-
Notifications
You must be signed in to change notification settings - Fork 6.5k
Closed
Labels
bugSomething isn't workingSomething isn't working
Description
Describe the bug
I am getting the following error with IMG2IMG
ValueError: At least one stride in the given numpy array is negative, and tensors with negative strides are not currently supported. (You can probably work around this by making a copy of your array with array.copy().)
The help topics online all seem to suggest the same fix with copying the array to fix the memory allocation for the array, I guess.
Full stack:
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
[<ipython-input-16-45b77f776f2b>](https://localhost:8080/#) in <module>
1398 raise SystemExit('\33[33mExecution interrupted by user.\33[0m')
1399 except Exception as e:
-> 1400 raise e
1401 finally:
1402 clean_env()
5 frames
[<ipython-input-16-45b77f776f2b>](https://localhost:8080/#) in <module>
1373 iteration = iteration
1374
-> 1375 do_gobig(path, opts())
1376
1377 NUM_ITERS = original_num_iters
[<ipython-input-1-fff68ce1e364>](https://localhost:8080/#) in do_gobig(gobig_init, opt)
1027 original_init = init.copy()
1028
-> 1029 image, result = diffuse_run()
1030 PROMPT = start_prompt #Fix for Style Prompts and NSP
1031
[<ipython-input-16-45b77f776f2b>](https://localhost:8080/#) in diffuse_run()
524 image = pipeout["sample"][0]
525 except BaseException as e:
--> 526 raise e
527 finally:
528 if pipeout and pipeout['nsfw_content_detected'][0] and ENABLE_NSFW_FILTER:
[<ipython-input-16-45b77f776f2b>](https://localhost:8080/#) in diffuse_run()
514 else:
515 with autocast("cuda"):
--> 516 pipeout = pipe(prompt=PROMPT, negative_prompt=NEG_PROMPT, num_inference_steps=STEPS, init_image=init, strength=INIT_SCALE, guidance_scale=SCALE, generator=gen_seed)
517 image = pipeout["sample"][0]
518 else:
[/usr/local/lib/python3.7/dist-packages/torch/autograd/grad_mode.py](https://localhost:8080/#) in decorate_context(*args, **kwargs)
25 def decorate_context(*args, **kwargs):
26 with self.clone():
---> 27 return func(*args, **kwargs)
28 return cast(F, decorate_context)
29
[/usr/local/lib/python3.7/dist-packages/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py](https://localhost:8080/#) in __call__(self, prompt, init_image, strength, num_inference_steps, guidance_scale, negative_prompt, eta, generator, output_type, return_dict, callback, callback_steps, **kwargs)
308 # Some schedulers like PNDM have timesteps as arrays
309 # It's more optimzed to move all timesteps to correct device beforehand
--> 310 timesteps_tensor = torch.tensor(self.scheduler.timesteps[t_start:], device=self.device)
311
312 for i, t in enumerate(self.progress_bar(timesteps_tensor)):
ValueError: At least one stride in the given numpy array is negative, and tensors with negative strides are not currently supported. (You can probably work around this by making a copy of your array with array.copy().)
Reproduction
Img2img run with negative prompts with DDIM scheduler
Logs
No response
System Info
Google Colab Env | Python 3.7 | Latest pull as of this morning (since the new merge of negative prompts broke my patches for negative prompts).
Metadata
Metadata
Assignees
Labels
bugSomething isn't workingSomething isn't working