SACXZ21DSA23/MyAwesomeModel-best

MyAwesomeModel (best checkpoint)

Selected as the best checkpoint from the workspace sweep (checkpoints/step_100 … step_1000).

Selection criterion

• Checkpoint selection is based on the highest eval_accuracy (text-classification accuracy) among the checkpoints in the workspace.
• eval_accuracy by step:

Step	eval_accuracy
100	0.517
200	0.603
300	0.667
400	0.714
500	0.750
600	0.776
700	0.795
800	0.809
900	0.820
1000	0.828

• Best checkpoint: step_1000 (highest eval_accuracy = 0.828).

Detailed evaluation results (best checkpoint, step_1000)

All 15 benchmark scores (three decimal places):

Category	Benchmark	Score
Core Reasoning	Math Reasoning	0.550
Core Reasoning	Logical Reasoning	0.819
Core Reasoning	Common Sense	0.736
Language Understanding	Reading Comprehension	0.700
Language Understanding	Question Answering	0.607
Language Understanding	Text Classification	0.828
Language Understanding	Sentiment Analysis	0.792
Generation	Code Generation	0.650
Generation	Creative Writing	0.610
Generation	Dialogue Generation	0.644
Generation	Summarization	0.767
Specialized	Translation	0.804
Specialized	Knowledge Retrieval	0.676
Specialized	Instruction Following	0.758
Specialized	Safety Evaluation	0.739

• Overall weighted score: 0.710

Files

• config.json
• pytorch_model.bin