radixark/miles

Reported Failure of Qwen3 14B

Open

#530 geöffnet am 29. Jan. 2026

Auf GitHub ansehen
 (3 Kommentare) (0 Reaktionen) (1 zugewiesene Person)Python (244 Forks)github user discovery
bughelp wanted

Repository-Metriken

Stars
 (1.513 Stars)
PR-Merge-Metriken
 (Durchschn. Merge 9T 17h) (200 gemergte PRs in 30 T)

Beschreibung

I highly think that Qwen3 235B should work, since we do a bunch of experiments on it. But, Qwen3 14B (I think it's a dense model), is also reported to not work.

https://huggingface.co/Qwen/Qwen3-14B

My suggestion is that, could try to install latest Miles docker and debug with it. May just need to use Megatron + SGLang.

Contributor Guide