220-bug-calling-sequential-lua-functions-in-transform-doesnt-work - Branch - mozilla-ai/encoderfile

fix(transform): add layer_norm to add_methods

#221Merged

Comparing

220-bug-calling-sequential-lua-functions-in-transform-doesnt-work

(

a8f80dc

) with

main

(

a9ce9fa

)

-1%

Untouched: 20

Skipped: 20

Benchmarks

Skipped (20)

Passed

bench_tok_cls_softmax[(128, 256, 32)]

encoderfile/benches/benchmark_transforms.rs

35.7 ms35.6 ms

embedding_postprocess[(16, 128, 768)]

encoderfile/benches/postprocessing.rs

25.2 ms25.2 ms

token_classification_postprocess[(16, 128)]

encoderfile/benches/postprocessing.rs

8.2 ms8.2 ms

token_classification_postprocess[(8, 16)]

encoderfile/benches/postprocessing.rs

3.6 ms3.6 ms

bench_l2_normalize[Axis(1)]

encoderfile/benches/l2_norm.rs

3 ms3 ms

bench_l2_normalize[Axis(0)]

encoderfile/benches/l2_norm.rs

1.6 ms1.6 ms

bench_embedding_l2_normalization[(32, 256, 768)]

encoderfile/benches/benchmark_transforms.rs

267 ms267 ms

embedding_postprocess[(64, 512, 1024)]

encoderfile/benches/postprocessing.rs

312.1 ms312.1 ms

bench_embedding_l2_normalization[(32, 128, 384)]

encoderfile/benches/benchmark_transforms.rs

65.3 ms65.3 ms

token_classification_postprocess[(64, 512)]

encoderfile/benches/postprocessing.rs

61.2 ms61.2 ms

embedding_postprocess[(8, 16, 384)]

encoderfile/benches/postprocessing.rs

4.6 ms4.6 ms

bench_tok_cls_softmax[(32, 128, 8)]

encoderfile/benches/benchmark_transforms.rs

1.8 ms1.8 ms

sequence_classification_postprocess[64]

encoderfile/benches/postprocessing.rs

48.2 µs48.3 µs

sequence_classification_postprocess[8]

encoderfile/benches/postprocessing.rs

-1%

11 µs11 µs

bench_embedding_l2_normalization[(16, 16, 16)]

encoderfile/benches/benchmark_transforms.rs

-1%

337 µs339.1 µs

bench_seq_cls_softmax[(128, 32)]

encoderfile/benches/benchmark_transforms.rs

-1%

292 µs294.5 µs

bench_tok_cls_softmax[(16, 16, 2)]

encoderfile/benches/benchmark_transforms.rs

-1%

183.6 µs186.2 µs

bench_seq_cls_softmax[(32, 8)]

encoderfile/benches/benchmark_transforms.rs

-2%

146.5 µs148.9 µs

bench_seq_cls_softmax[(16, 2)]

encoderfile/benches/benchmark_transforms.rs

-2%

136.6 µs139 µs

sequence_classification_postprocess[16]

encoderfile/benches/postprocessing.rs

-4%

33 µs34.3 µs

Commits

Click on a commit to change the comparison range

Base

main

a9ce9fa

-0.55%

add layer_norm to add_methods

a8f80dc

2 days ago

by besaleli

Home Terms Privacy Docs