Edit model card

smol_llama-220M-bees-internal

This model is a fine-tuned version of BEE-spoke-data/smol_llama-220M-GQA on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 2.6892
  • Accuracy: 0.4610

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: 4
  • eval_batch_size: 2
  • seed: 27634
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_ratio: 0.05
  • num_epochs: 2.0

Training results

Training Loss Epoch Step Validation Loss Accuracy
3.0959 0.1 50 2.9671 0.4245
2.9975 0.19 100 2.8691 0.4371
2.8938 0.29 150 2.8271 0.4419
2.9027 0.39 200 2.7973 0.4457
2.8983 0.49 250 2.7719 0.4489
2.8789 0.58 300 2.7519 0.4515
2.8672 0.68 350 2.7366 0.4535
2.8369 0.78 400 2.7230 0.4558
2.8271 0.88 450 2.7118 0.4569
2.7775 0.97 500 2.7034 0.4587
2.671 1.07 550 2.6996 0.4592
2.695 1.17 600 2.6965 0.4598
2.6962 1.27 650 2.6934 0.4601
2.6034 1.36 700 2.6916 0.4605
2.716 1.46 750 2.6901 0.4609
2.6968 1.56 800 2.6896 0.4608
2.6626 1.66 850 2.6893 0.4609
2.6881 1.75 900 2.6891 0.4610
2.7339 1.85 950 2.6891 0.4610
2.6729 1.95 1000 2.6892 0.4610

Framework versions

  • Transformers 4.36.2
  • Pytorch 2.1.0
  • Datasets 2.16.1
  • Tokenizers 0.15.0
Downloads last month
5
Safetensors
Model size
218M params
Tensor type
BF16
·

Finetuned from

Dataset used to train BEE-spoke-data/smol_llama-220M-bees-internal

Collections including BEE-spoke-data/smol_llama-220M-bees-internal