geohot
a8d13380f9
simpler, but not inverse
2026-06-25 10:24:01 -07:00
geohot
afb0463f10
fix bitcast
2026-06-25 10:11:13 -07:00
geohot
947d6a3c69
enforce a cast for bitcast
2026-06-25 10:00:14 -07:00
George Hotz and GitHub
dc57efc3a3
move down dtype decomps (try 2) ( #16746 )
...
* move down dtype decomps (try 2)
* gpt fixes
* fixes
* that did need reindex
* valid.where
* upd to master
* reindex
2026-06-25 09:49:43 -07:00
George Hotz and GitHub
b5e0bca764
move the load/store gate back to gater ( #16747 )
...
* move the load/store gate back to gater
* order
2026-06-25 09:29:32 -07:00
geohot
fa60fa96a6
hotfix: decomp in pyproject
2026-06-25 08:56:34 -07:00
George Hotz and GitHub
e72ecd101f
split decomp into files ( #16745 )
...
* split decomp into files
* that's in op
2026-06-25 08:40:02 -07:00
George Hotz and GitHub
54887cbbdf
move decomp files into codegen ( #16744 )
2026-06-25 08:19:02 -07:00
chenyu and GitHub
96c9a68205
imrpove some type casts ( #16743 )
2026-06-25 10:48:28 -04:00
chenyu and GitHub
db2257ad09
don't run tinyboxbenchmark on fork ( #16741 )
2026-06-25 09:12:56 -04:00
qazal and GitHub
daf8c8f509
viz: simpler test_link_sched_codegen, merge with the other codegen test ( #16739 )
...
* simpler test_link_sched_codegen
* merge test_codegen_tracing in the other test
* include BEAM.value
2026-06-25 15:02:00 +09:00
qazal and GitHub
695ffb3a08
viz: fix rendering custom Ops.PROGRAM source ( #16738 )
...
* failing test
* fix
* simplify test
2026-06-25 14:21:31 +09:00
George Hotz and GitHub
2f91a401a1
move late decomp after new style ( #16734 )
2026-06-24 22:04:59 -07:00
qazal and GitHub
cb9df8bd33
fix profile.sh BENCHMARK=3 exit status ( #16736 )
2026-06-25 13:41:03 +09:00
sirhcm and GitHub
63f0e08593
start benchmarks cleanup ( #16729 )
2026-06-24 23:33:46 -04:00
wozeparrot and GitHub
e0c69d7a12
llama: fused grad quantize ( #16731 )
2026-06-24 23:25:39 -04:00
chenyu and GitHub
65dd099b63
support Invalid in _pad_constant [PR] ( #16733 )
...
and simplified image_conv2d
2026-06-24 21:03:36 -04:00
George Hotz and GitHub
9d5d2253b3
refactor to image_valid_dims ( #16730 )
2026-06-24 16:53:09 -07:00
nimlgen and GitHub
b99b9e1875
do not assert index in mem coalesing ( #16723 )
2026-06-24 14:28:49 +03:00
f38836d14d
cuda: pass kernel dynamic shared memory in the graph runner ( #16699 )
...
The CUDA graph runner hardcoded sharedMemBytes=0 in the kernel node params, so a custom_kernel
using dynamic shared memory (extern __shared__) failed with CUDA error 700 on graph replay
while the eager path worked. Pass runtime.smem (0 for normal kernels, which use static shared).
Co-authored-by: Claude Opus 4.8 (1M context) <[email protected]
2026-06-24 14:20:10 +03:00
qazal and GitHub
535c806c9f
llama: profile.sh BENCHMARK=3 ( #16722 )
2026-06-24 19:20:51 +09:00
chenyu and GitHub
857a435af2
remove dead added_ox ( #16721 )
2026-06-23 23:59:17 -04:00
chenyu and GitHub
687ade119e
IMAGE hand_coded_optimizations update ( #16720 )
2026-06-23 21:55:28 -04:00
George Hotz and GitHub
0a8e61d0c5
switch to the new memory coaleser [pr] ( #16716 )
...
