Sizes of tensors must match except in dimension 0

Question

Sizes of tensors must match except in dimension 0

matthewchung74 opened this issue 4 years ago · comments

Hi @muellerzr , in Lesson_1, on this line
dbunch.show_batch(max_n=9, figsize=(6,7))

I'm getting this error

RuntimeError                              Traceback (most recent call last)
<ipython-input-18-5a9f6693bc49> in <module>()
----> 1 dbunch.show_batch(max_n=9, figsize=(6,7))

12 frames
/usr/local/lib/python3.6/dist-packages/torch/utils/data/_utils/collate.py in default_collate(batch)
     53             storage = elem.storage()._new_shared(numel)
     54             out = elem.new(storage)
---> 55         return torch.stack(batch, 0, out=out)
     56     elif elem_type.__module__ == 'numpy' and elem_type.__name__ != 'str_' \
     57             and elem_type.__name__ != 'string_':

RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 0. Got 180 and 403 in dimension 2 at /pytorch/aten/src/TH/generic/THTensor.cpp:689

Do you know if something in the API changed? I'm running a copy of your Lesson_1 in a colab. Also, thanks a ton for putting this together.

Zach Mueller · Answer 1 · Tue Jan 28 2020 10:38:12 GMT+0800 (China Standard Time)

Thank you for pointing this out. Yes this is a bug in the library. I'll investigate if it's present in the current pip version too. I've made the dev's aware of this

YHat.pub · Answer 2 · Tue Jan 28 2020 10:43:31 GMT+0800 (China Standard Time)

Awesome. Thanks!

Zach Mueller · Answer 3 · Tue Jan 28 2020 10:53:52 GMT+0800 (China Standard Time)

I've limited the install of fastai2. Once this issue is fixed in the main version we will use the newest versions.

YHat.pub · Answer 4 · Tue Jan 28 2020 14:06:03 GMT+0800 (China Standard Time)

Thanks. Do you know if there there is a workaround so I can finish working through the notebook?

On Mon, Jan 27, 2020 at 6:53 PM Zachary Mueller ***@***.***> wrote: Closed #3 <#3> . — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#3?email_source=notifications&email_token=AAM3RRE2ESAEE6GXOWDT4VTQ76M4FA5CNFSM4KMLLMSKYY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGOWHO7ZDQ#event-2984115342>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAM3RRGB6WFAT6D4IBVBFTLQ76M4FANCNFSM4KMLLMSA> .

-- Regards, Matt

Zach Mueller · Answer 5 · Tue Jan 28 2020 14:07:15 GMT+0800 (China Standard Time)

Everything is fixed now in all the notebooks using the most recent version (not the old folders though)