modal-labs / llm-finetuning

Guide for fine-tuning Llama/Mistral/CodeLlama models and more
MIT License
537 stars 83 forks source link

DeepSpeed documentation corrections #63

Open lhl opened 5 months ago

lhl commented 5 months ago

The link to the configs is wrong so also fixed, but more importantly the config path is wrong (points to /root not /workspace) - this causes runs to die (and is actually a bit of a PITA)

Note the docs for the same section here https://modal.com/docs/examples/llm-finetuning is uniquely wrong so I guess not synced, but add to the confusion