-
Notifications
You must be signed in to change notification settings - Fork 308
Pull requests: NVIDIA/TransformerEngine
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[TE/JAX] Enabling CudaGraph for custom calls with FFI
jax
#1228
opened Oct 7, 2024 by
phu0ngng
Loading…
4 of 13 tasks
Check for bf16 support and context parallel support for fused attention
#1227
opened Oct 7, 2024 by
mgoldfarb-nvidia
Loading…
8 of 13 tasks
[PyTorch] Drop FA2 as an installation requirement
#1226
opened Oct 7, 2024 by
cyanguwa
Loading…
8 of 13 tasks
[PyTorch] Add documentation for FP8 attention checkpointing
#1223
opened Oct 2, 2024 by
cyanguwa
Loading…
8 of 13 tasks
Fix bug in torch compile and seqdim is integer
#1217
opened Sep 29, 2024 by
wplf
Loading…
7 of 13 tasks
[PyTorch] Improve
get_attention_backend
#1214
opened Sep 27, 2024 by
cyanguwa
Loading…
8 of 13 tasks
[JAX] Expose sliding window attn to TE-JAX API
enhancement
New feature or request
jax
#1205
opened Sep 25, 2024 by
huanghua1994
Loading…
8 of 13 tasks
[PyTorch] Debug dtype casting in operation-based API
bug
Something isn't working
#1202
opened Sep 24, 2024 by
timmoon10
Loading…
7 of 13 tasks
Draft: Use fused push_send_recv kernel for TP AG and RS overlaps
#1200
opened Sep 24, 2024 by
erhoo82
Loading…
13 tasks
[PyTorch] Fused dbias-cast-transpose in bias operation
#1168
opened Sep 6, 2024 by
timmoon10
Loading…
7 of 13 tasks
[PyTorch/C] Exposed Userbuffers configuration option to control comm and compute stream priorities
enhancement
New feature or request
#1149
opened Aug 29, 2024 by
denera
Loading…
8 of 13 tasks
[PyTorch] Avoid saving fp8_tensors in certain scenarios
#1143
opened Aug 28, 2024 by
cyanguwa
Loading…
8 of 13 tasks
[PyTorch] Userbuffers support in operation-based API
#1142
opened Aug 27, 2024 by
timmoon10
Loading…
7 of 13 tasks
Don't save fp8 q/k/v/out tensors when using bf16 bprop
#1139
opened Aug 27, 2024 by
guyueh1
Loading…
13 tasks
Fix param input order for cudagraph
bug
Something isn't working
#1138
opened Aug 27, 2024 by
yifeis-nv
Loading…
4 of 13 tasks
Add high_precision_init_val to model params when using fp8_model_init
#1121
opened Aug 19, 2024 by
kunlunl
Loading…
8 of 13 tasks
Previous Next
ProTip!
no:milestone will show everything without a milestone.