Commit Graph

13011 Commits

Author SHA1 Message Date
oxrinz
f551a4bded add threefry const folding (#15787)
* prim threefry

* test fix

* clean test

* cleanup

* cleanup 2

* cleanup 3

* fix conflict markers in test_const_folding.py

* update test

* fix lint

* use const instead of value for test
2026-04-20 09:30:03 +08:00
qazal
b05b1010bf viz/cli: ux cleanups, show user python (#15817)
* small fixes

* print python trace

* jsonl

* cleanup fmt, fix tqdm

* print mode

* types

* less

* keep those

* fix

* everyone can print json

* pmc p2
2026-04-20 03:50:48 +03:00
chenyu
8b87b3522a more UOp empty cleanups [pr] (#15818) 2026-04-19 19:48:36 -04:00
chenyu
2a5a6236ac UOp.empty and UOp.empty_like (#15816)
* UOp.empty and UOp.empty_like

Tensor.empty and Tensor.empty_like use these, and removed _buffer_like

* import line
2026-04-19 16:01:01 -04:00
qazal
c6d8753ee1 viz/cli: --json support, refine docs (#15528)
* refine

* remove

* refine

* keep

* need to say this

* back

* feedback

* feedback

* json

* dur_ms

* et_ms

* remove useless thing

* docs

* respect NO_COLOR

* DEBUG also produces valid json
2026-04-19 21:53:38 +03:00
chenyu
50a7b82372 merge untag_and_append and append_after [pr] (#15815)
reads cleaner
2026-04-19 13:13:26 -04:00
chenyu
cace07c87a clean up untag_and_append [pr] (#15812)
replace_uop does not change, and ret.op is always AFTER
2026-04-19 11:23:59 -04:00
wozeparrot
f28ea84de2 llama: fused silu fp8 amax (#15798)
* llama: combined w13

* llama: fused swiglu+fp8

* llama: fix amax interleaving

* llama: don't need seperate matmul
2026-04-19 12:03:55 +08:00
chenyu
5bdfd4883f update test_assign (#15809)
clean up old skips and update tests
2026-04-18 21:25:44 -04:00
nimlgen
022d8c4a11 remove jit_cache usage in extra/examples (#15808)
* remove jit_cache usage in extra/examples

* cached
2026-04-18 23:00:18 +03:00
wozeparrot
06343092c8 llama: combined w13 (#15803) 2026-04-17 22:27:31 -07:00
Christopher Milan
6adf4c3cd9 MOCKGPU interfaces (#15796) 2026-04-17 21:56:29 -04:00
chenyu
8da308573f update test_assign_changes_alt with clone (#15802) 2026-04-17 20:17:37 -04:00
qazal
2581985532 viz/cli: multi device profiler output, print markers (#15795)
* yield

* all devices

* better

* add unittests

* markers like this

* profile_markers work

* less

* update README

* tiny and null
2026-04-17 23:40:10 +03:00
chenyu
0191cc73dc update arange range check (#15794)
it was not checking negative steps correctly
2026-04-17 16:07:50 -04:00
nimlgen
23ca680a3a run_linear (#15784)
* run_linear try 2

* x

* f

* tests

* ctx, cleaner

* r

* x
2026-04-17 22:44:16 +03:00
qazal
8fcaaede9a fix root cause of TestVizIntegration.test_link_sched_codegen flakiness (#15793) 2026-04-17 20:31:52 +03:00
googlefan256
482c8c1ec8 Fix no module named error (#15792) 2026-04-17 19:42:35 +03:00
qazal
a227dbece1 viz/cli: reconstruct DEBUG output (#15791)
* work

* work

* ext

* padding

* at time

* work

* reorder

* less flags

* num_rows

* feedback

* pmc
2026-04-17 18:27:58 +03:00
qazal
601d137e85 viz: rename to rewrites_data, only use ContextVar (#15790)
* viz: rename to rewrites_data

* tms also 0

* gt 0
2026-04-17 17:21:51 +03:00
qazal
afc3904e58 viz/cli: unit tests in CI (#15788)
* simple failing test

* test stdout

* cleanup sqttmap
2026-04-17 22:34:44 +09:00
qazal
9f2a578e26 unskip TestCall.test_call_gemm_uop [pr] (#15786) 2026-04-17 16:18:51 +03:00
qazal
7bdb3adbbf viz/cli: simplification and reordering (#15785)
* remove

* work

* this is all one thing

* the reorder
2026-04-17 15:16:07 +03:00
George Hotz
e1d13bc4fe add GGUF IQ4_XS support (#15766)
* add GGUF IQ4_XS support

