SmilingWolf 59cb0be595 SwinV2: add configurable act_layer argument
Defaults to "gelu", but makes it possible to pass "gelu_tanh".
Makes it easier to port weights from JAX/Flax, where the tanh
approximation is the default.
2024-03-05 22:04:17 +01:00
..
2023-10-22 21:36:23 -07:00
2023-10-05 08:58:41 -07:00
2023-05-10 12:16:30 -07:00
2023-12-16 15:10:45 -08:00
2023-11-22 16:32:41 -08:00
2023-05-10 12:16:30 -07:00
2024-02-12 15:00:32 -08:00
2023-05-12 21:48:46 +02:00
2023-10-05 08:58:41 -07:00
2023-10-05 08:58:41 -07:00
2023-10-05 08:58:41 -07:00
2024-02-10 20:14:47 -08:00
2023-10-05 08:58:41 -07:00
2023-09-15 11:41:56 -07:00
2023-11-23 21:48:14 -08:00
2023-10-05 08:58:41 -07:00
2023-10-05 08:58:41 -07:00