github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-04-29 03:00:14 -04:00
Code Issues Packages Projects Releases Wiki Activity
7,808 Commits 248 Branches 14 Tags
0568720a68d301a47ba7b2fd66cb957525e4d56f
Commit Graph

7808 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
George Hotz
0568720a68 delete revectorize (#9000) 
* delete revectorize

* test vectorized LLVM/CLANG

* idk about that

* was that the segfault?
 2025-02-10 18:32:35 +08:00
qazal
fd9f9ec772 realized base tensors become RESHAPE(BUFFER) [pr] (#8994) 2025-02-10 10:17:54 +01:00
George Hotz
910ae260cd dsp float4 fold + revectorize [pr] (#8995) 
* dsp float4 fold [pr]

* revectorize

* fix reg issue

* no bool vectorize

* cleanups

* no need for that
 2025-02-10 12:14:32 +08:00
George Hotz
e618efce22 COMMUTATIVE flipping is only for ints (#8996) 
* COMMUTATIVE flipping is only for ints [pr]

* no pr

* comm fixes this
 2025-02-10 12:01:28 +08:00
George Hotz
2983285315 use HEX_REG_QEMU_INSN_CNT from qemu as a DSP timer [pr] (#8993) 
* use HEX_REG_QEMU_INSN_CNT from qemu as a DSP timer [pr]

* add quantize test to dsp

* fix tests

* older onnx

* debug, let's see what's happening
 2025-02-10 11:07:35 +08:00
chenyu
9119716761 update Tensor.maximum (#8992) 
now it's just broadcast and UOp.maximum
 2025-02-09 21:26:27 -05:00
nimlgen
88add71c25 amd: increase sdma copy size (#8989) 
* amd: increase sdma max copy size

* rm this

* fix

* fx

* ops
 2025-02-09 20:53:35 +03:00
qazal
7eba5fb413 Tensor.empty is RESHAPE(BUFFER) (#8987) 
* empty is RESHAPE(BUFFER)

* eh

* add test_empty_buf

* can we unsupport this

* linter

* Revert "can we unsupport this"

This reverts commit 0f71e1aadb.
 2025-02-09 18:42:51 +01:00
qazal
44479f8ad6 raise ValueError in view reshape for negative dims [pr] (#8988) 2025-02-09 17:27:15 +01:00
nimlgen
c6c2373bc0 replace libpciaccess autogen with just pci regs (#8983) 
* replace libpciaccess autogen with just pci regs

* add pci.py
 2025-02-09 18:40:45 +03:00
qazal
55351ebb31 minimal failing test for #8975 [pr] (#8982) 2025-02-09 14:10:37 +01:00
nimlgen
e5a3f60fc2 am: remove libpciaccess dep (#8980) 
* am: remove libpciaccess dep

* offset in mockhwiface

* op

* fake regions
 2025-02-09 16:06:55 +03:00
nimlgen
52a69dd5e9 Revert "use am in training benchmarks (#8965)" (#8981) 
This reverts commit 107e616857.
 2025-02-09 15:43:45 +03:00
George Hotz
0b26cee2f1 fix some slow tests [pr] (#8979) 2025-02-09 15:57:04 +08:00
George Hotz
208097d488 try reducing testing deps [pr] (#8976) 
* reduce testing deps

* break out test models

* add PR to models, add models to metal

* okay, not that

* mac cleanup

* mac typo

* other typo
 2025-02-09 15:22:32 +08:00
George Hotz
6ffee2fca9 reduce speed example [pr] (#8978) 
* reduce speed example

* fast like a nascar
 2025-02-09 14:13:59 +08:00
Samuel Ayala
ac3765c043 use getpass instead of os.getlogin() (#8972) 2025-02-08 23:29:26 +03:00
qazal
308516e439 fix viz paginate + cleanups [pr] (#8973) 
* fix viz paginate [pr]

