Kiv Chen

icon location Mountain View, CA icon location @cmusv. Broke things @risingwavelabs @bytedance @aliyun @hyperledger and 2 others.

Results 21 comments of Kiv Chen

fix(meta): Stale SST Metrics without Alive Compactors #4380

PTAL @hzxa21 @wenym1

fix(meta): Stale SST Metrics without Alive Compactors #4380

> Can you also try this PR locally to verify the metric is correct in grafana dashboard before merging? To verify this metric, I started a full cluster and ran...

fix(meta): Stale SST Metrics without Alive Compactors #4380

I also observed that the object storage's `read-xx` metric dies with the compactor while some of the others do not. Should we start an new issue for it? @hzxa21

CLI: support MySQL engine

Hi, I'm also interested in working on MySQL support for sqllogictest. Is there any notable progress yet?

model(vlm): mistral 3.1

mmmu val: 0.522 cuda device: 2xA100 max throughput: about 60 tokens/s (heavy image prefill) ``` Benchmark time: 182.7642366886139 answers saved to: ./val_sglang.json Evaluating... {'Accounting': {'acc': 0.4, 'num': 30}, 'Agriculture': {'acc':...

model(vlm): mistral 3.1

Hi @FireMasterK, you are correct about the cause. It is now up to date.

model(vlm): mistral 3.1

readily fixed @zhaochenyang20

model(vlm): pixtral

`multi-image` modality has support in the model implementation, but is yet to be integrated to multi-modal data pipelines.(#4754). My private use case branch has a minor refactor about it, but...

model(vlm): pixtral

Hi @yhyang201 , I just pushed the final commits that resolves the two problems. The PR statement has also been updated to reflect additional modifications, especially multimodality support.

model(vlm): pixtral

> 1. When loading the model, the following error occurred: > > > > ``` > > File "/root/sglang/kgl/python/sglang/srt/model_loader/loader.py", line 146, in _initialize_model > > return model_class( > > File...