README.md

metadata

tags:
  - generated_from_trainer
datasets:
  - /pfs/lustrep4/scratch/project_462000259/noah/instruct-datasets/askscience
metrics:
  - accuracy
model-index:
  - name: layer_13,14,15
    results:
      - task:
          name: Causal Language Modeling
          type: text-generation
        dataset:
          name: >-
            /pfs/lustrep4/scratch/project_462000259/noah/instruct-datasets/askscience
          type: >-
            /pfs/lustrep4/scratch/project_462000259/noah/instruct-datasets/askscience
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.27968436193888074

layer_13,14,15

This model is a fine-tuned version of /pfs/lustrep4/scratch/project_462000259/noah/instruct_1bil/transfer/pythia-deduped-1b-chat-base/ on the /pfs/lustrep4/scratch/project_462000259/noah/instruct-datasets/askscience dataset. It achieves the following results on the evaluation set:

• Loss: 5.4570
• Accuracy: 0.2797

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: 24
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 192
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 6000

Training results

Framework versions

• Transformers 4.27.0
• Pytorch 1.12.1+gitcb6c422
• Datasets 2.11.0
• Tokenizers 0.13.3