microsoft
/
ml4f
зеркало из https://github.com/microsoft/ml4f.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность
143 коммитов 9 Ветки 14 Теги 3,2 MiB
Граф коммитов

143 Коммитов

Автор SHA1 Сообщение Дата
Michal Moskal 77018c8569 Add C "SDK" 2020-11-13 11:30:50 +01:00
Michal Moskal 6b5696dede Add more info to readme 2020-11-13 10:45:19 +01:00
Michal Moskal 13d53cb0d6 Add microSpeech sample 2020-10-20 14:50:15 +02:00
Michal Moskal f35af03449 Fix padding computation when strides are involved 
Throw on unsupported layer
 2020-10-20 14:50:01 +02:00
Michal Moskal 58cfbeb74c More tolerance for f16 errors 2020-10-20 14:49:24 +02:00
Michal Moskal 1bac014f52 v1.5.0 2020-10-16 15:21:19 +02:00
Michal Moskal 04f33a1ccf Save eval info in file 2020-10-16 15:21:10 +02:00
Michal Moskal 48988a31f1 1.4.0 2020-10-16 12:41:38 +02:00
Michal Moskal 27d4b075c5 Change output extension and add --basename option 2020-10-16 12:41:24 +02:00
Michal Moskal 549f8e02c0 v1.3.0 2020-10-15 13:54:52 +02:00
Michal Moskal 1e26221f28 Add --eval option 2020-10-15 13:54:36 +02:00
Michal Moskal 74c222e6cf Simplify build script 2020-10-15 13:20:50 +02:00
Michal Moskal e1a784372d Fix CI reciepe 2020-10-15 13:19:42 +02:00
Michal Moskal 398a6bb4a9 v1.2.0 2020-10-15 12:55:51 +02:00
Michal Moskal 03e58cd9be Fix files 2020-10-15 12:55:38 +02:00
Michal Moskal c6c96180f3 Use packaged module for node 2020-10-15 12:47:07 +02:00
Michal Moskal 81cddcbb9b Fix two errors in asm emit 2020-10-13 14:31:36 +02:00
Michal Moskal eeba84ed28 Store JS output as reference in Thumb 2020-10-13 14:06:31 +02:00
Michal Moskal 83641064fd Add --float16 option 2020-10-13 13:59:02 +02:00
Michal Moskal b45baa020c v1.1.1 2020-10-13 13:48:58 +02:00
Michal Moskal 905d05ae5d Use yarn for build 2020-10-13 13:45:26 +02:00
Michal Moskal ee90148ada 1.1.0 2020-10-13 13:41:50 +02:00
Michal Moskal 94ffb8e590 Add build action 2020-10-13 13:37:15 +02:00
Michal Moskal 9985c7704b Call testFloatConv first 2020-10-13 13:34:35 +02:00
Michal Moskal 2593ac2788 Allow for less precise f16 results 2020-10-13 13:34:17 +02:00
Michal Moskal 2e3019eb98 Test with f16 2020-10-13 13:33:59 +02:00
Michal Moskal c8c3520e55 Add breakpoint support 2020-10-13 13:33:47 +02:00
Michal Moskal 07957a0ffb Bugfix 2020-10-13 13:30:17 +02:00
Michal Moskal 2d3172b76d Add optimize flag 2020-10-13 13:05:52 +02:00
Michal Moskal a894e55dff F16 fixes 2020-10-13 12:34:45 +02:00
Michal Moskal 2e4422c607 Add pretty-printing 2020-10-13 12:12:31 +02:00
Michal Moskal 88edc4b04b Add f16 load impl 2020-10-13 12:00:04 +02:00
Michal Moskal 7338f8027d Add float16->float32 conversion 2020-10-12 18:57:41 +02:00
Michal Moskal 5a9718e8b5 Remove circular dep 2020-10-12 16:44:33 +02:00
Michal Moskal 8f996a39bc Add some more testing 2020-10-12 16:40:33 +02:00
Michal Moskal a7b71d0b08 Work on float16 2020-10-12 16:11:24 +02:00
Michal Moskal 531666470d Move JS runtime to TS 2020-10-12 13:34:15 +02:00
Michal Moskal 711ea1b4fc Add basis for load16 2020-10-12 13:29:33 +02:00
Michal Moskal cb038762b0 Byte buffer for weights 2020-10-12 13:11:00 +02:00
Michal Moskal 4ccf902d5b Refactor weight storing 2020-10-12 12:30:44 +02:00
Michal Moskal abdfe6b0af Try to estimate better allocations 2020-10-09 19:02:03 +02:00
Michal Moskal d810dd82c7 Add f16 vcvts 2020-10-09 19:01:50 +02:00
Michal Moskal 09f3d55fa9 Add .float16 support in asm 2020-10-09 16:57:15 +02:00
Michal Moskal 1b7aa68269 Don't use r1 as scratch 2020-10-09 16:20:08 +02:00
Michal Moskal 38c1b895ac Fix expf() assembly 2020-10-09 15:57:39 +02:00
Michal Moskal f1a91b31c8 Add --sample-model option 2020-10-09 15:33:25 +02:00
Michal Moskal bd53ca72d5 Add -t option 2020-10-09 13:32:52 +02:00
Michal Moskal cb14492709 Find good softmax testing data 2020-10-08 18:48:29 +02:00
Michal Moskal 36e157df3f Add pooling padding and Reshape 2020-10-08 18:22:18 +02:00
Michal Moskal 9304c8b54c Padding fixes 2020-10-08 18:13:01 +02:00