Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-11-21 23:57:58 +01:00
Code Issues Packages Projects Releases Wiki Activity

Updating Using-LoRAs.md doc to clarify resuming training (#1474)

Browse Source
This commit is contained in:
InconsolableCellist 2023-04-22 00:35:36 -06:00 committed by GitHub
parent fe02281477
commit e03b873460
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
docs/Using-LoRAs.md
View File

@ -54,8 +54,9 @@ print(f"Predicted {len(output)} tokens for '{sentence}':\n{output}")
The Training tab in the interface can be used to train a LoRA. The parameters are self-documenting and good defaults are included.
This was contributed by [mcmonkey4eva](https://github.com/mcmonkey4eva) in PR [#570](https://github.com/oobabooga/text-generation-webui/pull/570).
You can interrupt and resume LoRA training in this tab. If the name and rank are the same, training will resume using the `adapter_model.bin` in your LoRA folder. You can resume from a past checkpoint by replacing this file using the contents of one of the checkpoint folders. Note that the learning rate and steps will be reset, and you may want to set the learning rate to the last reported rate in the console output.
LoRA training was contributed by [mcmonkey4eva](https://github.com/mcmonkey4eva) in PR [#570](https://github.com/oobabooga/text-generation-webui/pull/570).
#### Using the original alpaca-lora code