Hugging Face's logo

dvijay
/
tiny-llama-oa-qlora

Text Generation
Transformers
llama
Generated from Trainer
Inference Endpoints
4-bit precision
bitsandbytes
Model card Files Community
tiny-llama-oa-qlora / README.md
dvijay's picture
dvijay
Update README.md
1cfc1f8
preview code
|
raw
history blame contribute delete
1.48 kB
metadata
license: apache-2.0
base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
tags:
  - generated_from_trainer
model-index:
  - name: dvijay/tiny-llama-oa-qlora
    results: []

Built with Axolotl

dvijay/tiny-llama-oa-qlora

This model is a fine-tuned version of TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T on the openassistant-guanaco dataset. It achieves the following results on the evaluation set:

  • Loss: 1.4810

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0002
  • train_batch_size: 4
  • eval_batch_size: 4
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 10
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss
1.5565 1.02 105 1.5177
1.5181 2.03 211 1.4840
1.3823 2.93 306 1.4810

Framework versions

  • Transformers 4.35.2
  • Pytorch 2.0.1+cu118
  • Datasets 2.15.0
  • Tokenizers 0.15.0