Conv1d optimize (#392)

* Reorder the conv1d loops in the cpu backend.

* Optimize the 1d convolution.

* Conv1D optimize.

* Fix some clippy lints.
This commit is contained in:
Laurent Mazare
2023-08-10 16:23:52 +02:00
committed by GitHub
parent 0b0fa56978
commit c8039579a5
6 changed files with 62 additions and 22 deletions

View File

@ -31,6 +31,7 @@ clap = { version = "4.2.4", features = ["derive"] }
cudarc = { version = "0.9.13", features = ["f16"] }
# TODO: Switch back to the official gemm implementation once it has caught up.
gemm = { version = "0.15.6", package = "candle-gemm" }
ggblas = "0.1.2"
hf-hub = "0.2.0"
half = { version = "2.3.1", features = ["num-traits", "rand_distr"] }
image = { version = "0.24.7", default-features = false, features = ["jpeg", "png"] }