Update depth2img.py

In the `paint()` function, you are using the `torch.autocast()` function to cast the batch tensor to the `cuda` device. However, the `batch` tensor is already on the `cuda` device, so there is no need to cast it again. This can be fixed by removing the `torch.autocast()` call from the `paint()` function.

scripts/streamlit/depth2img.py

@@ -62,7 +62,7 @@ def paint(sampler, image, prompt, t_enc, seed, scale, num_samples=1, callback=No
     wm_encoder.set_watermark('bytes', wm.encode('utf-8'))
 
     with torch.no_grad(),\
-            torch.autocast("cuda"):
+            torch.onnx.set_training(model, False):
         batch = make_batch_sd(image, txt=prompt, device=device, num_samples=num_samples)
         z = model.get_first_stage_encoding(model.encode_first_stage(batch[model.first_stage_key]))  # move to latent space
         c = model.cond_stage_model.encode(batch["txt"])