-
Notifications
You must be signed in to change notification settings - Fork 432
Pull requests: pytorch/FBGEMM
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[fbgemm_gpu] Adjust install and test steps based on installed variant
cla signed
#2702
opened Jun 7, 2024 by
q10
Loading…
Revert D57738223: Multisect successfully blamed "D57738223: [fp8 kv cache] wmma_gqa_attn_splitk" for one test failure
cla signed
fb-exported
#2695
opened Jun 6, 2024 by
Aya-ZIbra
Loading…
[comm][ROCm] move memory copy into one_shot_all_reduce
cla signed
module: rocm
#2693
opened Jun 6, 2024 by
wenkaidu
Loading…
Support MTIA in DenseTableBatchedEmbeddingBagsCodegen
cla signed
fb-exported
#2680
opened Jun 4, 2024 by
gnahzg
Loading…
Support stochastic rounding for FP32/BF16 -> FP8 conversion
cla signed
fb-exported
#2677
opened Jun 4, 2024 by
jianyuh
Loading…
Improve Int4 Quantization and Fix Groupwise Scaling
cla signed
fb-exported
#2655
opened May 31, 2024 by
jwfromm
Loading…
[ROCm] enable experimental gen_ai build
cla signed
module: rocm
#2610
opened May 20, 2024 by
jeffdaily
Loading…
all_to_one cuda support non-2d inputs
cla signed
fb-exported
#2575
opened May 9, 2024 by
IvanKobzarev
Loading…
add max norm support to PARTIAL_ROWWISE_ADAM
cla signed
fb-exported
#2567
opened May 7, 2024 by
zainhuda
Loading…
Pyre Configurationless migration for] [batch:9/28]
cla signed
fb-exported
#2557
opened May 3, 2024 by
connernilsen
Loading…
Pyre Configurationless migration for] [batch:6/29]
cla signed
#2548
opened Apr 29, 2024 by
connernilsen
Loading…
Integrate triton row and blockwise fp8 gemm to llm inference.
cla signed
fb-exported
#2547
opened Apr 29, 2024 by
choutim
Loading…
Previous Next
ProTip!
no:milestone will show everything without a milestone.