github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-31 01:38:20 -05:00
Code Issues Packages Projects Releases Wiki Activity
10,417 Commits 208 Branches 14 Tags
reprocess_node_experiments
Commit Graph

10417 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
nimlgen
51d6f372e4 nv get classes based on device (#5325) 
* nv get classes

* support in mockgpu

* choose sm based on gpu

* fix

* fix

* fix arch
 2024-07-08 18:25:05 +03:00
chenyu
7d049fc20c move getting 0 and min value of a dtype to dtype.py (#5328) 
cleanup getting base case for reduce ops
[run_process_replay]
 2024-07-08 10:51:56 -04:00
nimlgen
b0c5c58833 nv rm_control to rmctrl type (#5327) 
* nv rm_control to rmctrl type

* fix
 2024-07-08 17:24:33 +03:00
Elias Wahl
73bddc44f6 Fix fake dataloader (#5326) 2024-07-08 09:07:44 -04:00
chenyu
6856f915d6 Tensor.any and Tensor.all (#5320) 
does not work in ptx yet due to how boolean tensor is handled
 2024-07-07 14:36:00 -04:00
chenyu
2029cb7047 support passing None to Tensor.clip (#5319) 
passing None for no upper bound or no lower bound
 2024-07-07 13:04:22 -04:00
chenyu
296a1a36bb update Tensor.round doc and example (#5318) 
document rounding half to even and update examples to show
 2024-07-07 12:10:39 -04:00
chenyu
c1e330f302 Tensor.int and Tensor.bool (#5317) 2024-07-07 11:52:58 -04:00
nimlgen
778d1cdbee nv allocate local memory dynamically (#5277) 
* nv allocate local memory dynamically

* fix

* linter

* linter 2

* linter

* fixes
 2024-07-07 17:34:49 +03:00
qazal
ae10e936e7 UOps.VECTORIZE cleanups [run_process_replay] (#5314) 
* still render_cast

* one extra line ok

* these are all just vectorize

* save space

* behavior change can go in a different diff
 2024-07-07 10:49:08 +03:00
greg-niemeyer
77b2ce9fc9 Add UOps.VECTORIZE [run_process_replay] (#5289) 
* Add UOps.VECTORIZE to core

* Update vectorized cast tests

* Addresses code review comments

- Removes VECTORIZE from LLVMRenderer
- Add line breaks to unduly long lines
- Add noop CAST rule back
- Update asserts and add render_vectorize in
  CSytleLanguage renderer

* Add missing const folding rule for VECTORIZE

Also adds corresponding test

* Fixes test_const_vectorize_fold and add assert

- Use sane types with VECTORIZE in test_const_vectorize_fold
- Add assert that sanity checks the types for VECTORIZE

* Rename test_cast_vectorized_fold

Renames test_cast_vectorized_fold to test_noop_vectorize_fold
because the test targets a very specific rule and there are
other tests for VECTORIZE.

* Revert unrelated changes

---------

Co-authored-by: qazal <77887910+Qazalin@users.noreply.github.com>
Co-authored-by: qazal <qazal.software@gmail.com>
 2024-07-07 09:59:57 +03:00
qazal
2a7282c1e1 test: delete the extra cast in cstyle load [run_process_replay] [no_assert] (#5310) 
* test: delete the extra cast in cstyle load [run_process_replay] [no_assert]

* assert buf_uop

* ImageDType

* ptx is actually a 64bit address
 2024-07-07 09:12:49 +03:00
chenyu
cededd8eb4 minor multi cleanup (#5311) 
add type, move around and some newlines
 2024-07-06 21:55:59 -04:00
qazal
8a99514462 generalize the uops toposort spec to ptx (#5309) 
* generalize spec to ptx

* redundant assert

* extra print
 2024-07-07 00:06:30 +03:00
chenyu
ca0ef1700b use precise::sin in metal (#5307) 2024-07-06 12:47:27 -04:00
qazal
5c2ca7bad4 remove UOps.SPECIAL rendering from llvm (#5306) 2024-07-06 19:28:47 +03:00
chenyu
356e5d2e54 touchup multi dtype in elementwise (#5305) 
only need to check real once, also added type annotation
 2024-07-06 11:54:12 -04:00
qazal
7ddda9f9f1 hotfix: cache seen graphs in fusion (#5302) 2024-07-06 14:13:58 +03:00
qazal
11dfb19b20 track seen graphs in recursive group (#5301) 
* track seen

* maybe never add realized

* ahh it needs to track sts

* delete extra check

* cache typings

* minor cleanup
 2024-07-06 12:39:31 +03:00
qazal
d813617742 prescheduling refactor (#5300) 
* p1

* refactor tuple
 2024-07-06 12:04:03 +03:00
qazal
c1e166c08a fix dtype mismatch for bool ops in multi (#5299) 2024-07-06 11:36:40 +03:00
chenyu
fc03fc025e enable sin on METAL in test_dtype_alu (#5298) 2024-07-05 14:52:09 -04:00
qazal
b369e75ed0 refactor schedule creation (#5297) 2024-07-05 21:14:38 +03:00
qazal
5292d37db6 LoadOps.VIEW in the scheduler spec (#5296) 
* refactor to allow_buffer_view

