Commit Graph

216 Commits

Author SHA1 Message Date
Noah Kasmanoff
de3d2467a3
Update: Fast GeLU Approximation (#744)
* add: fast gelu approx

* fix docs

* Update gelu_fast_approx function documentation

* Update python/mlx/nn/layers/activations.py

Co-authored-by: Awni Hannun <awni.hannun@gmail.com>

* fix: test gelu

---------

Co-authored-by: Awni Hannun <awni.hannun@gmail.com>
2024-02-26 21:08:50 -08:00
Awni Hannun
fe1dabf272
Fix compile with non standard types (#745)
* refactor tree utils

* fix compile + tree code refactor

* Add an extra test

* add a few missing activations to docs

* hash structure

* Encode the full argument structure

---------

Co-authored-by: Angelos Katharopoulos <a_katharopoulos@apple.com>
2024-02-26 19:28:53 -08:00
Hinrik Snær Guðmundsson
08226ab491
added atleast *args input support (#710)
* added atleast list(array) input support

* function overloading implemented

* Refactoring

* fixed formatting

* removed pos_only
2024-02-26 11:17:59 -08:00
Chime Ogbuji
3b661b7394
Add linear warmup and schedule joining for use with existing schedules (#721)
* Add linear warmup to schedules for use with existing schedules

* Changed parameters for simplicity of most common case (0 initial value)

* Added ScheduleJoiner and updated documentation

* ScheduleJoiner -> join_schedules (ala optax #)

* black compliance

* Different evaluation of schedules

* nits

---------

Co-authored-by: Awni Hannun <awni@apple.com>
2024-02-26 07:28:48 -08:00
Awni Hannun
e6418781ab
Fix logsumexp edge case (#740)
* fix logsumexp

* fix inf constant

* also fix power grad

* fix ternary dispatch
2024-02-25 08:39:55 -08:00
Gabrijel Boduljak
22364c40b7
Upsample2d (#414)
Co-authored-by: Angelos Katharopoulos <a_katharopoulos@apple.com>
Co-authored-by: Awni Hannun <awni.hannun@gmail.com>
2024-02-23 09:55:04 -08:00
Noah Farr
d729a1991b
Fix arange with inf step (#686)
* Fix case for step=inf in arange and add inf check for start/stop

* Add test cases for arange

* Update ops.cpp to include climits header

* Fix arange

* Fix formatting

* Refactor

* Add missing include
2024-02-23 06:18:15 -08:00
Awni Hannun
5798256fcf
Shapeless compilation for some graphs (#687)
* shapeless compilation for some graphs

* update compile benchmark

* default compile a few activations

* buffer donation

* bugfix

* shapeless fix

* update tests to work for cpu and gpu fusion

* test kwargs

* add kwargs to compile

* Recompile when python arguments change

* no compile for tanh

* some constant tests

---------

Co-authored-by: Angelos Katharopoulos <a_katharopoulos@apple.com>
2024-02-19 21:43:54 -08:00
Awni Hannun
d0fda82595
fix tolist for half types (#702) 2024-02-19 09:44:27 -08:00
Hinrik Snær Guðmundsson
f883fcede0
Added support for atleast_1d, atleast_2d, atleast_3d (#694) 2024-02-19 09:40:52 -08:00
Srimukh Sripada
818cda16bc
Support LR schedulers (#334)
* Add a few LR schedulers

* Move parents's constructor call to the top

* Fix docstring

* refactor optimizers into two files

* add docs

* nit

* Fix Callable type annotation for python 3.8

---------

Co-authored-by: Awni Hannun <awni@apple.com>
Co-authored-by: Angelos Katharopoulos <a_katharopoulos@apple.com>
2024-02-15 11:26:20 -08:00
toji
85143fecdd
improved error msg for invalid axis(mx.split) (#685)
* improved error msg for invalid axis(`mx.split`)

* Apply suggestions from code review

Co-authored-by: Awni Hannun <awni.hannun@gmail.com>

* fixed formatting issue

---------

Co-authored-by: Awni Hannun <awni.hannun@gmail.com>
2024-02-15 07:25:38 -08:00
Diogo
35431a4ac8
Adds device context manager (#679) 2024-02-14 14:14:58 -08:00
Awni Hannun
ccf1645995
Custom primitive + RoPE fat op (#676)
* extensions start

