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
11,106 commits 280 branches 15 tags 325 MiB
Commit graph

4,667 commits

Author SHA1 Message Date
Sieds Lykles
d9560a631c
 		remove cast between ints if safe (#11946) 2025-09-01 05:56:49 +02:00
Sieds Lykles
a19d689481
 		fix vec dtype _min_max (#11944) 2025-09-01 03:24:07 +02:00
Sieds Lykles
f32f3464d6
 		Can safe cast from certain ints to floats (#11941) 
* add rule

* add some tests

* prevent infinite loop with bfloat16

* add some ints to double and float can_safe_cast

* add tests
 2025-09-01 00:51:24 +02:00
Sieds Lykles
1c6e43c203
 		Double cast is one cast if intermediate cast is safe (#11939) 
* add rule

* add some tests

* prevent infinite loop with bfloat16

* prevent more infinite rewrite
 2025-09-01 00:36:29 +02:00
b1tg
c1eeb3b99c
 		only skip AMD_LLVM (#11934) 
Co-authored-by: b1tg <b1tg@users.noreply.github.com>
 2025-08-31 18:15:47 +03:00
b1tg
75d380a77c
 		fix transcendentals in python renderer (#11932) 
* fix transcendentals in python renderer

* add test

---------

Co-authored-by: b1tg <b1tg@users.noreply.github.com>
 2025-08-31 09:37:17 -04:00
Sieds Lykles
d3252ccd85
 		fix special vmax when arg is UOp (#11930) 2025-08-31 06:54:39 +02:00
chenyu
af89be317e
 		relax rtol for bfloat16 test_dtype_alu (#11926) 2025-08-30 17:16:08 -04:00
qazal
c27b99d68f
 		viz: refactor to indexed rewrite traces (#11923) 2025-08-30 20:01:47 +03:00
qazal
bf0d055b39
 		viz: color by name (#11919) 2025-08-30 16:04:58 +03:00
Sieds Lykles
0bc34c000f
 		simplify range mod its own upper bound (#11917) 
* add rules

* add tests
 2025-08-30 08:37:35 +02:00
chenyu
561318fea7
 		Tensor.cos in test_stype_alu (#11916) 
* Tensor.cos in test_stype_alu

* need this fix anyway
 2025-08-29 20:26:36 -04:00
nimlgen
c6e342cdac
 		mockgpu: no hang if gpuocelot failed (#11915) 2025-08-30 00:44:49 +03:00
chenyu
26d03a86a1
 		test_symbolic_ops.py cleanup (#11895) 2025-08-29 17:11:59 -04:00
b1tg
b2cc06218a
 		python bfloat16 (#11912) 
* python bf16

* _to_torch_storage_type

---------

Co-authored-by: b1tg <b1tg@users.noreply.github.com>
 2025-08-29 15:18:02 -04:00
George Hotz
afad7d0cd1
 		remove dtype from range, it will be dtypes.index soon [pr] (#11914) 
* remove dtype from range, it will be dtypes.index soon [pr]

* a few more
 2025-08-29 09:52:07 -07:00
George Hotz
394c2d1db1
 		update Kernel API in tests + move optimize_local_size (#11907) 2025-08-28 15:12:47 -07:00
nimlgen
fa695ac1ce
 		ci: mac gpuocelot (#11906) 
* gm

* fix?

* ops

* imp

* xx

* add file
 2025-08-28 23:29:43 +03:00
George Hotz
b9b438c516
 		small updates from postopt (#11903) 
* tests from postopt

* modernize

* skip lin tests

* that's fixed?

