mirrors/tinygrad
2
0
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-06-24 02:14:17 +00:00
Code Issues Projects Releases Packages Wiki Activity
1,995 commits 280 branches 15 tags 325 MiB
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