[Fix] Add compressed_tensors as deps

[ocss884]

Motivation

close #4818

Modifications

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://slack.sglang.ai to discuss your PR.

[zhyncs]

@BBuf @HandH1998 @yundai424 I think that the best way is to make sgl-kernel’s W8A8 turn on by default for compressed_tensor. WDYT

[zhyncs]

In fact, I don't really want to introduce this dependency, at least in srt, because we have already implemented w8w8 int8 and fp8 in sgl-kernel with better performance. The part of compressed_tensors can be completely replaced by the one in sgl-kernel. cc @BBuf

[yundai424]

yeah i think the only reason we need compressed_tensors is because we kind of rely on some of the enum / constants in it for example here. It boils down to whether we can tolerate hard-coding these values

[zhyncs]

It boils down to whether we can tolerate hard-coding these values

I think using an enum is preferable to introducing this dependency.

[zhyncs]

@ocss884 @yundai424 Can you submit a PR to remove this dependency, thank you!