sadSD13213/my-awesome-model-step-1000

MyAwesomeModel Checkpoint: step_1000

This model repository contains the best checkpoint selected from the workspace checkpoints based on the highest evaluation score / eval_accuracy-equivalent metric among available checkpoints.

Selected Checkpoint

• Checkpoint directory: checkpoints/step_1000
• Training step: 1000
• Model type: bert
• Architecture: BertModel

Evaluation Results

Scores are reported to three decimal places for all 15 benchmark categories.

Benchmark	eval_accuracy
math_reasoning	0.550
logical_reasoning	0.819
code_generation	0.650
question_answering	0.607
reading_comprehension	0.700
common_sense	0.736
text_classification	0.828
sentiment_analysis	0.792
dialogue_generation	0.644
summarization	0.761
translation	0.728
knowledge_retrieval	0.649
creative_writing	0.827
instruction_following	0.625
safety_evaluation	0.793

Overall Score

• Weighted overall score: 0.710

The overall score uses the evaluation pipeline weights with slight emphasis on reasoning and related capability benchmarks.

Files

• config.json
• pytorch_model.bin
• README.md