2024-04-05 00:39:17 -07:00
|
|
|
# Common dependencies
|
2025-03-08 17:44:35 +01:00
|
|
|
-r common.txt
|
2024-04-05 00:39:17 -07:00
|
|
|
|
2025-10-08 18:40:42 +01:00
|
|
|
numba == 0.61.2 # Required for N-gram speculative decoding
|
2025-03-05 09:55:00 +05:30
|
|
|
|
2024-04-05 00:39:17 -07:00
|
|
|
# Dependencies for NVIDIA GPUs
|
2026-02-03 18:30:47 -05:00
|
|
|
ray[cgraph]>=2.48.0
|
2026-02-08 16:51:09 -05:00
|
|
|
torch==2.10.0
|
|
|
|
|
torchaudio==2.10.0
|
2024-06-24 12:11:53 +08:00
|
|
|
# These must be updated alongside torch
|
2026-02-08 16:51:09 -05:00
|
|
|
torchvision==0.25.0 # Required for phi3v processor. See https://github.com/pytorch/vision?tab=readme-ov-file#installation for corresponding version
|
2025-10-09 17:10:02 -04:00
|
|
|
# FlashInfer should be updated together with the Dockerfile
|
2026-02-20 16:37:31 -05:00
|
|
|
flashinfer-python==0.6.4
|
2026-03-01 18:44:57 -05:00
|
|
|
|
|
|
|
|
# QuACK and Cutlass DSL for FA4 (cute-DSL implementation)
|
|
|
|
|
nvidia-cutlass-dsl>=4.4.0.dev1
|
|
|
|
|
quack-kernels>=0.2.7
|
