diff options
| author | Laurent Mazare <laurent.mazare@gmail.com> | 2024-05-04 10:14:57 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-05-04 10:14:57 +0200 |
| commit | b13a82a4387a55df07bec4e2eb6f7a8ebd0b98a2 (patch) | |
| tree | aed5a019e7e053900ffa5be57ddfd20bdfad8582 /candle-metal-kernels | |
| parent | 59b18d974ec3cad6963b774aa245e23f8c80414f (diff) | |
| download | candle-b13a82a4387a55df07bec4e2eb6f7a8ebd0b98a2.tar.gz candle-b13a82a4387a55df07bec4e2eb6f7a8ebd0b98a2.tar.bz2 candle-b13a82a4387a55df07bec4e2eb6f7a8ebd0b98a2.zip | |
Separate quantized phi-3 implementation. (#2157)
* Separate quantized phi-3 implementation.
* Integrate the quantized phi3 model.=
* Small fixes, get the generation to work properly.
* Keep the old llama implementation around.
* Change the default.
Diffstat (limited to 'candle-metal-kernels')
| -rw-r--r-- | candle-metal-kernels/src/lib.rs | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/candle-metal-kernels/src/lib.rs b/candle-metal-kernels/src/lib.rs index c08e44fe..814ca0b9 100644 --- a/candle-metal-kernels/src/lib.rs +++ b/candle-metal-kernels/src/lib.rs @@ -350,7 +350,7 @@ pub fn call_unary_contiguous_tiled( let pipeline = kernels.load_pipeline(device, Source::Unary, kernel_name.0)?; let encoder = command_buffer.new_compute_command_encoder(); let tile_size = 2; - let tiles = length.div_ceil(tile_size); + let tiles = (length + tile_size - 1) / tile_size; encoder.set_compute_pipeline_state(&pipeline); |