* cleanups

* remove the extra font definition

* more work

* none for the first graph
 2025-02-08 20:26:57 +01:00
nimlgen
107e616857 use am in training benchmarks (#8965) 
* am in training benchmarks

* fix

* not needed anymore
 2025-02-08 20:20:47 +03:00
nimlgen
79de980565 am: do not fork pci bars (#8969) 2025-02-08 19:03:17 +03:00
chenyu
0cac941af1 move xpow to sym instead of late_rewrite (#8968) 
does not need to be in late_rewrite and can be simplified further
 2025-02-08 10:09:24 -05:00
qazal
e7182bbb2c fix "fatal bad object" log in process replay [pr] (#8966) 2025-02-08 11:57:38 +01:00
uuuvn
9b9c1e14da Late MTLCompiler load (#8963) 
Moved loading MTLCompiler (and trying to load normal llvm before it)
to MetalCompiler, like in CPUProgram with helper
 2025-02-08 17:29:23 +08:00
George Hotz
a3c78d47b3 speed docs + upgrades [pr] (#8964) 
* add some docs about speed [pr]

* better torch gemm

* enable locals on llvm/clang

* disable locals for beam speed on LLVM/CLANG

* 0x20 alignment in llvm allows ymm use
 2025-02-08 17:28:52 +08:00
George Hotz
5bdd6a1cc4 increase CI speed with more runners [pr] (#8961) 
* increase CI speed with more runners [pr]

* splits + cleanups [pr]

* more runners

* need that dep

* split that too

* can't be minimal

* move test readme

* bugfix + naming

* one more split

* bump to 22.04
 2025-02-08 09:04:36 +08:00
nimlgen
11d50324d8 am: tiny cleanups (#8958) 
* am: start cleanups

* am
 2025-02-07 23:44:43 +03:00
chenyu
cfd28517df move pow folding tests to test_schedule [pr] (#8955) 
not really belongs to test_const_folding
 2025-02-07 12:51:43 -05:00
George Hotz
c2b4c43edb handle stride 0 reduce (#8068) 
* handle stride 0 reduce [pr]

* more test fixups

* a few more

---------

Co-authored-by: qazal <qazal.software@gmail.com>
 2025-02-07 15:40:58 +01:00
qazal
cf21e27d78 little better VIEW simplifier pattern [pr] (#8954) 2025-02-07 12:55:54 +01:00
qazal
329013f577 fix UOp.metadata on KERNEL op [pr] (#8953) 
* fix UOp.metadata on KERNEL op [pr]

* hotfix: is not None
 2025-02-07 12:40:11 +01:00
George Hotz
4de084a835 cleanup ci, split docs/autogen, testing_minimal, LLVM Speed [pr] (#8952) 
* cleanup ci [pr]

* testing_minimal

* add hypothesis to minimal

* fail tiktoken import okay

* add LLVM speed test

* llvm speed w/o beam
 2025-02-07 19:01:59 +08:00
uuuvn
6090cbe3be Try to open llvm first when opening metal (#8949) 
* Try to open llvm first when opening metal

* Use more specific FileNotFoundError
 2025-02-07 18:58:37 +08:00
uuuvn
67b70e4f6c Fix incorrect __del__ (#8950) 
CPython doesn't make any guarantees about order in which globals like
`msg` or `libobjc` are destroyed when the interpreter shuts down

https://github.com/tinygrad/tinygrad/pull/8949 triggered the
unlucky ordering which lead to a bunch of errors at exit

There is also a bunch of other places where similar problems exist
 2025-02-07 18:21:44 +08:00
George Hotz
9ed2d0dfa2 refactor into subactions (#8946) 
* refactor into subactions

* this work?

* add shell

* move install opencl

* valid?

* support mac os x

* refactor other osx

* fix linux/osx

* fixes

* cleanups

* used everywhere

* no quotes

* quotes on true

* bugfixes

* this run?

* hardcode

* that

* process replay action

* fix checkout

* restore to branch

* fix caching

* fix osx python cache

* does replace function exist

* Revert "does replace function exist"

This reverts commit 622177c5a0.

