1,178 commits

Author	SHA1	Message	Date
wozeparrot	4d1a9dca41	fix: don't copy precompiled custom kernel outputs (#16084)	2026-05-07 14:02:38 -07:00
nimlgen	5fa0016ffc	supports_exec_item -> supports_uop (#16033)	2026-05-05 22:41:13 +03:00
wozeparrot	419d525553	feat: handle multioutput kernel grads (#16028)	2026-05-02 22:31:45 -07:00
George Hotz	5f441ecffc	unify reduce + reduce_axis (#15973) ... * unify reduce + reduce_axis * fix all tests * lil cleanups	2026-04-29 10:29:56 -07:00
nimlgen	4164666c72	programinfo (#15942) ... * programinfo * fix * m * x * x * changes * x * fix * rm	2026-04-27 23:12:03 +03:00
nimlgen	96165ff0d1	validate_with_cpu as rewrite (#15938) ... * validate_with_cpu as rewrite * compil * x * linter * moved * fix	2026-04-26 19:58:53 +03:00
nimlgen	d3378010ee	schedule() -> schedule_linear() in tests (batch 1) (#15915) ... * schedule_with_vars -> linear_with_vars in tests * tests batch 1 * batch 2 * estimate_uop * simpler * rm	2026-04-24 23:40:53 +03:00
b1tg	af93a677ae	llm: glm 4.5 air (#15771) ... * llm: glm 4.5 air * clean * clean * remove gguf_size	2026-04-22 22:47:37 +08:00
Christopher Milan	99a0debd62	Device.count() (#15842)	2026-04-21 16:46:38 -04:00
chenyu	9192c93b7e	Tensor.invalid -> Tesnor.invalids (#15849) ... matches ones and zeros, and to not share name with UOp.invalid	2026-04-21 11:19:51 -04:00
nimlgen	01ac1c8c15	remove all run_schedule from tests (#15846)	2026-04-21 12:02:10 +03:00
Christopher Milan	1a8ba4cbd6	CPU renderers use arch (#15839)	2026-04-20 23:38:29 -04:00
George Hotz	5819c0abed	fix gc in gguf (#15820) ... * fix gc in gguf * fix mypy	2026-04-20 10:15:03 +08:00
George Hotz	67ed4c4eb3	move gguf stuff from nn/state.py to llm/gguf.py (#15783) ... * move gguf stuff from nn/state.py to llm/gguf.py * docs	2026-04-20 09:41:43 +08:00
Kartik Vashishta	a1696e8413	objc: fix _classmethods_ dispatch flag (#14854) ... * objc: fix _classmethods_ dispatch flag * test: add objc _classmethods_ regression	2026-04-20 09:35:03 +08:00
chenyu	5bdfd4883f	update test_assign (#15809) ... clean up old skips and update tests	2026-04-18 21:25:44 -04:00
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	9f2a578e26	unskip TestCall.test_call_gemm_uop [pr] (#15786)	2026-04-17 16:18:51 +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
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
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
chenyu	f0c12a2004	another form of assign to itself (#15770)	2026-04-16 15:17:19 -04:00
chenyu	d147e2a549	update test_nested_after_contiguous_store (#15763) ... add kernel counts and some TODOs	2026-04-16 09:59:26 -04:00
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
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	10c262ced8	update tests that use UOp.size (#15753)	2026-04-15 21:58:27 -04:00
George Hotz	1ae6528bb6	move schedule into schedule (#15736) ... * move schedule into schedule * callify to root * sched docs	2026-04-15 11:03:25 +08:00
wozeparrot	2b8d303f75	allreduce in precast dtype (#15689)	2026-04-13 20:24:12 -07:00
George Hotz	7610bdc59e	block multistore, it's not supported (#15708)	2026-04-13 20:57:59 +08:00
George Hotz	4c1fb18a09	Revert "Revert "Tests for GatedDeltaNetBlock + fix multi after assign issue (…" (#15703) ... This reverts commit 0cec42db71.	2026-04-13 19:09:38 +08:00
