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.

Sign up for GitHub

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

Jump to bottom

Per token #39

Closed

horheynm wants to merge 2 commits into main from per-token

Member

horheynm commented Apr 25, 2024

Blocked on channel wise being merged.
Channel gets scale and zp from one axis, per token is along another dim, so can reuse that

horheynm added 2 commits

April 25, 2024 18:37


          draft

7fa62ff


          per token

df1e61d

bfineran reviewed

View reviewed changes

src/compressed_tensors/quantization/observers/base.py

-                          self._scale, self._zero_point = self.calculate_qparams(observed)
+                          if self.quantization_args.strategy == QuantizationStrategy.TOKEN:
+                              self._scale, self._zero_point = self.get_qparams_per_dim(
+                                  observed, dim=2

Contributor

bfineran Apr 29, 2024

dim should be 1

horheynm closed this

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet