- 11 Mar, 2025 33 commits
-
-
jmorganca authored
-
jmorganca authored
-
jmorganca authored
-
jmorganca authored
-
Michael Yang authored
-
Daniel Hiltgen authored
-
jmorganca authored
-
jmorganca authored
-
jmorganca authored
This reverts commit c7eae586b899083acebcd9b3847b89ea78c2850c.
-
Jesse Gross authored
This is useful for a few things: - Work around bugs, such as having 2 images in one batch - Keep the image in a single batch for fully connected attention - Improve performance by not evaluating embeddings multiple times
-
Jesse Gross authored
Currently we are using positions, which are relative to a sequence and may not be unique.
-
Jesse Gross authored
-
Jesse Gross authored
-
Michael Yang authored
-
Michael Yang authored
-
Patrick Devine authored
-
Michael Yang authored
-
Michael Yang authored
-
Michael Yang authored
-
Michael Yang authored
-
Michael Yang authored
-
Michael Yang authored
-
Patrick Devine authored
-
Michael Yang authored
-
Michael Yang authored
-
Jesse Gross authored
-
Patrick Devine authored
-
Michael Yang authored
-
Patrick Devine authored
-
Jesse Gross authored
-
Michael Yang authored
-
Patrick Devine authored
-
Daniel Hiltgen authored
-
- 10 Mar, 2025 7 commits
-
-
Michael Yang authored
fix: pad tensor item if ge zero
-
Michael Yang authored
this produces a nicer output since both positive and negative values produces the same width
-
Vincent Koc authored
-
Parth Sareen authored
-
Michael Yang authored
Better WantedBy declaration
-
frob authored
-
Xiaowei Zhu authored
-