Qwen 32b fine tune using axolotl

[ccocks]

I need someone to get me started with this. I am trying to use https://modal.com/docs/examples/llm-finetuning but it is very confusing because I don't see qwen examples

[ccocks]

• actually finetuning fluently-lm/FluentlyLM-Prinum

[NanoCode012]

Hey, we have qwen2 examples here: https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/qwen2

Your target model is of model_type qwen2 as well, so the above should work.

[NanoCode012]

In particular, you may want to reference the qlora_fsdp example. Feel free to convert it into full fine-tuning by removing

-adapter: qlora

[NanoCode012]

I haven't seen that blog before, but we also have modal integration natively. When running any of the axolotl commands, you can prepend with --cloud config.yaml to run it there https://axolotl-ai-cloud.github.io/axolotl/docs/cli.html#remote-compute-with-modal-cloud

[ccocks]

Thank you, this is what I was looking for.

…

________________________________ From: NanoCode012 ***@***.***> Sent: Monday, March 31, 2025 8:24 PM To: axolotl-ai-cloud/axolotl ***@***.***> Cc: blazgocompany ***@***.***>; Author ***@***.***> Subject: Re: [axolotl-ai-cloud/axolotl] Qwen 32b fine tune using axolotl (Discussion #2459) I haven't seen that blog before, but we also have modal integration natively. When running any of the axolotl commands, you can prepend with --cloud config.yaml to run it there https://axolotl-ai-cloud.github.io/axolotl/docs/cli.html#remote-compute-with-modal-cloud — Reply to this email directly, view it on GitHub<#2459 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AXA4YJJKLH4DVJ2XXIMGMZ32XIBGLAVCNFSM6AAAAAB2FMNUOOVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTENRYGMZTOMY>. You are receiving this because you authored the thread.Message ID: ***@***.***>