* gguf 21

* gguf 21

* use plus

* ggml_common autogen for constant arrays

* fix

* ggml_common in autogen

* inline
2026-04-17 14:43:39 +08:00
wozeparrot
9e60e4a7e7 llama: native fp8 (#15733) 2026-04-16 22:16:05 -07:00
George Hotz
a9b6cfece0 refactor llm into files (#15780)
* refactor llm into files

* chat.html

* tokenizer cleanup

* cleanup

* tests
2026-04-17 12:33:11 +08:00
chenyu
1fac03ce54 softmax and friends to mixin (#15778)
with detach now
2026-04-16 23:03:37 -04:00
George Hotz
ec00cefa5b llm is the only app (#15779)
* tinygrad/llm is the only app

* upd pyproject

* claude refs

* scoping

* min diff
2026-04-17 10:44:48 +08:00
qazal
0e69388f6b viz/cli: add DEBUG, optional number of rows (#15777)
* tabulate switch

* support DEBUG

* --top

* improve

* work

* feedback

* 0

* print_kernel both ways

* simplify
2026-04-17 04:36:47 +03:00
chenyu
2d196fb9bb move Tensor.size to mixin (#15775) 2026-04-16 17:56:17 -04:00
Christopher Milan
9f4b7bed25 add pickled jit regression test (#15774) 2026-04-16 16:59:09 -04:00
qazal
6d9320ffb3 add NO_COLOR (#15765)
* NO_COLOR in cli

* add in helpers

* rm flags

* docs

* fix that

* temp

* Revert "temp"

This reverts commit 7522e664f6.
2026-04-16 22:44:55 +03:00
qazal
12c653a743 remove opts arg in get_program, everything uses opts_to_apply [pr] (#15767)
* check Ops.BEAM in process replay

* remove opts from the get_program api

* lint

* simplify

* cleanup
2026-04-16 22:42:43 +03:00
chenyu
f0c12a2004 another form of assign to itself (#15770) 2026-04-16 15:17:19 -04:00
b1tg
4e88d875ba llm: glm 4.7 flash (#15738)
* glm 4.7

* test

* temperature, server enable_thinking

* --no-think

* remove think stuff
2026-04-16 22:42:04 +08:00
chenyu
d147e2a549 update test_nested_after_contiguous_store (#15763)
add kernel counts and some TODOs
2026-04-16 09:59:26 -04:00
qazal
126cda45f8 viz/cli: cleanups, add memory printer (#15762)
* simple repro

* use context

* work

* memory printer

* rm

* memory printer

* pylint
2026-04-16 22:44:47 +09:00
George Hotz
f57380cbc2 simplify GatedDeltaNetBlock using two state tensors (#15704)
* test double after

* simpler ssm

* no double test
2026-04-16 21:14:00 +08:00
nimlgen
c04f3eaa70 jit: capturedjit is linear (#15743)
* jit: capturedjit is linear

* x

* new beam

* test

* imp

* clean

* spec

* linter
2026-04-16 14:54:39 +03:00
George Hotz
d1cce7a476 put the ranges on store instead of after (#15759)
* put the ranges on store instead of after

* better assert

* fix stuff

* comment out slow rules i don't understand

* simpler rule

* closer

* return false for store

* fix loop

* only a few schedule failures remain

* remove stores to self

* all tests pass locally

* remove junk

* regression test and fix

* better test, bump broken torch count

* bugfix with regression test

* new fusion is better
2026-04-16 19:06:40 +08:00
George Hotz
d24466c844 CALL with return value is FUNCTION (#15758)
* CALL with return value is FUNCTION (GPT try)

* cleanups
2026-04-16 13:25:07 +08:00
chenyu
218d6b8988 delete old UOp.size [pr] (#15756) 2026-04-15 23:21:00 -04:00
wozeparrot
d090732270 usbgpu: reset endpoint for custom fw (#15754) 2026-04-15 20:01:27 -07:00
Muzammil
983a7bb576 exclude __del__ from TRACEMETA wrapping (#15747)
Session-Id: 019d9234-2531-75a0-a252-f0302cd9931f
2026-04-16 10:49:55 +08:00
chenyu
8bd4fead26 UOp.size -> prod(max_shape) (#15755)
and more test updates
2026-04-15 22:41:30 -04:00
chenyu
10c262ced8 update tests that use UOp.size (#15753) 2026-04-15 21:58:27 -04:00
qazal
96092d110c fix process_replay Ops.BEAM [pr] (#15752) 2026-04-16 07:35:28 +09:00
chenyu
41421c3b48 BUFFER size is their arg (#15750) 2026-04-15 18:08:29 -04:00
Christopher Milan
be8005c5dc DEV: secondary targets (#15748) 2026-04-15 17:26:20 -04:00
chenyu
507c02cecb fix symbolic contiguous_view_offset (#15749)
* fix symbolic contiguous_view_offset

* flatten
2026-04-15 16:54:38 -04:00