Logan Adams
|
13d19827b9
Merge branch 'master' into loadams/update-nv-inference-torch-ver
|
8 月之前 |
BacharL
|
1c9e5ef2d2
Check overflow on device without host synchronization for each tensor (#5115)
|
8 月之前 |
Nadav Elyahu
|
fac1df9983
tensorboard logging: avoid item() outside gas to improve performance (#5135)
|
8 月之前 |
Bruno Magalhaes
|
2d0a6bc20a
remove mandatory `index` key from output of `metric_function` in `DataAnalysis` map operation (#5112)
|
8 月之前 |
Bruno Magalhaes
|
a7864846a4
Avoid zero-sized microbatches for incomplete minibatches when doing curriculum learning (#5118)
|
8 月之前 |
Bruno Magalhaes
|
2b411103ad
Fix broadcast deadlock for incomplete batches in data sample for data analysis (#5117)
|
8 月之前 |
Bruno Magalhaes
|
3e5d400473
Solve tensor vs numpy dtype conflicts in data efficiency map-reduce. (#5108)
|
8 月之前 |
Logan Adams
|
e6c2cb03f7
Merge branch 'master' into loadams/update-nv-inference-torch-ver
|
8 月之前 |
Michael Wyatt
|
b20c46745b
add missing methods to MPS_Accelerator (#5134)
|
8 月之前 |
Logan Adams
|
bf600d2123
Comment out tests that fail to test others
|
8 月之前 |
Logan Adams
|
39c7fb6764
Update torch version
|
8 月之前 |
Logan Adams
|
ad726e2992
Fix whitespace issue
|
8 月之前 |
Logan Adams
|
2fd388844b
Merge branch 'master' into loadams/update-nv-inference-torch-ver
|
8 月之前 |
Logan Adams
|
083197ea1b
Update return codes in PyTest to properly error out if tests fail (#5122)
|
8 月之前 |
Logan Adams
|
ff5c624990
Try with newer docker image
|
8 月之前 |
Bruno Magalhaes
|
9c69662032
Fix issue with zero-sized file after merging file on curriculum `map_reduce` (#5106)
|
8 月之前 |
Logan Adams
|
d7d275d45d
Update nv-inference torch version
|
8 月之前 |
Michael Wyatt
|
d532f64365
Update inference test model names (#5127)
|
8 月之前 |
Michael Wyatt
|
d67d4e5b2f
disable compile tests for torch<2.1 (#5121)
|
8 月之前 |
Bruno Magalhaes
|
9aaead5520
removed if condition in `if collate_fn is None` (#5107)
|
8 月之前 |
Max Kovalenko
|
3c0bd31288
BF16 optimizer: Improve device utilization by immediate grad update (#4975)
|
8 月之前 |
Mingzhen WANG
|
0d9cfa01e2
[NPU]ZeRO-Infinity feature compatibility (#5077)
|
8 月之前 |
Masahiro Tanaka
|
f295aea09e
Stop tracking backward chain of broadcast (ZeRO3) (#5113)
|
8 月之前 |
Michael Wyatt
|
41bc9feed0
Update version.txt after 0.13.2 release (#5119)
|
8 月之前 |
Masahiro Tanaka
|
18179807f5
Remove optimizer step on initialization (#5104)
|
8 月之前 |
Lev Kurilenko
|
25a02047ae
Add HIP conversion file outputs to .gitignore (#5111)
|
8 月之前 |
Lev Kurilenko
|
d04a8386d1
Skip Triton import for AMD (#5110)
|
8 月之前 |
Masahiro Tanaka
|
0a10bd427e
Fix alignment of optimizer states when loading (#5105)
|
8 月之前 |
André Storhaug
|
2518cc429d
Support `exclude_frozen_parameters` for `zero_to_fp32.py` script (#4979)
|
8 月之前 |
Heyang Qin
|
3c811c966b
47% FastGen speedup for low workload - refactor allocator (#5090)
|
8 月之前 |