nimlgenandGitHub 8d41b3eb3f beam=16 makes gpt2 gpu-time < 5ms on 3090 (#2154)

2023-10-27 10:21:27 -10:00

..

codegen

wmma: refactor tensor cores using existing local dims (#2097)

2023-10-25 13:10:46 -04:00

features

beam=16 makes gpt2 gpu-time < 5ms on 3090 (#2154)

2023-10-27 10:21:27 -10:00

nn

hlb cifar touchups (#2113)

2023-10-18 16:26:15 -07:00

renderer

minor triton casting refactor (#2118)

2023-10-20 12:11:55 -07:00

runtime

init cudagraph (#2153)

2023-10-27 16:19:50 -04:00

shape

MulNode.__lt__ rule (#2086)

2023-10-17 13:18:35 -07:00

graph.py

uops graphing + phi (#2120)

2023-10-19 22:26:28 -07:00

helpers.py

Add dictionary keys to reduce db size (#2131)

2023-10-24 10:49:22 -04:00

jit.py

hip multigpu training (#1878)

2023-10-24 17:35:53 -04:00

lazy.py

stable diffusion < 324ms (#2129)

2023-10-24 14:56:12 -04:00

mlops.py

add mops to graph, refactor IMAGE (#2100)

2023-10-17 21:27:51 -07:00

ops.py

perf: lazyop as dataclass (#1603)

2023-10-25 17:54:30 -04:00

realize.py

Disable logging in early compile2 and lower kernel counts (#2090)

2023-10-16 20:15:24 -07:00

tensor.py

Tensor.uniform set default to standard uniform (#2158)

2023-10-27 16:15:30 -04:00