ColossalAI/colossalai
Hanks cc40fe0e6f
[fix] multi-node backward slowdown (#6134)
* remove redundant memcpy during backward

* get back record_stream
2024-11-14 17:45:49 +08:00
..
_analyzer
_C
accelerator [misc] fit torch api upgradation and remove legecy import (#6093) 2024-10-18 16:48:52 +08:00
amp [plugin] support get_grad_norm (#6115) 2024-11-05 18:12:47 +08:00
auto_parallel
autochunk
booster [zero] support extra dp (#6123) 2024-11-12 11:20:46 +08:00
checkpoint_io [checkpointio] fix hybrid plugin model save (#6106) 2024-10-31 17:04:53 +08:00
cli
cluster [FP8] rebase main (#5963) 2024-08-06 16:29:37 +08:00
context
device
fx
inference [shardformer] fix linear 1d row and support uneven splits for fused qkv linear (#6084) 2024-10-10 14:34:45 +08:00
interface [plugin] support get_grad_norm (#6115) 2024-11-05 18:12:47 +08:00
kernel [misc] fit torch api upgradation and remove legecy import (#6093) 2024-10-18 16:48:52 +08:00
lazy [fp8] Merge feature/fp8_comm to main branch of Colossalai (#6016) 2024-08-22 09:21:34 +08:00
legacy [fp8] Merge feature/fp8_comm to main branch of Colossalai (#6016) 2024-08-22 09:21:34 +08:00
logging [fp8] Merge feature/fp8_comm to main branch of Colossalai (#6016) 2024-08-22 09:21:34 +08:00
moe [hotfix] moe hybrid parallelism benchmark & follow-up fix (#6048) 2024-09-10 17:30:53 +08:00
nn
pipeline [misc] fit torch api upgradation and remove legecy import (#6093) 2024-10-18 16:48:52 +08:00
quantization [fp8] add fallback and make compile option configurable (#6092) 2024-10-18 13:55:31 +08:00
shardformer [hotfix] fix flash attn window_size err (#6132) 2024-11-14 17:11:35 +08:00
tensor [fp8] support fp8 amp for hybrid parallel plugin (#5975) 2024-08-07 18:21:08 +08:00
testing [fp8] Merge feature/fp8_comm to main branch of Colossalai (#6016) 2024-08-22 09:21:34 +08:00
utils [checkpointio] fix hybrid plugin model save (#6106) 2024-10-31 17:04:53 +08:00
zero [fix] multi-node backward slowdown (#6134) 2024-11-14 17:45:49 +08:00
__init__.py
initialize.py [fp8] hotfix backward hook (#6053) 2024-09-11 16:11:25 +08:00