* rope custom op

* fix build

* docs + rope benchmark

* fix test

* Add a Metal kernel for RoPE

* Fix position of traditional

* transform tests

* Move rope computation to float and fix tests

* Fix the test and a typo

* change to fast

* fix no metal build

---------

Co-authored-by: Angelos Katharopoulos <a_katharopoulos@apple.com>
2024-02-14 14:04:25 -08:00
Noah Farr
0c65517e91
Return empty array when repeats is 0 in mx.repeat (#681)
* Return empty array when repeats is 0

* Add test case for repeats = 0
2024-02-13 17:49:31 -08:00
Gabrijel Boduljak
e54cbb7ba6
Pooling layers (#357)
Co-authored-by: Angelos Katharopoulos <a_katharopoulos@apple.com>
Co-authored-by: Awni Hannun <awni@apple.com>
2024-02-12 22:08:13 -08:00
Angelos Katharopoulos
40c108766b
Quantized matmul fix (#677)
* Fix qmv for small or unaligned matrices

* Fix qmm
2024-02-12 18:54:21 -08:00
Nripesh Niketan
0dbc4c7547
feat: Update pre-commit-config.yaml (#667) 2024-02-11 06:08:20 -08:00
Awni Hannun
b96be943dc
bug fix (#658) 2024-02-09 16:50:45 -08:00
Abdussamet Türker
b670485185
Remainder negative numerator bug fixed (#641)
Co-authored-by: Angelos Katharopoulos <a_katharopoulos@apple.com>
2024-02-09 16:49:14 -08:00
Diogo
b57bd0488d
Metadata support for safetensors (#639)
* metadata support for safetensors

* aliases making it alittle more readable

* addressing comments

* python binding tests
2024-02-08 19:33:15 -08:00
Awni Hannun
5c03efaf29
Compile docs (#653)
* compile docs

* docs nits + comments
2024-02-08 11:21:50 -08:00
LeonEricsson
7dccd42133
updated calls to use loc &scale (#643) 2024-02-08 09:01:59 -08:00
Awni Hannun
1b97b2958b
Compile with capture (#629)
* Simple kernel generation

* Remove the generate kernel from graph_utils

* fix multi-output with compile

* fuse with stopgrad

* v1 input, output capture in compile

* cleanup tree update with visitor update

* nit

* remove todo

* state for model, optional explicit init and more pure optimizer steps

* move learning rate to state

* add lr to opt state, some fixes in capture

* fix optim

* update tuple of containers as well

* fix stream for compiled output

* rng state for compile

* nit

* updates and comments

---------

Co-authored-by: Angelos Katharopoulos <a_katharopoulos@apple.com>
2024-02-07 17:29:22 -08:00
Awni Hannun
e5e816a5ef
fix sequential with empty modules at end (#647) 2024-02-07 13:22:27 -08:00
Noah Farr
5fd11c347d
Add loc and scale to random.normal (#638)
* Add loc and scale to random.normal

* Add tests for loc and scale for random.normal

* Run pre-commit hooks

* Fix code review
2024-02-07 11:49:59 -08:00
Aryan Gupta
ef73393a19
Feat: Add weights argument in BCE Loss and tests (#620) 2024-02-07 09:39:52 -08:00
Angelos Katharopoulos
ea406d5e33
CI change (#645)
* CI update

* Skip large binary test for now

* Upgrade pip

* Add proper env variable skipping

* Update the CI

* Fix workflow name

* Set the low memory flag for the tests

* Change build process

* Add pip upgrade

* Use a venv

* Add a missing env activate

* Add setuptools

* Add twine upload back

* Re-enable automatic release builds
2024-02-07 06:04:34 -08:00
Awni Hannun
d40a04f8dc
minor fixes (#631)
* minor fixes

