Stefan He

How to run it?

2

It said 'RCTRootView.h' file not found

Question: What's the limit of the corpus size?

1

Like the question said, I was wondering has any one reached the limit that the data/ could contain? Or: Is it no limitation actually, but just make the model performance...

bug: error: image envd-quick-start:dev not found

1

### Are you use the envd server? - [ ] Yes, I am using the envd server. - [X] No, I am not using the envd server. ### Describe the...

[Dashboard | Core] Some Ray subprocesses exited unexpectedly: dashboard [exit code=-11]

5

### What happened + What you expected to happen `ray start --head --metrics-export-port=8080 --block --dashboard-agent-listen-port=52365 --dashboard-host=0.0.0.0 --disable-usage-stats --verbose` will fail in my latest built docker image Error log in terminal:...

[BUG] Single Node Multiple GPU HorovodJob Failure with pyflyte-fast-execute

2

### Describe the bug # Bug Report: Single Node Multiple GPU HorovodJob Failure with pyflyte-fast-execute ## Issue Description When executing a Flyte workflow containing a Multiple GPU HorovodJob using `pyflyte-fast-execute`,...

[Bug] KeyError: 'model.layers.18.mlp.down_proj.weight_scale'

3

### Checklist - [x] 1. I have searched related issues but cannot get the expected help. - [x] 2. The bug has not been fixed in the latest version. -...

## Motivation Gemlite is already integrated (See https://github.com/sgl-project/sglang/pulls?q=is:pr+gemlite+is:closed) in both TorchAO and SGLang, but it's missing in the quantization docs: https://docs.sglang.ai/references/quantization.html Also I benchmarked Gemlite and found it's indeed faster...