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
|
2026-04-07 21:55:48 -04:00
|
|
|
torch==2.11.0
|
|
|
|
|
torchaudio==2.11.0
|
2024-06-24 12:11:53 +08:00
|
|
|
# These must be updated alongside torch
|
2026-04-07 21:55:48 -04:00
|
|
|
torchvision==0.26.0 # 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
|
2026-03-30 18:36:18 +02:00
|
|
|
flashinfer-python==0.6.7
|
|
|
|
|
flashinfer-cubin==0.6.7
|
2026-03-10 21:17:35 -07:00
|
|
|
# Cap nvidia-cudnn-frontend (transitive dep of flashinfer) due to
|
|
|
|
|
# breaking changes in 1.19.0
|
|
|
|
|
nvidia-cudnn-frontend>=1.13.0,<1.19.0
|
2026-03-01 18:44:57 -05:00
|
|
|
|
|
|
|
|
# QuACK and Cutlass DSL for FA4 (cute-DSL implementation)
|
2026-04-02 13:02:37 -04:00
|
|
|
nvidia-cutlass-dsl>=4.4.2
|
|
|
|
|
quack-kernels>=0.3.3
|
