Brian Keene 1865299a30 Metal shaders for memory efficient self attention on large sequences (#964) ...

* Metal shaders for efficient self attention on large sequences

Updated fast attention: GEMM-ified with Steel primitives
Uses flash attention 1 for scale correction

* more compiler silencing

* Address rebase issues

* Templatize kernel instantiation, revise cpu bindings

* Safer writes to output

* Permit batch size > 1

* Numerical fixes for sdpa self attention

* Re-enable test, remove unused variable

* add benchmarking script

* Disable sdpa prior to perf tuning, and simplify tests for per-patch CI

2024-06-03 09:16:19 -07:00

..

mlx_tests.py

Make MLX build on x64 macOS (#901)

2024-03-27 06:14:29 -07:00

mpi_test_distributed.py

Comms (#1097)

2024-05-23 17:04:02 -07:00

test_array.py

feat: Added dlpack device (#1165)

2024-05-31 12:29:01 -07:00

test_autograd.py

stable cumprod grad at 0 (#1167)

2024-05-31 12:28:42 -07:00

test_bf16.py

bump pre commit and fix format (#373)

2024-01-04 16:28:52 -08:00

test_blas.py

Fix matvec vector stride bug (#1168)

2024-05-29 12:18:28 -07:00

test_compile.py

Fix a couple bugs (#1161)

2024-05-28 15:18:18 -07:00

test_constants.py

feat: Add numpy constants (#428)

2024-01-11 06:47:29 -08:00

test_conv.py

Add groups to 2-D convolutions (#1129)

2024-05-22 20:01:44 -07:00

test_device.py

Adds device context manager (#679)

2024-02-14 14:14:58 -08:00

test_eval.py

Shared events for synchronization + async eval (#998)

2024-04-17 06:16:02 -07:00

test_fast_sdpa.py

Metal shaders for memory efficient self attention on large sequences (#964)

2024-06-03 09:16:19 -07:00

test_fast.py

No copy command encoder (#986)

2024-04-11 21:15:36 -07:00

test_fft.py

Metal FFT for powers of 2 up to 2048 (#915)

2024-04-11 21:40:06 -07:00

test_graph.py

Multi output primitives (#330)

2024-01-08 16:39:08 -08:00

test_init.py

Make shape a tuple (#591)

2024-01-30 13:11:01 -08:00

test_linalg.py

Implemented Cholesky on CPU (#1119)

2024-05-17 12:31:59 -07:00

test_load.py

Fix logsumexp edge case (#740)

2024-02-25 08:39:55 -08:00

test_losses.py

Feat: Add weights argument in BCE Loss and tests (#620)

2024-02-07 09:39:52 -08:00

test_metal.py

Reset peak memory (#1074)

2024-05-03 17:12:51 -07:00

test_nn.py

Make sure 0 is represented in the quantization (#1016)

2024-04-19 19:47:26 -07:00

test_ops.py

Fix multi-block sort stride management (#1169)

2024-05-31 11:10:54 -07:00

test_optimizers.py

Treate 'minimum' differently in cosine decay (#1138)

2024-05-20 08:00:48 -07:00

test_quantized.py

Rename block sparse (#1149)

2024-05-22 07:48:34 -07:00

test_random.py

Implementation of mlx.random.multivariate_normal (#502) (#877)

2024-04-09 13:50:12 -07:00

test_reduce.py

Add GPU support for uint64/int64 reductions (#569)

2024-01-31 11:18:04 -08:00

test_tree.py

Add isort pre-commit and run (#68)

2023-12-08 11:31:47 -08:00

test_upsample.py

Upsample with bicubic interpolation (#967)

2024-04-10 15:47:22 -07:00

test_vmap.py

Adds radians and degrees (#1011)

2024-04-22 11:17:49 -07:00