51616 b55b9a5643 workaround to work with deepspeed (avoid dtype cast of hypernet)		2025-01-08 19:33:08 +00:00
..
configs.py	add repeat prompt	2025-01-08 16:17:07 +00:00
data_utils.py	fix repeat logic	2025-01-08 17:14:19 +00:00
hooks.py	better logging	2024-12-23 04:03:19 +00:00
intx_sft.py	workaround to work with deepspeed (avoid dtype cast of hypernet)	2025-01-08 19:33:08 +00:00
model_loading.py	workaround to work with deepspeed (avoid dtype cast of hypernet)	2025-01-08 19:33:08 +00:00
modeling_utils.py	workaround to work with deepspeed (avoid dtype cast of hypernet)	2025-01-08 19:33:08 +00:00
pooling.py	add hyperlora w/ mean pooler + biggest head	2024-12-21 09:56:19 +00:00
training_utils.py	fix unseen split + wandb tags	2025-01-07 10:46:03 +00:00
utils.py	biashyperinit + better config + bigger ctx_num_10 + perceiver args	2025-01-04 15:42:08 +00:00