George Hotz	0cec42db71	Revert "Tests for GatedDeltaNetBlock + fix multi after assign issue (#15700)" (#15702) ... This reverts commit 6f5d756282.	2026-04-13 19:06:44 +08:00
George Hotz	6f5d756282	Tests for GatedDeltaNetBlock + fix multi after assign issue (#15700) ... * broken after/assign test * test for GatedDeltaNet * better comments * fix issue 1 with multi kernel * fix 2 * fix * linter * public api + cleanup	2026-04-13 18:43:23 +08:00
chenyu	e706f408cb	suppress test warnings from numpy (#15688)	2026-04-11 22:33:20 -04:00
Graham Robbins	4ca844e96b	add Q1_0 gguf type (#15683) ... * add Q1_0 * better description * fix trailing whitespace --------- Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>	2026-04-11 18:17:24 +08:00
wozeparrot	457508d5a0	llama: save more 2 (#15681)	2026-04-11 01:03:36 -07:00
George Hotz	b5a9465b13	llm: add support for moonlight (deepseek MLA) (#15466) ... * add gguf Q5_0 * it works * rebase * simpler test * class * less diff * dicts * normal names * simplify * this * simpler * work * work	2026-04-11 10:32:48 +08:00
George Hotz	9092f2a8c0	llm: add shared_expert and rope_dim support from qwen35 (#15673) ... * llm: add shared_expert and rope_dim support from qwen35 * refactor into FFNBlock and TransformerBlock * norms where they belong	2026-04-10 19:18:27 +08:00
George Hotz	35e3983840	Add Q5_0, Q5_1, and bfloat16 GGUF types (#15644)	2026-04-08 17:16:19 +08:00
Christopher Milan	acf239e4d2	specify renderer in DEV, <dev>_<ren>=1 is deprecated (#15551)	2026-03-31 18:35:14 -04:00
nimlgen	5181c8e23a	llm: fix nan in kvcache (#15552)	2026-04-01 00:38:45 +03:00
b1tg	a63392a565	llm: pairwise ranking topk for MoE expert selection (#15499)	2026-03-31 12:46:39 +08:00
nimlgen	0d6fc0f571	jit: graphing in uops (#15489) ... * jit: graphing as rewrite rule * f * +metal,cuda * x * cl * x * x * simpler * f * m * x * revert? * revert2 * back * back * t * x * m * x * c * x * l * x * comment * smaller * rv * x * x	2026-03-27 19:09:02 +03:00
Christopher Milan	bc180a963c	deprecate <dev>=1 in favor of DEV=<dev> (#15467) ... * start work on target * add test * update actions to use DEV * update docs * update readmes * tests need that too * update example * update tests (comments) * fix that test * ruff * mypy * oops * remove getenvs * don't add Target yet * and the test * lint * and docs * more stuff * assert * few more fixes * test assert	2026-03-26 03:48:03 -04:00
George Hotz	fe2690399b	llm: support assistant prefill + refactor to TransformerConfig (#15457) ... * llm: support assistant prefill * refactor to ModelConfig * TransformerConfig * more	2026-03-25 10:50:48 +08:00
George Hotz	a33ac869aa	llm server: temperature + test client (#15444) ... * improvements to the llm server * eval script * eval llm * better eval gets 58.71 * cleanups * add temperature, but multinomial is absurdly slow * claude is so smart * lint * remove slop * no more stop	2026-03-24 21:07:15 +08:00
chenyu	c491345766	pass device into Tensor._frompy (#15385) ... * pass device into Tensor._frompy with this, canonicalize_device is the only usage of Device in tensor.py * export_model.py	2026-03-20 05:09:01 -04:00
George Hotz	3b75d8a7a2	fix double after bug in rangeify (#15381)	2026-03-20 14:53:46 +08:00
Christopher Milan	0c89340a1e	automatically emulate unsupported (tiny) floats [skip_process_replay] (#15366)	2026-03-20 02:31:44 -04:00