* switch to the new memory coalese
* move that stuff
* copy in allowed length logic
* mulitple buffers
* new coalese is better
* fine
* earlier
* fixes
* work
* work
* valid
* stack on index const
2026-06-23 18:03:48 -07:00
wozeparrot and GitHub
dfea9e7994
llama: fused silu mul quantize mxfp8 ( #16704 )
2026-06-23 16:59:50 -07:00
chenyu and GitHub
ce87d80911
better _drop_valid_stmts [pr] ( #16719 )
...
also dropped the unused is_increasing
2026-06-23 19:35:01 -04:00
George Hotz and GitHub
5a2b3b7b06
early dtype decomp ( #16718 )
...
* early dtype decomp
* simplify
* cleanup
* that goes there
* doing too much
* stupid symbolic rules
2026-06-23 16:07:20 -07:00
sirhcm and GitHub
116045cc8e
ci: remove tensorflow from testoptim ( #16717 )
2026-06-23 18:11:48 -04:00
nimlgen and GitHub
7c1d0b6d9a
hcq2: use shrink(bitcast) ( #16713 )
...
* hcq2: use shrink(bitcast)
* x
2026-06-23 18:11:39 +03:00
George Hotz and GitHub
c9dc1d63cc
small changes from new codegen ( #16712 )
...
* small changes from new codegen
* shrink/flatten
2026-06-22 17:44:15 -07:00
sirhcm and GitHub
da98fae9e1
ci: try parallelizing tc tests ( #16710 )
2026-06-22 20:43:32 -04:00
chenyu and GitHub
15988b5941
contiguous to mixin and cleanups [PR] ( #16711 )
2026-06-22 20:18:18 -04:00
sirhcm and GitHub
cbfcf36e44
ci: remove generate_dataset and CL misc ( #16709 )
2026-06-22 18:01:07 -04:00
nimlgen and GitHub
f9c8c697d6
hcq2: drop args after inner deps ( #16708 )
2026-06-22 23:26:11 +03:00
chenyu and GitHub
0138480910
dropout and scaled_dot_product_attention to mixin ( #16707 )
2026-06-22 16:17:45 -04:00
chenyu and GitHub
33b635d23a
Tensor.train -> TRAINING [PR] ( #16705 )
...
* Tensor.train -> TRAINING [PR]
* doc
2026-06-22 15:13:22 -04:00
chenyu and GitHub
625d8bbd0d
TRAINING ContextVar ( #16703 )
2026-06-22 13:03:08 -04:00
wozeparrot and GitHub
fe9b19b12d
llama: more mp mem fixes ( #16701 )
...
* llama: more mp mem fixes
* clean: unused
* fix: batch
2026-06-22 10:54:35 -04:00
chenyu and GitHub
267af9c601
full_like to CreationMixin [PR] ( #16702 )
2026-06-22 09:33:23 -04:00
chenyu and GitHub
97da54b9d6
more method to CreationMixin [PR] ( #16698 )
2026-06-22 00:01:22 -04:00
chenyu and GitHub
fd0dc40689
clean up CreationMixin and DTypeMixin [PR] ( #16697 )
2026-06-21 21:13:40 -04:00
chenyu and GitHub
2d8b802958
contiguous in wino conv ( #16696 )
...
also fixed test_counters
2026-06-21 17:11:46 -04:00
chenyu and GitHub
ba1d3baae8
masked_select and nonzero to mixin [PR] ( #16695 )
...
with a .data stub
2026-06-21 15:10:44 -04:00
chenyu and GitHub
d80a41d559
some rand method to RandMixin [PR] ( #16693 )
2026-06-21 12:16:51 -04:00
wozeparrot and GitHub
5164c21b44
gemm: keep shape thru mxfp8 quantize ( #16692 )
2026-06-20 22:28:53 -07:00
chenyu and GitHub
58ff75272e
const_like and invalids to mixin [PR] ( #16690 )
...
* const_like and invalids to mixin [PR]
* empty_like
* einsum
* type
2026-06-21 00:02:29 -04:00
chenyu and GitHub
b50da5c205
move Tensor.__getitem__ to mixin [PR] ( #16689 )
2026-06-20 22:01:45 -04:00
chenyu and GitHub
4618d27129
final const cleanups [PR] ( #16688 )
2026-06-20 21:38:16 -04:00
chenyu and GitHub
9ae0a93d0e
more const cleanups [PR] ( #16682 )
2026-06-20 20:41:43 -04:00
George Hotz and GitHub
30830850a9
small changes from new codegen ( #16681 )
...
* small changes from new codegen
* revert that
2026-06-19 18:29:01 -07:00
geohot
sirhcmandGitHub