vnm_intent_detection / model_metadata.json
Tnt3o5's picture
chore: upload quantized ONNX model from Kaggle
918c154 verified
{
"model_name": "Vietnamese-Sentiment-visobert fine-tuned",
"max_length": 512,
"requires_token_type_ids": true,
"input_names": [
"input_ids",
"attention_mask",
"token_type_ids"
],
"num_intents": 5,
"intents": [
"definition",
"execute_sql",
"greeting_chitchat",
"refine_query",
"support_previous_answer"
],
"fp32_size_mb": 372.45,
"int8_size_mb": 93.85,
"compression_ratio": 74.8,
"latency_fp32_ms": 147.51,
"latency_int8_ms": 51.28,
"speedup": 2.88
}