"ValueError: Stop argument for islice() must be None or an integer: 0 <= x <= sys.maxsize."

[jnc-nj]

def restore_func(images, restorer):
    for idx in range(len(images)):
        img = cv2.cvtColor(_img, cv2.COLOR_RGB2BGR)
        # restore faces and background if necessary
        cropped_faces, restored_faces, r_img = restorer.enhance(
            img,
            has_aligned=False,
            only_center_face=False,
            paste_back=True)
        r_img = cv2.cvtColor(r_img, cv2.COLOR_BGR2RGB)
        yield r_img

def accum_func(r_img):
    return r_img

job = Parallel(n_jobs=batch_size/2, return_as='generator', pre_dispatch='1.5*n_jobs', verbose=10)(delayed(accum_func)(r_img) for r_img in restore_func(images, restorer))`

I run into the following error for the above code:

File "/workspace/SadTalker/src/utils/face_enhancer.py", line 41, in enhancer_list
   gen = enhancer_generator_no_len(images, method=method, bg_upsampler=bg_upsampler, batch_size=batch_size)
 File "/workspace/SadTalker/src/utils/face_enhancer.py", line 129, in enhancer_generator_no_len
   job = Parallel(n_jobs=batch_size/2, return_as='generator', pre_dispatch='1.5*n_jobs', verbose=10)(delayed(accum_func)(r_img) for r_img in restore_func(images, restorer))
 File "/usr/local/lib/python3.8/dist-packages/joblib/parallel.py", line 1942, in __call__
   next(output)
 File "/usr/local/lib/python3.8/dist-packages/joblib/parallel.py", line 1580, in _get_outputs
   self._start(iterator, pre_dispatch)
 File "/usr/local/lib/python3.8/dist-packages/joblib/parallel.py", line 1563, in _start
   if self.dispatch_one_batch(iterator):
 File "/usr/local/lib/python3.8/dist-packages/joblib/parallel.py", line 1426, in dispatch_one_batch
   islice = list(itertools.islice(iterator, big_batch_size))
ValueError: Stop argument for islice() must be None or an integer: 0 <= x <= sys.maxsize.

The base examples here work fine, which is quite confusing considering i'm following the sample code here

Any advice would be appreciated. Thanks in advance!

[jnc-nj]

solved, silent failure when gpu ooms.
reducing batch_size fixes this.

[tomMoral]

Thanks for the report. I think this is related to the behavior reported in #1489? The error reporting should be improved with #1491 hopefully.