RuntimeError: "slow_conv2d_cpu" not implemented for 'Half'

[v-puppala]

I ran the short code snippet in the readme and ran it on my own colab notebook. I received an error on the final line (result=whisper.decode(model, mel, options)

Here's error message and traceback:

RuntimeError Traceback (most recent call last)
in
17 # decode the audio
18 options = whisper.DecodingOptions()
---> 19 result = whisper.decode(model, mel, options)
20
21 # print the recognized text

10 frames
/usr/local/lib/python3.7/dist-packages/torch/nn/modules/conv.py in _conv_forward(self, input, weight, bias)
302 _single(0), self.dilation, self.groups)
303 return F.conv1d(input, weight, bias, self.stride,
--> 304 self.padding, self.dilation, self.groups)
305
306 def forward(self, input: Tensor) -> Tensor:

RuntimeError: "slow_conv2d_cpu" not implemented for 'Half

[Batuhanapa]

18 options = whisper.DecodingOptions(fp16 = False), it worked for me

[kumbarisagar]

i actually got the same error because when i turned float 32 operations into float16 operations

[kumbarisagar]

how to solve it??\

[marinone94]

Are you running on CPU?

[Batuhanapa]

how to solve it??\

İf u wanna use float16 type u should work on GPU

[kumbarisagar]

Yes , i am running on CPU on raspberry pi

…

On Fri, Oct 7, 2022, 12:10 PM Batuhanapa ***@***.***> wrote: how to solve it??\ İf u wanna use float16 type u should work on GPU — Reply to this email directly, view it on GitHub <#92 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOJYIZ7AZKE7GXJQEJGYUJ3WB7AV3ANCNFSM6AAAAAAQUQPDLA> . You are receiving this because you commented.Message ID: ***@***.***>

[megatron100]

I am interested in your project. May I know why raspberry pi, are using it to record conversation?

[ooluwaseunabel]

I ran into the same error. I was able to resolve it after changing this line:
options = whisper.DecodingOptions()

to
options = whisper.DecodingOptions(fp16 = False)

[Randalf96]

thanks man it worked for me 2 !!!! life saver

[ideelahmed]

thank you for blessing us with your awesomeness. Sh*t works!

[StepanTrnkaCNC]

Where did you change those options, please?

[chazmofdoom]

Thank you! This is the solution!

[clevadani]

Worked like a charm.

[halqadasi]

I added this option (--fp32) and it works for me, for example:

!python inference_realesrgan.py -n realesr-general-x4v3 -i upload -g 0 -dn 0.2 --fp32

[andrez4vz]

Great! It's works. Thanks man!

[AbdulHaseeb007]

updating ultralytics to 8.0.109 also works, just in case u have changed your ultralytics version