* tests

* fix multi
 2024-07-05 19:43:50 +03:00
hikettei
1ab7a4cff0 Handling Multiple UnaryOps.BITCAST in Function for Proper Kernel Fusion [run_process_replay] (#5172) 
* [Patch] added an option not to ignore view replacing when doing bitcast

* added the testcase

* [Add] reproduced bitcast cannot be fused into a single kernel in the unittest

---------

Co-authored-by: qazal <77887910+Qazalin@users.noreply.github.com>
 2024-07-05 19:16:44 +03:00
chenyu
43c3f73fbc handcode_bert_opt.py (#5295) 
similar to handcode_resnet50_opt.py, one file to check bert kernels without dataset.
 2024-07-05 11:01:20 -04:00
nimlgen
d7835a705c hotfix: fix metal with vars (#5294) 
* hotfix: fix metal with vars

* one more place
 2024-07-05 16:53:40 +03:00
nimlgen
8a548b0b6e metal support offset (#5293) 2024-07-05 16:13:05 +03:00
qazal
1cefbb33ab uop graph tests + type_verify cleanup (#5292) 
* test_cast_alu_fold

* test_double_cast_fold + these should assert
 2024-07-05 13:00:01 +03:00
qazal
341c4a29d1 hotfix: use dtype.scalar() for rendering cast [run_process_replay] [no_assert] (#5290) 2024-07-05 11:29:35 +03:00
chenyu
87d27c45ec minor _broadcast cleanup (#5286) 
`any(x==0 for x in y)` is `0 in y`.

also `get_args(ConstType)` instead of hard coded `float, int, bool`
 2024-07-04 14:25:24 -04:00
SnakeOnex
8c03816ae9 fix README example (#5284) 
* fixed README example

* README test

* changed py -> python markdown code flags in REAME
 2024-07-04 11:15:07 -04:00
nimlgen
2778b6046c new memory scheduler (#5278) 
* new memory schedule algo

* works

* fix

* fix

* linter

* tiny fixes

* do not optimize copy buffers

* mpre comments

* tiny cleanups
 2024-07-04 18:06:04 +03:00
nimlgen
84b3e3bb6f hcq exec no embedded signal (#5142) 2024-07-04 13:29:21 +03:00
Tobias Fischer
0c3a35e5c2 Stable Diffusion v2 Inference (#5283) 
* model implementation

* clip fix, more qol options
 2024-07-03 22:47:10 -04:00
chenyu
e5ba385f03 remove first contiguous in multi from_sharded (#5121) 
second contiguous guarantees lbs are contiguous going into MultiLazyBuffer, don't need the first contiguous
 2024-07-03 19:42:56 -04:00
chenyu
f1ff65e763 remove "no-nans-fp-math"="true" for LLVM (#5282) 
fixed isnan for llvm (still have issue with < nan)
 2024-07-03 17:52:50 -04:00
chenyu
3929a9dc94 fix UOp.cmp_tuple for ALU (#5280) 
* fix UOp.cmp_tuple for ALU

for ALU, use self.arg instead of self.op to compare

* skip that?
 2024-07-03 14:59:05 -04:00
qazal
a9d6a6c339 verify_lazyop with multi reduce (#5276) 
* outsource the assert to the implicit movement op check

* tests
 2024-07-03 20:15:42 +03:00
George Hotz
16e3b8b013 uops work from lowerer [run_process_replay] (#5279) 2024-07-03 09:40:00 -07:00
chenyu
622b7bd556 simpler TinyJit inside TinyJit detection (#5219) 
* simpler TinyJit inside TinyJit detection

suggested in 73395b998b (commitcomment-143660402)

* cannot repro...

* clear the way out

* finally clear
 2024-07-03 12:28:53 -04:00
gip
04ef0fd328 fix: message when applegpu tools missiong (#5236) 2024-07-03 09:07:09 -07:00
reddyn12
d3e244d8b7 prev speed improvements (#5252) 
Co-authored-by: reddyn <nikidsniper@gmail.com>
 2024-07-03 09:06:01 -07:00
nimlgen
21d41f06a2 nv follows HCQCompatAllocRes protocol (#5275) 
* nv follows HCQCompatAllocRes protocol

* fix amd
 2024-07-03 11:34:10 +03:00
Vyacheslav Pachkov
d3e4e21759 add return type for HCQCompatAllocator _alloc (#5267) 
Co-authored-by: nimlgen <138685161+nimlgen@users.noreply.github.com>
 2024-07-03 10:25:44 +03:00
chenyu
191463a919 add timing to SDXL (#5273) 2024-07-02 23:29:54 -04:00
chenyu
b2c3a28a5e nn.RMSNorm (#5272) 
the norm itself has no significant value to add to Tensor method, but we would want Tensor.normalize
 2024-07-02 21:39:01 -04:00
chenyu
9a2a82a77f test stable diffusion unet in ci (#5268) 
unet is parameterized now so can test a smaller one is ci
 2024-07-02 21:37:52 -04:00
chenyu
ce52b10f6f add a flag DISABLE_LOOP_COLLAPSE (#5270) 
workaround if user encountered UNMUL error
 2024-07-02 20:01:11 -04:00
George Hotz
e53b164e1a small changes from lowerer (#5266) 2024-07-02 15:03:54 -07:00