* var with ddof >= nelements
2024-02-05 13:27:49 -08:00
Awni Hannun
d75ae52ecd
Compile primitive (#571)
* Compiled primitive with basic binary, unary graph-level fusion
2024-02-05 06:51:22 -08:00
Awni Hannun
5c3ac52dd7
fix test (#627) 2024-02-04 16:18:03 -08:00
Avikant Srivastava
11a9fd40f0
fix: handle linspace function when num is 1 (#602)
* fix: handle linspace function when num is 1

* add comment

* fix test case

* remove breakpoint
2024-02-04 11:03:49 -08:00
Daniel Strobusch
4fd2fb84a6
make python array SupportsAbs conform (like numpy) (#624) 2024-02-04 09:31:02 -08:00
Daniel Strobusch
9852af1a19
fix "shape" docstring. (#623) 2024-02-04 09:21:22 -08:00
AtomicVar
83f63f2184
Add Margin Ranking Loss (#536) 2024-02-02 10:57:31 -08:00
Awni Hannun
cb6156d35d
Fix eval in trace bugs (#612)
* Fix eval in trace bugs

* comment nit
2024-02-02 09:57:12 -08:00
Awni Hannun
e88e474fd1
Reduce vmap + some fixes (#601) 2024-02-01 11:30:28 -08:00
David Koski
601c6d6aa8
Fix for AdaDelta (#603)
- state was being read from parameter "s"
- but being stored in parameter "u"
2024-02-01 09:56:27 -08:00
Angelos Katharopoulos
ba8d6bf365
Change the transformer to norm_first by default (#599) 2024-01-31 12:55:30 -08:00
Sugato Ray
4a5f3b21bb
Add py.typed to support PEP-561 (type-hinting) for mlx (#588)
* Add `py.typed` to support PEP-561 (type-hinting)

This adds support for type-hinting information as laid in [PEP-561](https://peps.python.org/pep-0561/).

* add py.typed to MANIFEST.in
2024-01-31 12:05:42 -08:00
Vijay Krish
fcc5ac1c64
Add GPU support for uint64/int64 reductions (#569) 2024-01-31 11:18:04 -08:00
nathan
bad67fec37
Added TeX line breaks to mlx.optimizers.Lion docstring (#595)
Fixes the "misplaced &" MathJax error in documentation.
2024-01-30 19:37:34 -08:00
Angelos Katharopoulos
0de5988f92
Custom VJP and checkpointing (#541)
* Implement custom_vjp and checkpointing
* Add a dependency management primitive
* Change the eval order to deep branches first
* Add graph depth tracking to the array
2024-01-30 16:04:45 -08:00
Jacket
143e2690d5
Fix SGD implementation (#473) 2024-01-30 15:50:46 -08:00
Awni Hannun
09b9275027
Make shape a tuple (#591)
* shape tuple

* also remove simplify from docs

* rebase
2024-01-30 13:11:01 -08:00
Andre Slavescu
d3a9005454
Softshrink mapping + op (#552)
* Added Softshrink mapping + op

* formatting

* docs + nits in docstring

---------

Co-authored-by: Awni Hannun <awni@apple.com>
2024-01-30 12:56:28 -08:00
Jacket
3f7aba8498
Implement diagonal operator (#562)
* Implement diagonal operator

This implements mx.diagonal in operator level, inspired by
@ManishAradwad.

* added `mx.diag` with tests

* corrected few things

* nits in bindings

* updates to diag

---------

Co-authored-by: ManishAradwad <manisharadwad@gmail.com>
Co-authored-by: Awni Hannun <awni@apple.com>
2024-01-30 09:45:48 -08:00
Awni Hannun
3c2f192345
Propagate nans in binary ops (#579)
* propagate nans in binary ops

* handle empty matmul

* cpu minimum/maximum propagate nan

* benchmark maximum

* add min as well

* throw on negative indices with full

* verbose on linux

* fix matmul for zero K
2024-01-29 11:19:38 -08:00
Angelos Katharopoulos
37d98ba6ff
No gil eval (#565) 2024-01-26 22:03:52 -08:00
Awni Hannun
8993382aaa
Buffer Donation (#519)
* buffer donation

* fix to move shared pointer

* format

* gpu in place for copy and binary

* revert ops test

* cpu in place

* a little cleanup

* remove useless bench
2024-01-26 16:30:33 -08:00