commaai
/
tinygrad
mirror of https://github.com/commaai/tinygrad.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
3,057 Commits 10 Branches 0 Tags 52 MiB
Commit Graph

3057 Commits

Author SHA1 Message Date
George Hotz 82fd932921
Lower schedule 2 (#2561) 
* ls2

* fix types

* simpler

* cleaner
 2023-12-01 20:25:49 -08:00
George Hotz 217cda81ba hotfix: no metalgraph if there's weird ops 2023-12-01 19:32:55 -08:00
George Hotz 6733425095
lower schedule (#2559) 
* lower schedule

* remove RAND, and don't put load in the JIT yet

* better fix for that test
 2023-12-01 19:17:46 -08:00
Christopher Mauri Milan 077567f62d
Remove as_buffer for TORCH (#2554) 
* remove as_buffer for torch

* enable torch zerocopy if on cpu

* remove as_buffer even on torch:cpu
 2023-12-01 18:51:38 -08:00
chenyu 05a5357dd9
fix handcode_resnet50_opt.py (#2558) 2023-12-01 20:51:21 -05:00
chenyu 86fbd413f3
update test_real_world configs (#2557) 2023-12-01 20:03:52 -05:00
andresgit 00523d5656
New fix accessing elements created by padding (#2529) 
* pad slice test cases, many failing

* fix failing test cases

check mask if we are outside the base buffer
also create a multi-view if in that case we reshape to an empty shape

* real_offset calculation more readable

---------

Co-authored-by: chenyu <chenyu@fastmail.com>
 2023-12-01 19:08:10 -05:00
George Hotz bfdce1f0e7 hotfix: make openpilot test deterministic 2023-12-01 15:37:23 -08:00
George Hotz 9c306be282 better name for fast path 2023-12-01 15:32:47 -08:00
chenyu 67f4e03724
rewrite 0 size loadop into a CONST (#2556) 
* rewrite 0 size loadop into a CONST

* check alloc size

* EMPTY is better

* Revert "EMPTY is better"

This reverts commit 574fe0f9ed28f1b97da5a81afdfd2cd5d9a94ff9.

* no ast is created

* fix test
 2023-12-01 18:29:06 -05:00
George Hotz 4447188051 gate METAL_FAST_LOAD 2023-12-01 15:28:40 -08:00
chenyu e9426f4fe4
simpler get_contraction (#2552) 
* simpler get_contraction

* and test
 2023-12-01 18:02:52 -05:00
George Hotz eb595588bb device.py cleanups for -1 line 2023-12-01 14:59:33 -08:00
George Hotz f9b1de598f hotfix: metal fastpath on sonoma 2023-12-01 14:55:34 -08:00
George Hotz f5de21e753
fast path for copy (#2548) 
* fast copy

* ruff first

* flat_mv on malloc

* order + webgpu test
 2023-12-01 11:34:47 -08:00
wozeparrot 28183c7438
feat: reword (#2549) 2023-12-01 10:56:18 -08:00
George Hotz 4c984bba7e
bump version to 0.8.0, clean CI, remove requests (#2545) 
* bump version to 0.8.0, clean CI, remove requests

* why was that even there
 2023-12-01 10:42:50 -08:00
nimlgen ff47be3a01
ruff check whitespaces (#2547) 2023-12-01 10:42:20 -08:00
George Hotz 8fd8399437
remove flake8 (#2544) 2023-12-01 09:48:41 -08:00
George Hotz d8175a4380
simple fix (#2543) 2023-12-01 09:42:15 -08:00
qazal 04483f8187
refactor llvm consts (#2537) 2023-12-01 09:39:40 -08:00
nimlgen badc97f824
hip & cuda to gpuctypes (#2539) 
* cuda with gpuctypes

* hip gpuctypes

* graphs

* rename + linter happy

* use cpu_time_execution

* no ji in build_kernel_node_params

* remove hip_wrapper

* hip fix

* no arc

* smalle changes

* no clean moduke in cudacpu
 2023-12-01 09:25:27 -08:00
qazal 0fb4ff30c8
share duplicate renders with cstyle (#2538) 2023-12-01 08:10:36 -08:00
chenyu 7fec966b5e
bye bye NOOP (#2534) 
* bye bye NOOP

* SIN

* NEG
 2023-11-30 23:10:35 -08:00
Joe Donovan fa549d198d
Remove `type: ignore` comments (#2533) 
* remove some type ignore comments and fix errors

* remove unnecessary get_args import

* revert triton changes

* remove changes not in tinygrad
 2023-11-30 22:15:55 -08:00
George Hotz 12fa846122
zero copy (#2531) 
* zero copy

* zero copy test

* loads coder in milliseconds

* zero copy for cpu and torch

* src_from_buffer is None

* SLOW_METAL_COPY there
 2023-11-30 18:38:41 -08:00
Matthias Kronberg 5394a05b9d
Fix: Get item from ndarray before casting to int (#2525) 
Directly casting is deprecated and will error in the future.
 2023-11-30 18:34:31 -08:00
George Hotz 2c363b5f0b
new style device (#2530) 
* cpu tests pass