* Revert "fix osx python cache"

This reverts commit e70d55cd93.

* user on osx to fix untar issue

* that
 2025-02-07 18:06:44 +08:00
Ahmed Harmouche
133cacadde Autogen webgpu dawn, removing wgpu-py dependency (f16 support part 1) (#8646) 
* Switch to dawn, all tests passing locally

* Use dawn-python

* Skip failing test

* Skip midcast and fix timestamp on metal ci

* Autogen webgpu

* Try fetch dawn lib again

* /usr/lib

* Without lib prefix

* Test autogen diff

* Delete webgpu support, move everything to ops_webgpu

* mypy fix

* Simplify, refactor

* Line savings

* No ResultContainer

* Type annotation for result

* Some more simplifications

* Why was this explicit sync used at all?

* Refactor: delete functions that are only used once

* Create shader module inline

* Clear unit tests cache, maybe that solves it

* That wasn't it

* Try deleting cache to pass failing weight compare

* weights_only=False for pytorch 2.6

* Simplify ctype array creation

* Remove nanosecond precision timestamps

* Simplify error handling

* Refactor, add back type annotations

* Deleted custom submit function, refactor

* read_buffer simplify

* Fix use after free, refactor

* Simplify supported_features

* Runtime docs

---------

Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>
 2025-02-07 15:16:59 +08:00
George Hotz
dbda72f91d hotfix: raise line limit to 11200 for new webgpu backend 2025-02-07 14:29:20 +08:00
George Hotz
b1e1319972 ci speed on the enterprise plan [pr] (#8942) 2025-02-07 11:18:12 +08:00
Bhavya Gada
3b67712892 [bounty] Fix LLVM=1 NO_DEVECTORIZE=1 python3 test/test_ops.py TestOps.test_strided_conv2d_simple (#8937) 
* fix LLVM=1 NO_DEVECTORIZE=1 python3 test/test_ops.py TestOps.test_strided_conv2d_simple

* remove expectedFailure

---------

Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>
 2025-02-07 10:07:54 +08:00
George Hotz
f54242849d failing test for the devectorize [pr] (#8940) 
* failing test for the devectorize [pr]

* add DEVECTORIZE to method_cache
 2025-02-07 09:44:54 +08:00
nimlgen
ee1a0fb8ec am_smi: print device name (#8939) 2025-02-07 03:01:25 +03:00
chenyu
a092b6395d Tuple -> tuple, List -> list [pr] (#8936) 2025-02-06 14:21:19 -05:00
chenyu
d5183e1584 remove unneeded annotation import (#8934) 2025-02-06 13:12:35 -05:00
chenyu
00d72a5144 setitem isinstance cleanup [pr] (#8932) 2025-02-06 11:44:57 -05:00
qazal
81e241150a hotfix: save 1 line (#8931) 
* hotfix: save 1 line

* no unwrap
 2025-02-06 17:26:05 +02:00
qazal
eb1144be8b hotfix: only check current graph when excluding nodes in viz (#8930) 2025-02-06 16:58:53 +02:00
George Hotz
3cc05081f4 llvm no devectorize, the right way (#8901) 
* closer

* env flag + transcendental issue
 2025-02-06 22:53:49 +08:00
George Hotz
8b16c65bca add compile3 benchmark [pr] (#8929) 2025-02-06 22:49:31 +08:00
qazal
79fb5c6470 hotfix: test_shard_no_recompile shouldn't rely on schedule order [pr] (#8928) 2025-02-06 16:27:59 +02:00
George Hotz
1249e8dd3b objc fast msg, try 2 [pr] (#8927) 2025-02-06 19:06:21 +08:00
nimlgen
86feb98dcd am: add support for 7600 (#8910) 
* am: start to add support for 7600

* test_tiny passes

* mmhub 3 0 2

* cleaner
 2025-02-06 14:04:07 +03:00