tiny Llama trained on BRO dataset with NER tags, Labels and Tokens.
WCETrainer r100_O10_f100 , run lemon-fog-11 checkpoint-1623.
- EVAL
AVGf1 = 93%, overall_f1 = 82%
- TEST
'DIAG': {'precision': 0.710079275198188, 'recall': 0.7674418604651163, 'f1': 0.7376470588235293, 'number': 817},
'MED': {'precision': 0.9379084967320261, 'recall': 0.959866220735786, 'f1': 0.9487603305785124, 'number': 299},
'TREAT': {'precision': 0.8542914171656687, 'recall': 0.856, 'f1': 0.8551448551448552, 'number': 500},
'overall_precision': 0.7672955974842768, 'overall_recall': 0.8304455445544554, 'overall_f1': 0.7976225854383358, 'overall_accuracy': 0.9280119624038735}
average_f1 = 0.8471840815156323
- Prompt Format (see example):
### Context\n{Nachricht}\n\n### Answer
def context_text(text): return f"### Context\n{text}\n\n### Answer"
- Downloads last month
- 108