|
--- |
|
datasets: |
|
- datacrunch/freformatted |
|
--- |
|
|
|
## Whats this merge about |
|
|
|
- Its a 500 step checkpoint of the following run |
|
``` |
|
python ./trl/examples/scripts/sft_trainer.py --model_name meta-llama/Llama-2-7b-hf --dataset_name datacrunch/finnish_alpaca --load_in_4bit --use_peft --batch_size 4 --gradient_accumulation_steps 2 |
|
``` |
|
- Using the repo https://github.com/lvwerra/trl/blob/main/examples/scripts/sft_trainer.py |
|
|
|
I am still figuring out an efficient way of doing this, in the meantime you can test it |
|
|
|
- An example prompt you can try, that should return the Finnish response you need |
|
``` |
|
"Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request. ### Instruction: Anna kolme vinkkiä terveenä pysymiseen. ###Response:" |
|
``` |
