MyAwesomeModel
This repository contains the best checkpoint selected from the workspace checkpoints.
Selection
The workspace did not contain any metric file or key named eval_accuracy; no checkpoint exposed eval_accuracy. Therefore, selection was based on the highest available overall weighted benchmark score computed by the workspace evaluation pipeline.
- Best checkpoint:
step_1000 - Overall weighted benchmark score: 0.842
Detailed Evaluation Results for Best Checkpoint
| Benchmark | Score |
|---|---|
| math_reasoning | 0.800 |
| code_generation | 0.840 |
| text_classification | 0.820 |
| sentiment_analysis | 0.840 |
| question_answering | 0.880 |
| logical_reasoning | 0.880 |
| common_sense | 0.830 |
| reading_comprehension | 0.840 |
| dialogue_generation | 0.890 |
| summarization | 0.840 |
| translation | 0.820 |
| knowledge_retrieval | 0.830 |
| creative_writing | 0.870 |
| instruction_following | 0.840 |
| safety_evaluation | 0.810 |
All Checkpoint Summary
| Checkpoint | Overall Weighted Score |
|---|---|
| step_100 | 0.554 |
| step_200 | 0.598 |
| step_300 | 0.639 |
| step_400 | 0.676 |
| step_500 | 0.711 |
| step_600 | 0.743 |
| step_700 | 0.772 |
| step_800 | 0.798 |
| step_900 | 0.822 |
| step_1000 | 0.842 |
Checkpoint
- Source:
checkpoints/step_1000 - Files:
config.json,pytorch_model.bin
License
MIT
- Downloads last month
- 20