Angelos Katharopoulos 22f9b8a6fc More JIT fixes

2025-07-03 17:33:25 -07:00

..

jit

More JIT fixes

2025-07-03 17:33:25 -07:00

kernels

Use segmented_mm to calculate the MoE gradient

2025-07-02 16:24:21 -07:00

allocator.cpp

Add memory cache to CUDA backend (#2221)

2025-05-30 12:12:54 -07:00

allocator.h

Add memory cache to CUDA backend (#2221)

2025-05-30 12:12:54 -07:00

binary.cpp

Improve metal elementwise kernels (#2247)

2025-06-06 11:37:40 -07:00

binary.h

Fixes for large arrays with a few ops (#1299)

2024-07-30 17:18:39 -07:00

CMakeLists.txt

Fix the JIT

2025-07-03 15:57:43 -07:00

compiled.cpp

Improve metal elementwise kernels (#2247)

2025-06-06 11:37:40 -07:00

conv.cpp

Collection of refactors (#2274)

2025-06-13 10:44:56 -07:00

copy.cpp

Improve metal elementwise kernels (#2247)

2025-06-06 11:37:40 -07:00

custom_kernel.cpp

Fix unintuitive metal kernel caching (#2242)

2025-06-06 20:08:15 -07:00

device.cpp

Collection of refactors (#2274)

2025-06-13 10:44:56 -07:00

device.h

Collection of refactors (#2274)

2025-06-13 10:44:56 -07:00

distributed.cpp

Move common gpu primitives to backend/gpu (#2145)

2025-05-05 13:45:29 -07:00

eval.cpp

Generalize gpu backend (#2138)

2025-04-30 09:08:17 -07:00

event.cpp

Generalize gpu backend (#2138)

2025-04-30 09:08:17 -07:00

fence.cpp

fix input coherent kernel launch (#2153)

2025-05-05 17:30:50 -07:00

fft.cpp

Fix fft for integer overflow (#2161)

2025-05-09 14:25:12 -07:00

hadamard.cpp

Move common gpu primitives to backend/gpu (#2145)

2025-05-05 13:45:29 -07:00

indexing.cpp

Fix scatter_axis on single axis arrays

2025-07-03 14:50:23 -07:00

jit_kernels.cpp

Fix metal jit

2025-07-03 14:09:05 -07:00

kernels.h

Fix metal jit

2025-07-03 14:09:05 -07:00

logsumexp.cpp

Move common gpu primitives to backend/gpu (#2145)

2025-05-05 13:45:29 -07:00

make_compiled_preamble.sh

Dispatch bf16 at run time when using the JIT (#1584)

2024-11-15 16:54:36 -08:00

matmul.cpp

Fix metal jit

2025-07-03 14:09:05 -07:00

matmul.h

Collection of refactors (#2274)

2025-06-13 10:44:56 -07:00

metal.cpp

Generalize gpu backend (#2138)

2025-04-30 09:08:17 -07:00

metal.h

Generalize gpu backend (#2138)

2025-04-30 09:08:17 -07:00

no_metal.cpp

start cuda circle config (#2256)

2025-06-10 21:19:47 -07:00

nojit_kernels.cpp

Fix metal jit

2025-07-03 14:09:05 -07:00

normalization.cpp

Collection of refactors (#2274)

2025-06-13 10:44:56 -07:00

primitives.cpp

Make sliceUpdate general (#2282)

2025-06-12 16:48:54 -07:00

quantized.cpp

5bit quants (#2226)

2025-05-30 12:12:10 -07:00

reduce.cpp

Move common gpu primitives to backend/gpu (#2145)

2025-05-05 13:45:29 -07:00

reduce.h

Reductions update (#1351)

2024-11-04 22:25:16 -08:00

resident.cpp

Generalize gpu backend (#2138)

2025-04-30 09:08:17 -07:00

resident.h

Wired (#1510)

2024-10-25 09:35:33 -07:00

rope.cpp

Fast primitives decide when to use the fallback (#2216)

2025-06-02 13:26:37 -07:00

scaled_dot_product_attention.cpp

Fix unintuitive metal kernel caching (#2242)

2025-06-06 20:08:15 -07:00

scan.cpp

Move common gpu primitives to backend/gpu (#2145)

2025-05-05 13:45:29 -07:00

slicing.cpp

Move common gpu primitives to backend/gpu (#2145)

2025-05-05 13:45:29 -07:00

softmax.cpp

Move common gpu primitives to backend/gpu (#2145)

2025-05-05 13:45:29 -07:00

sort.cpp

Move common gpu primitives to backend/gpu (#2145)

2025-05-05 13:45:29 -07:00

ternary.cpp

Improve metal elementwise kernels (#2247)

2025-06-06 11:37:40 -07:00

ternary.h

Add some internal GPU apis (#1177)

2024-06-04 09:24:26 -07:00

unary.cpp

CUDA backend: unary ops (#2158)

2025-06-09 06:45:08 -07:00

unary.h

Add some internal GPU apis (#1177)

2024-06-04 09:24:26 -07:00

utils.cpp

Move some dims utils to common (#2223)

2025-05-29 06:48:30 -07:00

utils.h

Improve metal elementwise kernels (#2247)

2025-06-06 11:37:40 -07:00