Commits FluxML/Flux.jl - SemanticDiff

Skip the redundant GPU index checks in label cross-entropies

CarloLucibello committed 2 days ago

1240f836

NVTX ranges in training steps, a profiling guide, and an auto GC pacer fix (#2831)

CarloLucibello committed 2 days ago

Verified 9873b9fb

Return the loss on the device from Reactant `trainstep!` (#2825)

CarloLucibello committed 3 days ago

Verified 7d4ca947

Update NCCL requirement from 0.1.1 to 0.1.1, 1.0 (#2820)

dependabot[bot] committed 3 days ago

Verified 56cffa42

Fix AlphaDropout gradient on GPU with ChainRules ≥ 1.74.3 (#2821)

CarloLucibello committed 3 days ago

Verified 1ef2e197

Fix ConvTranspose pad=:same output size with both stride and dilation (#2814)

parth-patel010 committed 6 days ago

Verified 77f9cb16

Make `outputsize` and `@autosize` work with `Embedding` (#2805)

parth-patel010 committed 8 days ago

Verified eca62fea

Bump JuliaRegistries/TagBot from 1.25.13 to 1.25.15 (#2808)

dependabot[bot] committed 8 days ago

Verified 14024416

Error clearly when adaptive pooling input is smaller than the output (#2810)

parth-patel010 committed 8 days ago

Verified d12f2ebc

Use the given rng to place the zeros of sparse_init (#2812)

parth-patel010 committed 8 days ago

Verified f11dd1ac

Add Lux + ReactantNitro versions of the GPT-2 and ResNet examples (#2803)

CarloLucibello committed 13 days ago

Verified 3b933245

Refresh ResNet and GPT-2 examples and benchmarks (#2794)

CarloLucibello committed 13 days ago

Verified 422d73f7

Bump version from 0.16.14 to 0.16.15

CarloLucibello committed 13 days ago

Verified 095766de

One GC pacer per model for train! and trainstep!, with a full GC after a step that compiles (#2798)

CarloLucibello committed 13 days ago

Verified a22f35ef

Make `g` of `WeightNorm` with `dims = :` an array, so that it is trained (#2797)

parth-patel010 committed 13 days ago

Verified eee35111

Mooncake: move the rule for mse here from Mooncake's Flux extension (#2772)

AstitvaAggarwal committed 13 days ago

Verified f45b4779

Count a change of exactly `min_dist` as no improvement in `early_stopping` and `plateau` (#2793)

parth-patel010 committed 14 days ago

Verified 20a0a8f5

Add a precompile workload (#2792)

CarloLucibello committed 14 days ago

Verified 3b9a07ea

Fix Enzyme zero gradient for a loss returning a NamedTuple with aux arrays (#2790)

CarloLucibello committed 15 days ago

Verified 5d4c57fa

GPT-2 example: add a CUDA.jl version (#2765)

CarloLucibello committed 15 days ago

Verified 07ac41c7

Fix the gradient of `logitcrossentropy` for a 0-dimensional label (#2788)

parth-patel010 committed 15 days ago

Verified fa4f7c93

Bump version from 0.16.13 to 0.16.14