Files
tinygrad/test
chenyu c07daf40e7 move attention upcast (#7830)
still upcast before softmax, but faster because intermediate buffer can be stored in half (as long as qk is within half range).
2024-11-22 17:10:51 -05:00
..
2024-11-22 16:39:39 -05:00
2024-11-22 17:10:51 -05:00
2020-12-15 23:44:08 -08:00
2024-11-11 20:18:04 +08:00
2024-11-22 15:55:02 -05:00
2024-11-03 11:26:10 +08:00
2024-07-12 20:43:36 -07:00
2024-10-29 14:01:44 +08:00
2024-10-06 14:46:58 -04:00
2024-11-21 12:44:48 +08:00
2023-12-07 17:07:05 -08:00
2024-05-15 23:46:08 +03:00
2024-11-22 15:55:02 -05:00
2024-10-12 18:20:44 +08:00
2024-10-12 18:20:44 +08:00
2024-10-12 18:20:44 +08:00
2024-11-13 12:28:41 -05:00
2024-11-05 11:35:15 -05:00