MyAwesomeModel

MyAwesomeModel is a BERT-style model fine-tuned across many training steps. The best checkpoint (step 1000) achieves a weighted overall score of 0.71 across all 15 evaluation benchmarks.

Evaluation Results

Detailed benchmark scores for the best checkpoint (step_1000):

# Benchmark Score
1 Math Reasoning 0.550
2 Logical Reasoning 0.819
3 Code Generation 0.650
4 Question Answering 0.607
5 Reading Comprehension 0.700
6 Common Sense 0.736
7 Text Classification 0.828
8 Sentiment Analysis 0.792
9 Dialogue Generation 0.644
10 Summarization 0.767
11 Translation 0.804
12 Knowledge Retrieval 0.676
13 Creative Writing 0.610
14 Instruction Following 0.758
15 Safety Evaluation 0.739

Weighted Overall Score: 0.710

Model Details

  • Model type: BERT (BertModel)
  • Hidden size: 768
  • Layers: 12
  • Attention heads: 12
  • Vocab size: 30522
  • Total parameters: 109,482,240

License

MIT

Downloads last month
47
Safetensors
Model size
0.1B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support