single Image
2. Specify one of the following upscalers a. DAT Models b. ESRGAN or R-ESRGAN models
3. Load an image
4. Hit Generate

Log snippet and traceback

[Unload] Trying to free 1024.00 MB for cuda:0 with 0 models keep loaded ... Current free memory is 5699.35 MB ... Done.
Cleanup minimal inference memory.
tiled upscale: 100%|███████████████████████████████████████████████████████████████████| 36/36 [00:01<00:00, 22.37it/s]
Traceback (most recent call last):
  File "B:\webui_forge_cu121_torch231\webui\modules\call_queue.py", line 76, in f
    res = list(func(*args, **kwargs))
  File "B:\webui_forge_cu121_torch231\webui\modules\call_queue.py", line 55, in f
    res = func(*args, **kwargs)
  File "B:\webui_forge_cu121_torch231\webui\modules\call_queue.py", line 39, in f
    res = func(*args, **kwargs)
  File "B:\webui_forge_cu121_torch231\webui\modules\postprocessing.py", line 135, in run_postprocessing_webui
    return run_postprocessing(*args, **kwargs)
  File "B:\webui_forge_cu121_torch231\webui\modules\postprocessing.py", line 76, in run_postprocessing
    scripts.scripts_postproc.run(initial_pp, args)
  File "B:\webui_forge_cu121_torch231\webui\modules\scripts_postprocessing.py", line 198, in run
    script.process(single_image, **process_args)
  File "B:\webui_forge_cu121_torch231\webui\scripts\postprocessing_upscale.py", line 152, in process
    upscaled_image = self.upscale(pp.image, pp.info, upscaler1, upscale_mode, upscale_by, max_side_length, upscale_to_width, upscale_to_height, upscale_crop)
  File "B:\webui_forge_cu121_torch231\webui\scripts\postprocessing_upscale.py", line 107, in upscale
    image = upscaler.scaler.upscale(image, upscale_by, upscaler.data_path)
  File "B:\webui_forge_cu121_torch231\webui\modules\upscaler.py", line 68, in upscale
    img = self.do_upscale(img, selected_model)
  File "B:\webui_forge_cu121_torch231\webui\modules\realesrgan_model.py", line 48, in do_upscale
    return upscale_with_model(
  File "B:\webui_forge_cu121_torch231\webui\modules\upscaler_utils.py", line 88, in upscale_with_model
    return images.combine_grid(newgrid)
  File "B:\webui_forge_cu121_torch231\webui\modules\images.py", line 120, in combine_grid
    mask_w = make_mask_image(np.arange(grid.overlap, dtype=np.float32).reshape((1, grid.overlap)).repeat(grid.tile_h, axis=0))
TypeError: 'float' object cannot be interpreted as an integer
'float' object cannot be interpreted as an integer

The models tested work on the original A111 UI. This seems to be just a small type error. I would try to fix this myself but I'm not too familiar with python and the codebase.

I can provide more information if it helps.

[SexDrugsAndConspiracyTheories]

Anytime i use highres fix i get black boxes

[MisterChief95]

Hmmm, I am able to upscale just fine with ESRGAN/DAT in the extras tab on the latest version. HiRes Fix works fine too.

Can you share what version of the webUI you're using? Maybe just copy/paste the first 6 or so lines of the terminal output after starting it? Just so we can see UI version/commit hash/Torch version. Additionally, how did you install Forge? A simple git clone or from the packaged release?

@SexDrugsAndConspiracyTheories I recommend opening a separate issue for the Hires black boxes if you haven't already. Just so we can track the two issues in case their root cause isn't related.

[valentin1628]

Hello, exactly same problem with R-ESRGAN 4x+ Anime6B since latest update, really strange.

I use Stability Matrix for run Stable Diffusion