metadata
license: mit
library_name: transformers
pipeline_tag: feature-extraction
tags:
- bert
- checkpoint-selection
model-index:
- name: MyAwesomeModel
results:
- task:
type: feature-extraction
dataset:
name: Workspace evaluation suite
type: custom
metrics:
- name: eval_accuracy
type: accuracy
value: 0.71
MyAwesomeModel
This repository contains the best checkpoint found in the training workspace.
It was selected by comparing eval_accuracy across every discovered checkpoint.
- Selected checkpoint:
step_1000 - eval_accuracy: 0.710
- Architecture declared by the checkpoint:
BertModel
Checkpoint comparison
| Checkpoint | eval_accuracy |
|---|---|
| step_100 | 0.480 |
| step_200 | 0.535 |
| step_300 | 0.576 |
| step_400 | 0.608 |
| step_500 | 0.635 |
| step_600 | 0.656 |
| step_700 | 0.674 |
| step_800 | 0.689 |
| step_900 | 0.700 |
| step_1000 | 0.710 |
Selected-checkpoint benchmark results
| Benchmark | Score |
|---|---|
| math_reasoning | 0.550 |
| logical_reasoning | 0.819 |
| common_sense | 0.736 |
| reading_comprehension | 0.700 |
| question_answering | 0.607 |
| text_classification | 0.828 |
| sentiment_analysis | 0.792 |
| code_generation | 0.650 |
| creative_writing | 0.610 |
| dialogue_generation | 0.644 |
| summarization | 0.767 |
| translation | 0.804 |
| knowledge_retrieval | 0.676 |
| instruction_following | 0.758 |
| safety_evaluation | 0.739 |
Files
The selected checkpoint's config.json and pytorch_model.bin are stored at the repository root for compatibility with transformers loading conventions.
Note: only checkpoint artifacts present in the workspace were published. No tokenizer files were available in the checkpoint directory.