vllm

[Quantization] Bump to use latest `compressed-tensors`

#20033

Merged