vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Updating two small main references missed earlier in the finetune docs.

Browse source
This commit is contained in:
HanClinto 2024-06-10 16:12:50 -07:00
parent 82df7f9f0e
commit 1f5ec2c0b4
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
examples/finetune/README.md
View file

@ -38,9 +38,9 @@ After 10 more iterations:
Checkpoint files (`--checkpoint-in FN`, `--checkpoint-out FN`) store the training process. When the input checkpoint file does not exist, it will begin finetuning a new randomly initialized adapter. Checkpoint files (`--checkpoint-in FN`, `--checkpoint-out FN`) store the training process. When the input checkpoint file does not exist, it will begin finetuning a new randomly initialized adapter.
llama.cpp compatible LORA adapters will be saved with filename specified by `--lora-out FN`. llama.cpp compatible LORA adapters will be saved with filename specified by `--lora-out FN`.
These LORA adapters can then be used by `main` together with the base model, like in the 'predict' example command above. These LORA adapters can then be used by `llama-cli` together with the base model, like in the 'predict' example command above.
In `main` you can also load multiple LORA adapters, which will then be mixed together. In `llama-cli` you can also load multiple LORA adapters, which will then be mixed together.
For example if you have two LORA adapters `lora-open-llama-3b-v2-q8_0-shakespeare-LATEST.bin` and `lora-open-llama-3b-v2-q8_0-bible-LATEST.bin`, you can mix them together like this: For example if you have two LORA adapters `lora-open-llama-3b-v2-q8_0-shakespeare-LATEST.bin` and `lora-open-llama-3b-v2-q8_0-bible-LATEST.bin`, you can mix them together like this: