Safetensors
ParagonLight's picture
upload loras
0a8e533
metadata
license: other
library_name: peft
tags:
  - llama-factory
  - lora
  - generated_from_trainer
base_model: /data1/model/llama2/meta-llama/Llama2-13b
model-index:
  - name: cnn_dailymail_no_sys
    results: []

cnn_dailymail_no_sys

This model is a fine-tuned version of /data1/model/llama2/meta-llama/Llama2-13b on the cnn_dailymail_no_sys dataset. It achieves the following results on the evaluation set:

  • Loss: 0.9911

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0001
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 3
  • total_train_batch_size: 24
  • total_eval_batch_size: 24
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 20
  • num_epochs: 10.0

Training results

Training Loss Epoch Step Validation Loss
1.0297 0.14 50 1.0376
1.0043 0.28 100 1.0226
1.0241 0.42 150 1.0131
1.0134 0.56 200 1.0083
1.0284 0.7 250 1.0041
1.0126 0.85 300 0.9974
0.9779 0.99 350 0.9922
0.9239 1.13 400 0.9984
0.9456 1.27 450 0.9971
0.9064 1.41 500 0.9911
0.9238 1.55 550 0.9884
0.9053 1.69 600 0.9863
0.9066 1.83 650 0.9829
0.932 1.97 700 0.9802
0.7339 2.11 750 1.0370
0.7678 2.25 800 1.0321
0.7538 2.39 850 1.0240

Framework versions

  • PEFT 0.9.0
  • Transformers 4.38.2
  • Pytorch 2.2.1
  • Datasets 2.18.0
  • Tokenizers 0.15.2