* torch works

* works

* metal works

* fix ops_disk

* metal jit works

* fix openpilot

* llvm and clang work

* fix webgpu

* docs are rly broken

* LRU works on metal

* delete comment

* revert name to ._buf. LRU only on Compiled

* changes

* allocator

* allocator, getting closer

* lru alloc

* LRUAllocator

* all pass

* metal

* cuda

* test examples

* linearizer

* test fixes

* fix custom + clean realize

* fix hip

* skip tests

* fix tests

* fix size=0

* fix MOCKHIP

* fix thneed

* copy better

* simple

* old style metal copy

* fix thneed

* np reshape

* give cuda a device
 2023-11-30 17:07:16 -08:00
chenyu e56511b59a
more type annotation for tensor and lazy (#2528) 
* more type annotation for tensor and lazy

* don't need that
 2023-11-30 17:50:22 -05:00
Davi Silva ddeec24fa8
Cleanup & fix llama.py (#2524) 
* docs, cleanup crap

* comma AI

* fix 70B

* this is why lexical scope exists
 2023-11-30 16:00:17 -05:00
chenyu 7d26452305
call ruff with --preview (#2522) 
some checks are ignored without --preview
 2023-11-30 13:59:00 -05:00
chenyu 5db0cdfbd3
support list of ints (or other Tensorable) in tensor indices (#2520) 
* support list of ints (or other Tensorable) in tensor indices

* enable some index test cases
 2023-11-30 12:46:33 -05:00
chenyu bd941a0df1
first version of test_indexing (#2515) 
* first version of test_indexing

* move to test/imported
 2023-11-30 00:03:59 -05:00
chenyu d210f6a786
minor device.py cleanups (#2510) 2023-11-29 18:16:25 -05:00
qazal 370cfbb957
Cleanup vectorized hip renders (#2497) 
* add typedefs and make_dtypen functions

use ext_vector_type for half16 kernels

* remove the old test_render because we just use whatever cstyle has

* align vectors
 2023-11-29 14:02:12 -08:00
George Hotz abfc99187d
cleanup realize (#2505) 
* delete reallocs

* cleaner

* that's real

* less lines
 2023-11-29 11:38:38 -08:00
George Hotz 3dedeaae74
rebalance tests (#2504) 
* rebalance

* balance

* parallel apt-get for all

* .local/lib/python3.11/site-packages

* what is user doing

* is that path right

* Update test.yml

* okay where are you

* site-packages
 2023-11-29 11:18:22 -08:00
George Hotz 065aff747e
make webgpu test reliable (#2502) 
* remove retry that doesn't work

* fix cleanup

* process exit in cleanup

* add space
 2023-11-29 10:02:24 -08:00
George Hotz 6707f2588e
use copyin (#2500) 
* it's always copyin

* all RawBuffer are RawBufferCopyIn

* cleanups

* this fixes it

* requirements='C'

* more correct
 2023-11-29 09:34:00 -08:00
George Hotz 947711a532
split metal and webgpu tests (#2501) 2023-11-29 09:32:09 -08:00
chenyu 3eb3c74675
metal ci tests everything (#2499) 
* metal ci tests everything

* pretty good

* METAL
 2023-11-29 12:04:37 -05:00
George Hotz 889acefe85
Support weird loads in Image (#2498) 
* image support weird loads

* umm, that was always wrong

* openpilot compile fails with a weird error

* image test passes

* we have valids now

* clean that up

* no more required opts

* add fastvits test, fix bug

* minor cleanups
 2023-11-29 08:30:46 -08:00
George Hotz e333672675
realize cleanup (#2496) 
* move that logic

* revert that change

* clean up transfer and asserts

* what's that junk
 2023-11-28 21:08:39 -08:00
George Hotz 5629fc368c
Use Buffer.STORE at the end of ASTs (#2494) 
* work

* store broken

* interpreteds work

* this passes

* symbolic cpu

* fix tests

* fix opt tests

* images fail

* fix InterpretedFlopCounter

* stupid hack for images
 2023-11-28 20:11:37 -08:00
Liam cf0c9096a9
Removing METAL Skips as CI works (#2488) 
* Test metal CI

* remove metal and CI restrictions

* enable dtype tests for metal ci
 2023-11-28 19:46:59 -08:00
Jake 5588922884
Update cuda_matmul.py (#2495) 2023-11-28 19:46:01 -08:00
George Hotz cdc3b95729 if you don't appreciate a 15 second timeout, you get a 10 second timeout 2023-11-28 17:44:09 -08:00
George Hotz d87a246439
move to new cached fetch (#2493) 
* move to new cached fetch

* extra.utils is over

* loads

* bump download cache

* bump timeout
 2023-11-28 17:36:55 -08:00
George Hotz ab5d14d4ba
MEM -> LOAD (#2492) 
* MEM -> LOAD

* keep legacy working
 2023-11-28 16:46:37 -08:00
chenyu a739c6646e
fp16 in gpt2 attention (#2491) 
* fp16 in gpt2 attention

* HALF
 2023-11-28 19:27:03 -05:00