-
Notifications
You must be signed in to change notification settings - Fork 463
Insights: pytorch/FBGEMM
Overview
-
- 0 Merged pull requests
- 21 Open pull requests
- 5 Closed issues
- 4 New issues
Could not load contribution data
Please try again later
21 Pull requests opened by 8 people
-
Allow manual specification of kernels in fp8 rowwise
#2951 opened
Aug 8, 2024 -
Add configuration knob for ENSEMBLE_ROWWISE_ADAGRAD, frontend
#2955 opened
Aug 8, 2024 -
Add int4 to int4 CPU Sequential TBE API
#2956 opened
Aug 8, 2024 -
Switch hipcub::DeviceRadixSort::SortPairs call to rocprim::device_radix_sort_pairs
#2960 opened
Aug 9, 2024 -
Enable pipeline prefetching
#2963 opened
Aug 9, 2024 -
Add a few missing files to cmake
#2967 opened
Aug 11, 2024 -
Add kv cache related ops (#65)
#2968 opened
Aug 11, 2024 -
Change Torch_CHECK condition in reorder_batched_ad_lengths_gpu
#2973 opened
Aug 12, 2024 -
ensemble rowwise adagrad (fbgemm frontend)
#2981 opened
Aug 13, 2024 -
Add FP8 x INT4 Gemm to Quantize Benchmarks
#2984 opened
Aug 13, 2024 -
Minor CK FP8 Tuning Improvements
#2987 opened
Aug 14, 2024 -
Improve Inference UX, add tests, and add inference API docstrings (#2295)
#2988 opened
Aug 14, 2024 -
Add masked_index_benchmark
#2989 opened
Aug 14, 2024 -
Fix memory utils organization
#2990 opened
Aug 14, 2024 -
Reduce prefetch SM usage when using pipeline prefetching
#2991 opened
Aug 14, 2024 -
Fix get_unique_indices_v2 registration
#2993 opened
Aug 14, 2024 -
Enable int4 to int4 CPU STBE in fbgemm_gpu TBE API
#2994 opened
Aug 15, 2024 -
Add a CPU nbit to float dequantization op that supports torch.quintMxN type and QuantizedCPU backend
#2995 opened
Aug 15, 2024 -
Add int4 to int4 CPU Sequence TBE kernel
#2996 opened
Aug 15, 2024 -
Add unit test for int4 to int4 sequence CPU TBE
#2997 opened
Aug 15, 2024 -
Test moving doc into TBE file
#2998 opened
Aug 15, 2024
5 Issues closed by 3 people
-
[Question] Is there FP8 embedding support for training?
#2920 closed
Aug 14, 2024 -
Why is there no implementation of adamw optimizer. Is there a plan for development?
#2969 closed
Aug 14, 2024 -
fail to compile version v0.8.0 on cuda 12.4
#2950 closed
Aug 11, 2024 -
fbgemm_gpu_py.so: undefined symbol: _ZNK3c105Error4whatEv
#2938 closed
Aug 9, 2024 -
macos install fbgemm error.
#2945 closed
Aug 9, 2024
4 Issues opened by 4 people
-
float conversion emulation routines
#2985 opened
Aug 14, 2024 -
fbgemm_gpu test fail
#2977 opened
Aug 13, 2024 -
cmake is desynchronized from internal TARGETS files, many files not being built
#2966 opened
Aug 11, 2024 -
DLRM run failed .
#2961 opened
Aug 9, 2024
2 Unresolved conversations
Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.
-
[Question FBGEMM_GPU] Adam optmizer not optimized
#2824 commented on
Aug 13, 2024 • 0 new comments -
Remove redundant torch.abs in sim check
#2822 commented on
Aug 10, 2024 • 0 new comments