github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-22 13:28:06 -05:00
Code Issues Packages Projects Releases Wiki Activity
1,995 Commits 205 Branches 14 Tags
805eef10dd94ed0dc618a0667dff24bfe8ec0099
Commit Graph

300 Commits

Author SHA1 Message Date
Casey Primozic
805eef10dd Add tensorflow GEMM benchmark script (#1000) 
* Modelled closely after the existing torch benchmark script but just adapted slightly for tensorflow
 2023-06-18 10:57:45 -07:00
Diogo
d2b837c1d9 Adds floor/ceil (#989) 
* floor ceil impl

* control casting in numpy
 2023-06-17 10:56:21 -07:00
George Hotz
fe71282ba1 faster RDNA assembly backend (#990) 
* fast asm

* torch gemm
 2023-06-16 12:06:38 -07:00
George Hotz
ba56ee6020 RDNA assembly backend ($1000 bounty) (#787) 
* Revert "Revert "ops rdna""

This reverts commit 0400315078.

* Revert "Revert "writing 2""

This reverts commit 325a3bf2cf.

* no dump

* 2x 2

* simple asm

* local size

* sub

* lil work

* support args != 3

* assembler work

* generate that

* ptx assembler

* begin index renderer

* max

* ptx loops

* gemms work

* valid works

* asm working a bit more

* close

* passing all ops tests

* ptx is a codegen only, not a backend

* ptx

* float16 support

* rdna goes here

* install types

* make amd disassemble

* ansilen for pretty print

* fix ptx log2/exp2

* assemblyinstruction

* new asm

* working gemm

* fix cmp

* more passing

* mod

* ptx works again

* rdan3 add works

* log exp

* sin is sin 2pi

* fix types

* progress

* loops work

* rdna xyz

* better addressing

* cleanups

* handle exception in early process

* div support

* rdna float4

* locals work

* fix neg index

* cast

* smaller diff

* yaml

* import only if selected

* fromimport

* types

* this all needs rewriting

* a few more
 2023-06-16 09:33:18 -07:00
Yahya Lmallas
804c45b5fc FIX: Can't pickle local object (#979) 
_early_exec_process is a local function that is defined whiting the scope of another function, should be global
 2023-06-14 12:32:17 -07:00
Steven Anderson
e54b6c5e7f One hot (#972) 
* passing with 1d indices

* passing all test

* cleanup

* using safe_numpy for scalar
 2023-06-12 10:13:29 -07:00
Diogo
2d4370b487 Adds tril & triu support (#936) 
* triu & tril support

* lint and kernel count error

* switched shape indicies

* larger shape tests

* reverted numpy removal until #942 is resolved
 2023-06-09 22:13:20 -07:00
Steven Anderson
c0e558b77c Test nllloss (#958) 
* works but slow

* work with NC and NCd1 it still slow

* refactor

* support for k dimensions

* without numpy
 2023-06-09 09:00:29 -07:00
Diogo
6b1280f01c fixes to Onnx ops LayerNormalization/Prelu and added OptionalHasElement/OptionalGetElement (#956) 
* prelu and where casting

* typing for safe_numpy

* optional

* get rid of tracing in ci

* cleanup and resolved layernorm issues

* removed debug print
 2023-06-08 16:09:19 -07:00
Diogo
666d151f8a Onnx slice fixups (#952) 
* resolved some slice test errors and added some more debugging logs

* use same device in cumsum

* increased float priority

* onnx debug ouput match input
 2023-06-07 19:44:30 -07:00
M4tthewDE
664d6cc7e5 Implement onnx MeanVarianceNormalization (#943) 2023-06-06 10:28:19 -07:00
Steven Anderson
079ea217a3 fix test_pow_type - autocasting for Pow with inputs of diff type (#937) 2023-06-05 15:22:35 -07:00
M4tthewDE
70f12fdb57 Fix wrong op version being used if versions equal (#934) 2023-06-05 07:45:10 -07:00
Steven Anderson
79613eb83e Test min (#932) 
* fix __neg__ defaulting to float32 due to 0.0

* fixed __neg__ always defaulting to float32

* fixed openpilot (OpenCL) Test
 2023-06-05 00:03:30 -07:00
George Hotz
fbf17f0031 intel benchmark matmul gets 60 TFLOPS? 2023-06-04 17:01:50 +00:00
Steven Anderson
657e642e3a Fixed test suite for Clip (#912) 
* Fixed test suite for Clip

* fixed issue with clip when taking large negative numbers as min

* Remove typings
 2023-06-04 09:01:01 -07:00
George Hotz
afd0be8a9c intel example 2023-06-04 06:43:09 +00:00
George Hotz
ed1963b899 Fast DiskTensor to other Tensor (#916) 
* make disktensors fast

* loading

* loader for sd and llama
 2023-06-03 12:25:41 -07:00
George Hotz
791530045d Refactor LoadOps (#910) 
* test

* work

* upd test

* loadops

* cleanups

* real ones

* remove LazyNumpyArray

* fix assign test

* remove range

* np.require

* llama uses arange kernels

* no caching consts

* fix enet

* torch load support

* tests cleanup

* fix shufflenet

* fix image

* fix torch_load test
 2023-06-03 09:40:43 -07:00
Steven Anderson
513aeb2f66 Fixed all ConstantOfShape test suite (#907) 2023-06-02 11:26:40 -07:00
Steven Anderson
301f7b54c6 ConstantOfShape ONNX test fixed. (#890) 
* ConstantOfShape ONNX test fixed.

