Add KV-Cache int8 quant support

[YanyunDuanIEI]

Add KV-Cache int8 quant support

Support [layer_level] and [group_level] KV-Cache int8 quant.

• [layer_level] use common scale factors for each layer.
• [group_level] group the head_size according to group_size, with each group_size, the scaling factor of key/value corresponding to the same value.

---

KV-Cache int8 quant (Click to Expand)

Get the scaling factor by calibration

Support to calibrate the KV-cache by datasets:

• [examples/int8/calibrate.py] calibrate and save to pth.
• [export_kv_params.py] save scaling factors to json.

Using KV-Cache int8

• kv_cache_dtype="int8"
• kv_quant_params_path=kv_quant_params_path
• kv_quant_group=kv_quant_group

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @YanyunDuanIEI.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[YanyunDuanIEI]

Would it be viable to hasten the review process?