Instructions to use Xenova/whisper-medium with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers.js
How to use Xenova/whisper-medium with Transformers.js:
// npm i @huggingface/transformers import { pipeline } from '@huggingface/transformers'; // Allocate pipeline const pipe = await pipeline('automatic-speech-recognition', 'Xenova/whisper-medium');
File size: 2,840 Bytes
180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a 5c0869e 180390a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 | {
"per_channel": false,
"reduce_range": false,
"per_model_config": {
"decoder_model": {
"op_types": [
"Div",
"Concat",
"Gather",
"Less",
"ReduceMean",
"MatMul",
"Erf",
"Add",
"Expand",
"Sub",
"ConstantOfShape",
"Range",
"Sqrt",
"Slice",
"Squeeze",
"Cast",
"Equal",
"Where",
"Constant",
"Pow",
"Softmax",
"Shape",
"Reshape",
"Mul",
"Transpose",
"Unsqueeze"
],
"weight_type": "QInt8"
},
"decoder_model_merged": {
"op_types": [
"Div",
"Gather",
"Concat",
"Less",
"ReduceMean",
"MatMul",
"Erf",
"Add",
"Expand",
"Sub",
"ConstantOfShape",
"Range",
"Sqrt",
"Slice",
"Squeeze",
"Cast",
"Equal",
"Where",
"If",
"Constant",
"Pow",
"Softmax",
"Shape",
"Reshape",
"Mul",
"Transpose",
"Unsqueeze"
],
"weight_type": "QInt8"
},
"decoder_with_past_model": {
"op_types": [
"Div",
"Concat",
"Gather",
"Constant",
"Pow",
"Sub",
"Mul",
"Transpose",
"Add",
"MatMul",
"Softmax",
"Sqrt",
"Slice",
"Shape",
"Reshape",
"ReduceMean",
"Erf",
"Unsqueeze"
],
"weight_type": "QInt8"
},
"encoder_model": {
"op_types": [
"Div",
"Gather",
"Concat",
"Constant",
"Pow",
"Sub",
"Unsqueeze",
"Mul",
"Conv",
"Transpose",
"MatMul",
"Softmax",
"Sqrt",
"Shape",
"Reshape",
"ReduceMean",
"Erf",
"Add"
],
"weight_type": "QUInt8"
}
}
} |