togetherchat-dev-7b-v2

This model is a fine-tuned version of togethercomputer/LLaMA-2-7B-32K on 25000 entries for 3 epochs.

Model description

Model can be used for text-to-code generation and for further fine-tuning, Colab notebook example (on free T4 GPU) soon!

Datasets used:

• evol-codealpaca-80k - 10000 entries
• codealpaca-20k - 10000 entries
• open-platypus - 5000 entries

Intended uses & limitations

Please remember that model may (and will) produce inaccurate informations, you need to fine-tune it for your specific task.

Training and evaluation data

See 'Metrics'

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 10
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 40
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3

Training results

Framework versions

• Transformers 4.33.1
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.13.3