github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-04-07 03:00:26 -04:00
Code Issues Packages Projects Releases Wiki Activity
1,847 Commits 241 Branches 14 Tags
810f03dafaaa19083c3fdcc805b3bc10b58e57ef
Commit Graph

52 Commits

Author SHA1 Message Date
George Hotz
810f03dafa conv3d + unet3d (#772) 
* conv3d, needs test

* test passes, padding wrong on unet

* unet3d

* no conv3d on images
 2023-05-12 13:54:07 -07:00
George Hotz
46d419060b start on mlperf models 2023-05-10 16:30:49 -07:00
Jacky Lee
d13629cb26 ResNet: match implementation with Nvidia and PyTorch (#770) 
* Match ResNet implementation with pytorch and nvidia

* Reduce number of Epochs
 2023-05-10 09:01:22 -07:00
Jacky Lee
06ed958abd Fix train_resnet example (#744) 
* Fix ResNet example

* Scientific notation
 2023-04-12 13:48:39 +05:30
George Hotz
d6f4219952 LayerNorm2d for 2 lines 2023-03-20 16:58:43 -07:00
George Hotz
b1206bcb18 third try at torch loading (#677) 
* third try at torch loading

* numpy fixed

* fix enet compile

* load_single_weight supports empty weights

* oops, CPU wasn't the default

* so many bugs
 2023-03-10 19:11:29 -08:00
George Hotz
1a039306d2 good changes from llama branch (#671) 
* good changes from llama

* transpose behavior changed
 2023-03-09 20:51:22 -08:00
George Hotz
b14d31d6db ConvNeXt + extras (#657) 
* simple convnext implementation

* shorter function names

* need to realize the random functions now

* creating an optimizer realizes all params

* assign contiguous

* fix lazy lazy

* why was i doing that...add convnext to tests

* LazyNumpyArray

* enable assert + comment

* no two tiny
 2023-03-06 22:10:56 -08:00
George Hotz
2e56a4793e rename log_softmax, support dim, fix onnx Softmax 2023-02-24 10:11:24 -08:00
Jacky Lee
cb679cd051 Fix weight initialization (#566) 
* Fix weight initialization

* Use scaled_uniform in serious_mnist
 2023-02-19 11:25:29 -08:00
Kirill
7944cfdadc Remove Tensor.data (#565) 2023-02-18 16:36:12 -08:00
Jacky Lee
e172f0087a BatchNorm2D -> BatchNorm2d (#558) 
* BatchNorm2D -> BatchNorm2d

* Fix typo
 2023-02-16 12:31:49 -08:00
Lucas Keller
56a06280c5 Testing/utils (#548) 
* New unittest for utils.py

Unit test fetch in basic ways. Would have tested more fetches, but
downloading stuff for tests is annoying and mocking is more
dependencies.

* Remove unused imports
 2023-02-10 12:08:20 -06:00
George Hotz
a0d169eb59 fix efficientnet 2022-09-28 14:23:01 -07:00
Comma Device
a734df98fa TEST_ENET for openpilot compiler 2022-08-31 13:23:36 -04:00
George Hotz
368c0ce2f6 NUM=-2 for ants 2022-07-02 15:47:10 -07:00
George Hotz
0cb99d72e9 NUM=-1 is a small efficientnet for small people 2022-07-02 15:11:51 -07:00
George Hotz
8cf1aed0f4 don't track_running_stats, parameters must require_grad 2022-07-02 14:38:45 -07:00
George Hotz
67ff6b52fd move padding to convs in enet 2022-06-26 23:14:31 -07:00
George Hotz
892ac661e1 enet readability 2022-06-07 10:23:05 -07:00
George Hotz
0ee21ba115 add ViT test and car 2022-06-05 17:12:43 -07:00
George Hotz
c8b569a8c7 cleaner comments 2022-05-14 21:28:39 -07:00
cjg91
7025c9bbeb Transfer learning for ResNet (#295) 
* Transfer learning for ResNet

* moved ResNet depth specifics into the class
 2022-01-15 23:22:10 -05:00
George Hotz
55d792b065 Revert "fixup resnet" 
This reverts commit 4eabe677ed.
 2022-01-15 20:22:01 -08:00
George Hotz
4eabe677ed fixup resnet 2022-01-15 20:21:02 -08:00
George Hotz
c0c2c0b041 support larger ViT models 2021-12-12 10:45:10 -08:00
George Hotz
e28cdfb0cf clean up resnet 2021-11-30 16:14:54 -05:00
George Hotz
8f5779eeaa very minor change 2021-11-30 15:54:03 -05:00
George Hotz
d31ef0ae48 make vit names match pytorch 2021-11-30 11:34:14 -05:00
George Hotz
4b7c31b5b7 break vit into it's own file 2021-11-30 11:19:22 -05:00
George Hotz
46bbbcf7f0 model touchups 2021-11-30 11:13:34 -05:00
George Hotz
835869974c clean up vit code 2021-11-30 10:58:03 -05:00
George Hotz
c39824bc62 oops, forgot some stars 2021-11-30 00:46:14 -05:00
George Hotz
908db3bdea support bias in conv like linear 2021-11-30 00:44:59 -05:00
George Hotz
bd21304e3c linear takes in weight and bias 2021-11-30 00:38:47 -05:00
George Hotz
535f02cc64 use sequential 2021-11-30 00:25:39 -05:00
George Hotz
de938c2d9d vit is now tested 2021-11-30 00:23:06 -05:00
George Hotz
aff810e722 unify transformer block 2021-11-29 18:58:15 -05:00
George Hotz
58ed46963e fix broadcastdot 2021-11-29 18:54:57 -05:00
George Hotz
033b04494a resnet pretrained is broken 2021-11-29 18:13:52 -05:00
George Hotz
dca076dbf1 remove dumb nn ops 2021-11-29 18:05:31 -05:00
George Hotz
8097b8f7d6 vit works 2021-11-29 16:28:14 -05:00
George Hotz
f909ab194f gelu with broken test 2021-11-29 15:00:50 -05:00
George Hotz
1eafa5580e layernorm with learnable parameters 2021-11-29 13:03:57 -05:00
George Hotz
c7f795ca1e added dot affine 2021-11-29 12:55:56 -05:00
George Hotz
30eb3afbe1 add bias term to transformer 2021-11-29 12:45:27 -05:00
George Hotz
99b6051467 add ff_dim to transformer 2021-11-29 12:40:52 -05:00
George Hotz
641b1dbb40 remove ane, start supporting ops_torch 2021-10-30 17:47:00 -07:00
George Hotz
7d12482d80 refactor efficientnet loading 2021-10-30 17:02:17 -07:00
Sebastian Kreft
3358770182 chore(efficientnet): don't use eval when loading weights (#286) 
Because the weights are being loaded from a third party internet address, it's unsafe to use eval. Also with the change I think the code became a little bit more clear as now it's clearer which keys are being transformed.

Co-authored-by: Seba Kreft <sebastian.kreft@houm.com>
 2021-10-22 15:10:04 -07:00