llama-finetuned / README.md
ipryzk's picture
Update README.md
3b18e3c verified
metadata
language:
  - en
metrics:
  - loss: 0.6
base_model:
  - meta-llama/Llama-3.1-8B-Instruct
data:
  - >-
    stored within this directory (respectively training_set and validation_set
    for training and validation)
notes:
  - Howdy! This is Ian Prazak's fine-tuned LLaMa-8B Instruct model.
  - >-
    The model was trained over just one epoch (work in progress) with a dataset
    of approximately 30,000 formatted examples in JSONL.
  - The base LLaMa tokenizer was used.
  - The current model is llama-fact-full-test.nemo
  - >-
    If the model only runs in its base state, please load it with the
    llama-fact.ckpt to incorporate the updated weights.
  - The metrics listed above represent the current validation loss and accuracy
  - Thank you for checking it out! :)