From b13a82a4387a55df07bec4e2eb6f7a8ebd0b98a2 Mon Sep 17 00:00:00 2001 From: Laurent Mazare Date: Sat, 4 May 2024 10:14:57 +0200 Subject: 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. --- candle-metal-kernels/src/lib.rs | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'candle-metal-kernels') 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); -- cgit v1.2.3