Blocksmith
Training Procedure
The T5-small model was fine-tuned on the Minecraft log dataset and a text summarising dataset (Xsum) using the Adam optimizer with a learning rate of 2e-05 for 1 epoch. Early stopping was not implemented.
Model description
Blocksmith is a natural language processing model designed to generate concise summaries of Minecraft logs. It is based on the Transformer architecture, specifically the T5-small model, and trained on a dataset of Minecraft logs.
Intended uses & limitations
Blocksmith is intended for analyzing player behavior, identifying potential issues or bugs, and generating insights for game improvement. However, the model may have limitations in handling specific log formats or game versions, and its summaries might be biased towards the content of the training data.
Training procedure
The T5-small model was fine-tuned on the Minecraft log dataset and a text summarising dataset (Xsum) using the Adam optimizer with a learning rate of 2e-05 for 1 epoch. Early stopping was not implemented.
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 1
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
|---|---|---|---|---|---|---|---|---|
| No log | 1.0 | 11 | 2.8271 | 34.8098 | 17.0245 | 32.5651 | 32.2774 | 14.8182 |
Framework versions
- Transformers 4.42.4
- Pytorch 2.3.1+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1