vLLM 0.16?
#2
by
MMaxHugg
- opened
What do you mean with Version "0.16.0+" required?
"Requires vLLM with NVFP4 support (0.16.0+), Transformers 5.0.0+"
Actually there is only a vLLM v0.15.1 docker image out there and only one 0.16 rc tag - was this a mistake or do we really need to wait for vLLM 0.16?
I only tested with a pull from main and my own fork that has pending PR changes that fix RTX Blackwell P2P, and tp=2 issues.
Not sure if the 15 release works.