* removed redundant if statement

* value is optional and should default to a float32 tensor with value of 0

* fixed: default parameters are created at function definition, bad for mutable objects.
 2023-06-02 07:34:25 -07:00
kposborne2
ae83e9844c add output_padding to transposed conv (#875) 2023-06-01 00:03:22 -07:00
Friedrich Carl Eichenroth
740304ef9d Small Onnx Parser Improvements (#885) 
* wip

* rename onnx_version to onnx_model_versioN

* add type

* add types

* small cleanup

* revert some changes from before

* add todo

* dumb fix
 2023-06-01 00:01:01 -07:00
Marcello Fuschi
3924aae8ed Fix ONNX dropout and unify the implementation (#857) 
* Fix ONNX dropout and unify the implementation

* Use tensor rand method for dropout

* Change approach for RNG in ONNX Dropout

* Fix style

* Test legacy RNG seeding

* Remove the necessity for legacy RNG in Tensor class
 2023-05-31 07:40:47 -07:00
skobsman
2e393f7ef2 InstanceNormalization ONNX test fixed. (#870) 2023-05-30 16:07:44 -07:00
Friedrich Carl Eichenroth
f91f28d9e2 fix a bunch of tests (#856) 2023-05-29 17:48:26 -07:00
zk-tarts
174c65b7d9 add onnx Binarizer op (#850) 
Co-authored-by: zk-tarts <>
 2023-05-29 13:15:50 -07:00
M4tthewDE
4408c25e9a Add Onnx op Shrink (#851) 
* Add onnx Shrink operation

* Fix soft/hard shrink onnx test
 2023-05-29 13:15:39 -07:00
Friedrich Carl Eichenroth
6f2b3755ca set axis default to 0 (#854) 2023-05-29 13:15:28 -07:00
Friedrich Carl Eichenroth
3b158f7a5f fix onnx versions greater or equal 10 (#853) 2023-05-29 13:04:06 -07:00
Diogo
1a5d72f812 Onnx ops And, Or, Xor, Not (#847) 
* onnx and, or, xor, not

* added bool type to llvm and clang

* removed float conversion

* switched where op to use tensor func
 2023-05-29 11:09:20 -07:00
SnakeOnex
844e6d0753 conv1d & conv3d onnx tests (#835) 
* conv1d onnx

* [Work in progress] conv1d + enforcing full padding tuple length

* make ONNX padding reorder not hardcoded, works for 1D and 3D convs now

* conv2d interprets padding based on the input tensor dimensions
 2023-05-29 10:16:45 -07:00
Marcello Fuschi
6d49925a26 Add max_pool2d dilation (#833) 2023-05-28 15:16:48 -07:00
cheeetoo
21d27d31a9 Fix a couple pad tests (#827) 
* fix pad bug

* float type hint for value

* convert pads to list

* update Pad type signature

* Change | to Union since not supported in < python 3.10
 2023-05-28 12:06:46 -07:00
Mattis Megevand
606b841d3f LR Schedulers (#755) 
* lr schedulers + test

* lr scheduler test moved + integration test

* integration test for all lr scheduler

* lr scheduler test now deterministic

* changed optimizer + parameters for lr sched test
 2023-05-27 07:47:49 -07:00
George Hotz
87fa5af70a ptx example 2023-05-26 19:28:51 -07:00
George Hotz
26014a0fa1 add convtranspose (#809) 
* add convtranspose

* onnx convtranspose
 2023-05-26 12:35:03 -07:00
wozeparrot
7351eb4b61 feat: put temperary file in the same directory as the destination file (#805) 2023-05-25 20:46:02 -07:00
Diogo
c19ef0fcce Add sin/cos/tan (#794) 
* added sin/cos/tan

* fix lint

* added onnx ops support
 2023-05-25 09:04:56 -07:00
George Hotz
0400315078 Revert "ops rdna" 
This reverts commit 81a11d891d.
 2023-05-21 13:02:18 -07:00
George Hotz
325a3bf2cf Revert "writing 2" 
This reverts commit dddd6c42f0.
 2023-05-21 13:02:17 -07:00
George Hotz
dddd6c42f0 writing 2 2023-05-21 12:52:36 -07:00
George Hotz
81a11d891d ops rdna 2023-05-21 11:45:38 -07:00
George Hotz
90fff82c8a Rdna (#776) 
* assembler maybe

* custom asm

* rdna3 on quiet

* trigger crashes

* fixed notes

* non-fatal rdna2 crash

* Crash4

* improve rdna sniffer

* comments

* improve sniffer

* asm

* 131 TFLOPS RDNA3

* opt simple matmul

* todos
 2023-05-16 05:33:57 -07:00
George Hotz
89b8b39d9c fix mypy 2023-05-13 21:25:36 -07:00
George Hotz
e0b2035023 fast imagenet eval, gets 76.14% across the set 2023-05-13 21:18:31 -07:00
George Hotz
46d419060b start on mlperf models 2023-05-10 16:30:49 -07:00
George Hotz
cb7c22beeb fix mypy 2023-05-06 19:18:54 +00:00
George Hotz
5190037cbc rocm: disassembler for shader 2023-05-06 19:07:52 +00:00
George Hotz
42256c0d9d rocm sniffer dumps code 2023-05-05 18:36:53 +00:00