-
Notifications
You must be signed in to change notification settings - Fork 233
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
MaskedLMHead should support dtype=bfloat16 #1195
Comments
hi I have developed a solution to your problem |
This is fixed on #1242, and we have a test enforcing this for all layers. |
Describe the bug
I claim that MaskedLMHead should support a dtype argument of tf.bfloat16 (and tf.float16) so that users can look at the effect of reducing their memory usage. This matters more as the vocab gets larger.
To Reproduce
In google corp colab I do "File -> Save a copy as GitHub Gist", enter an OTP, and then there is a message that "github auth fails" so I'll just include the code inline:
Expected behavior
Lack of a crash.
The loop checking dtypes (assert w.dtype == dtype, ("Wrong type: " + w.name)) should arguably pass - unless we
are hitting some subtle case of wanting mixed types.
Additional context
The error I get is:
and I suspect we need to pass in a few dtype= parameters in the code.
Would you like to help us fix it?
yes
The text was updated successfully, but these errors were encountered: