octen-embedding-0.6b-onnx-int4 / conversion-metadata.json
geoffsee's picture
Upload ONNX int4 and tokenizer
18c1c2c verified
raw
history blame contribute delete
432 Bytes
{
"model_id": "Octen/Octen-Embedding-0.6B",
"opset": 18,
"max_seq_length": 512,
"export_mode": "direct_fp16",
"export_batch_size": 4,
"dynamic_batch": true,
"device": "mps",
"pooling": "last_token",
"normalize": "l2",
"padding_side": "left",
"weights_dtype": "float16",
"output_dtype": "float32",
"hidden_size": 1024,
"files": {
"onnx_fp16": "model.fp16.onnx",
"tokenizer_dir": "tokenizer"
}
}