* skip, not failure
 2025-08-28 12:34:52 -07:00
Ben Waldron
ea1be2e4cd
 		[bounty] Remove using reshape to register symbolic shape (#11771) 
* Modify tests and start work towards removing symbolic reshape

* Refactor symbolic reshape

* fix small error

* much cleaner + fix more tests

* Can remove this now

* Update test_symbolic_ops and test_tiny

* Couple more tests

* Unused import

* More tests and add EXPAND to Tensor.empty

* Fix test beam search

* all int

* Fix rangeify by adding shrink

* Remove OOB check and so fix test_symbolic_jit

* test_symbolic_jit doesn't need OOB Context anymore either

* Should remove that test now

* Cleanups part 1

* fix linters

* Final cleanups

* Don't reassign inside for loop

---------

Co-authored-by: chenyu <chenyu@fastmail.com>
 2025-08-28 12:30:49 -04:00
Ben Waldron
17ecaf4682
 		Add test_variable_empty (#11889) 
* Add test_variable_empty

* Move test and add TODO

---------

Co-authored-by: chenyu <chenyu@fastmail.com>
 2025-08-28 11:38:27 -04:00
Nino Risteski
54be477152
 		rope cache optim for jit prune in llm.py (#11678) 
* rope cache optim for jit prune

* rope test

* tests in test attention

* Revert "rope test"

This reverts commit 69ede543d0.

* lint
 2025-08-28 08:31:29 -07:00
quortus
5f8fe9a331
 		Replace ASSIGN with STORE in test_linearizer (#11821) 2025-08-28 07:33:20 -07:00
geohotstan
4e8370309c
 		Support onnx If OP (#11648) 
* start

* tiny clean up

* whoops, didn't mean to accidentally fix this

* fix .to(device), kinda hacky and this fix makes it slower?

* merge properly

* FINALLY figured out slowness, also hack pylint for now

* add DEBUGONNX print for subgraph

* oops

* WOOOOOOOO SHAPE CACHE 50% SPEED INCREASE

* small fix, but maybe all deterministic Tensor creation in fp should be cached

* cache condition

* sliiiightly cleaner

* better abstraction?

* remove sam from model_benchmark

* remove shape cache speed up for now

* less lines

* isinstance fix

---------

Co-authored-by: chenyu <chenyu@fastmail.com>
 2025-08-28 10:17:35 -04:00
George Hotz
6d6f0dada7
 		support for tuple ranges (#11890) 
* support for tuple ranges

* breaks it
 2025-08-28 07:02:31 -07:00
chenyu
beb5982165
 		FUSE_ATTENTION (#11884) 2025-08-27 19:59:17 -04:00
nimlgen
44816218b5
 		memplan: fix large buffers planning (#11878) 
* memplan: fix large buffers planning

* fix

* fix dsp
 2025-08-27 23:54:27 +03:00
nimlgen
4006366752
 		Revert "memplan: fix large buffers planning (#11876)" (#11877) 
This reverts commit 7f90497efc.
 2025-08-27 22:36:14 +03:00
nimlgen
7f90497efc
 		memplan: fix large buffers planning (#11876) 
* memplan: fix large buffers planning

* fix
 2025-08-27 22:04:15 +03:00
Jordan Chalupka
e9789d8a70
 		Add mxfp4 support (#11873) 
* bump ggml url

* map mxfp4 to tensor

* tests
 2025-08-27 10:56:56 -07:00
Sieds Lykles
d39365809a
 		add ctx to z3_renderer arg (#11867) 
* add ctx to z3_renderer arg

* update symbolic fuzzer

* rewrite u1,u2,u3

* update fuzz_fast_idiv

* remove imports
 2025-08-27 03:38:15 +02:00
chenyu
7028cb4167
 		clean up TestBitcastConstFolding (#11856) 2025-08-26 15:26:47 -04:00
George Hotz
b268755d51
 		small changes from postopt (#11854) 2025-08-26 11:56:16 -07:00
Sieds Lykles
a3aeef45cc
 		associative variation of where branch-merging (#11851) 
* add rule and test

* change comment
 2025-08-26 19:27:05 +02:00
b1tg
1dd613cb89
 		test float_to_bf16 round-to-even behavior (#11849) 
Co-authored-by: b1tg <b1tg@users.noreply.github.com>
 2025-08-26 12:16:10 -04:00
b1tg
409399c609
 		fix nan in float_to_bf16 (#11843) 
Co-authored-by: b1tg <b1tg@users.noreply.github.com>
 2025-08-26 11:42:25 -04:00
chenyu
f28f613f85
 		improved float_to_bf16 (#11848) 
round instead of truncate
 2025-08-26 11:14:06 -04:00
chenyu
337e979a59
 		call dtypes.as_const in Tensor(list) (#11840) 2025-08-25 22:08:26 -04:00
chenyu
ac3449b0c8
 		truncate_fp16 cleanup (#11838) 
native `@` is default
 2025-08-25 19:03:41 -04:00
qazal
a1f6823060
 		viz: memory layout in client side (#11830) 
* viz: memory layout in client side

* update test_viz
 2025-08-25 14:49:33 +03:00
Sieds Lykles
a286a1a6f7
 		Fast idiv try removing factors of two before cast (#11824) 
* try removing factors of two

* dont return if None

* add test
 2025-08-24 20:04:25 +02:00
George Hotz
6540bb32a6
 		move into codegen late [pr] (#11823) 2025-08-24 10:23:25 -07:00
Sieds Lykles
dd69114573
 		Revert "Better div nesting (#11811)" (#11818) 
This reverts commit 952f729b07.
 2025-08-24 18:11:24 +02:00
Sieds Lykles
952f729b07
 		Better div nesting (#11811) 
* remove check

* use fold_divmod_congruence instead of simplify

* adjust tests

* shorten line
 2025-08-24 04:17:40 +02:00
Sieds Lykles
e652062f92
 		tweak divmod_folding condition (#11810) 2025-08-24 02:59:02 +02:00
Sieds Lykles
07d4ed7e4c
 		one more symbolic add variation (#11807) 2025-08-24 01:15:04 +02:00
qazal
0d86288bd7
 		viz: calculate timeline fixed points in client side (#11805) 
* viz: calculate timeline fixed points in client side

* 26 bytes / event

* math
 2025-08-24 01:44:40 +03:00
George Hotz
a75da49951
 		use AxisType for UPCAST/UNROLL (#11800) 
* use AxisType for UPCAST/UNROLL

* fixes

* fix the bug

* fix hack

* bad test

* flaky test
 2025-08-23 14:44:48 -07:00
qazal
2407fecdae
 		viz bytepack format (#11792) 
* viz bytepack format

Training a 1B llama yields ~20M profiler events.

With JSON serialization, the browser tries to load 6GB to memory. This OOMs since each tab is limited to <3-4GB memory usage. Using a packed format, we only need ~600MB.

**Design decisions:**

- Timestamps are in microseconds relative to start time. They're stored in u32, which can express up to ~1 hr of trace events.
- Strings (kernel names, metadata, etc) are deduped.
- Buffer sizes are in u64 nbytes.

More optimization possible:

- The string lookup is a JSON dumped array, we can compress this.
- Can store less for memory by moving the layout to client.

**Results**

|  | Events | JSON | bytepack |
|----------------|---------|-------------|-------------|
| DP=8 llama 1B train (`command: [1]`) | 24M | 5.8GB | 640MB |
| examples/beautiful_mnist.py | 16K | 3.7MB | 745KB |
| examples/gpt2.py | 55K | 12.54MB | 1.40MB |

`[1]`: `VIZ=1 FAKEDATA=1 OFFLOAD_OPTIM=1 DP=8 BS=8 GRADIENT_ACC_STEPS=2 BLOCK_REORDER=0 LR=3e-4 TRAIN_ON_VAL=1 DEFAULT_FLOAT=bfloat16 OPTIM_DTYPE=bfloat16 LLAMA3_SIZE=1B WARMUP_STEPS=36 DECAY_STEPS=360 SEQLEN=8192 PYTHONPATH=. AMD=1 AMD_LLVM=0 MODEL=llama3 python3 examples/mlperf/model_train.py`

* python reference decoder

* 27 bytes / event, 1hr hard limit
 2025-08-23 23:50:21 +03:00
qazal
b12d1d866c
 		count bytes per kernel in test_viz (#11801) 
Currently at ~100 bytes/kernel with JSON.
 2025-08-23 23:35:27 +03:00