parakeet.js 0.0.2 → 0.0.3

package/docs/parakeet-transformers-js/js_step_by_step_v3.json

@@ -1,450 +0,0 @@
-{
-  "audio_length": 176017,
-  "sample_rate": 16000,
-  "model_config": {
-    "vocab_size": 1025,
-    "blank_idx": 1024,
-    "max_tokens_per_step": 10,
-    "subsampling_factor": 8,
-    "pred_layers": 2,
-    "pred_hidden": 640
-  },
-  "steps": [
-    {
-      "step": 0,
-      "t": 0,
-      "emitted_tokens": 0,
-      "probs_shape": [
-        1025
-      ],
-      "token": 1024,
-      "token_text": "<blk>",
-      "is_blank": true,
-      "step_prediction": 1,
-      "top_5_tokens": [
-        {
-          "token": 1024,
-          "text": "<blk>",
-          "prob": 41.226837
-        },
-        {
-          "token": 34,
-          "text": "▁I",
-          "prob": 33.77882
-        },
-        {
-          "token": 248,
-          "text": "▁We",
-          "prob": 33.430401
-        },
-        {
-          "token": 139,
-          "text": "▁So",
-          "prob": 33.349823
-        },
-        {
-          "token": 108,
-          "text": "▁W",
-          "prob": 32.639244
-        }
-      ],
-      "top_5_probs": [],
-      "action": "blank",
-      "advance": 1,
-      "advance_reason": "duration_step"
-    },
-    {
-      "step": 1,
-      "t": 1,
-      "emitted_tokens": 0,
-      "probs_shape": [
-        1025
-      ],
-      "token": 1024,
-      "token_text": "<blk>",
-      "is_blank": true,
-      "step_prediction": 4,
-      "top_5_tokens": [
-        {
-          "token": 1024,
-          "text": "<blk>",
-          "prob": 42.59375
-        },
-        {
-          "token": 123,
-          "text": "▁And",
-          "prob": 35.228672
-        },
-        {
-          "token": 34,
-          "text": "▁I",
-          "prob": 34.526535
-        },
-        {
-          "token": 155,
-          "text": "▁The",
-          "prob": 33.929031
-        },
-        {
-          "token": 248,
-          "text": "▁We",
-          "prob": 33.60059
-        }
-      ],
-      "top_5_probs": [],
-      "action": "blank",
-      "advance": 4,
-      "advance_reason": "duration_step"
-    },
-    {
-      "step": 2,
-      "t": 5,
-      "emitted_tokens": 0,
-      "probs_shape": [
-        1025
-      ],
-      "token": 1024,
-      "token_text": "<blk>",
-      "is_blank": true,
-      "step_prediction": 4,
-      "top_5_tokens": [
-        {
-          "token": 1024,
-          "text": "<blk>",
-          "prob": 41.437325
-        },
-        {
-          "token": 34,
-          "text": "▁I",
-          "prob": 35.343742
-        },
-        {
-          "token": 123,
-          "text": "▁And",
-          "prob": 34.755867
-        },
-        {
-          "token": 155,
-          "text": "▁The",
-          "prob": 33.781651
-        },
-        {
-          "token": 157,
-          "text": "▁H",
-          "prob": 33.757645
-        }
-      ],
-      "top_5_probs": [],
-      "action": "blank",
-      "advance": 4,
-      "advance_reason": "duration_step"
-    },
-    {
-      "step": 3,
-      "t": 9,
-      "emitted_tokens": 0,
-      "probs_shape": [
-        1025
-      ],
-      "token": 1024,
-      "token_text": "<blk>",
-      "is_blank": true,
-      "step_prediction": 1,
-      "top_5_tokens": [
-        {
-          "token": 1024,
-          "text": "<blk>",
-          "prob": 40.797409
-        },
-        {
-          "token": 34,
-          "text": "▁I",
-          "prob": 35.102684
-        },
-        {
-          "token": 123,
-          "text": "▁And",
-          "prob": 34.858852
-        },
-        {
-          "token": 574,
-          "text": "▁Oh",
-          "prob": 34.581413
-        },
-        {
-          "token": 261,
-          "text": "▁You",
-          "prob": 34.249733
-        }
-      ],
-      "top_5_probs": [],
-      "action": "blank",
-      "advance": 1,
-      "advance_reason": "duration_step"
-    },
-    {
-      "step": 4,
-      "t": 10,
-      "emitted_tokens": 0,
-      "probs_shape": [
-        1025
-      ],
-      "token": 1024,
-      "token_text": "<blk>",
-      "is_blank": true,
-      "step_prediction": 1,
-      "top_5_tokens": [
-        {
-          "token": 1024,
-          "text": "<blk>",
-          "prob": 39.746513
-        },
-        {
-          "token": 34,
-          "text": "▁I",
-          "prob": 35.508808
-        },
-        {
-          "token": 155,
-          "text": "▁The",
-          "prob": 34.634434
-        },
-        {
-          "token": 163,
-          "text": "▁O",
-          "prob": 34.146984
-        },
-        {
-          "token": 819,
-          "text": "▁",
-          "prob": 34.010185
-        }
-      ],
-      "top_5_probs": [],
-      "action": "blank",
-      "advance": 1,
-      "advance_reason": "duration_step"
-    },
-    {
-      "step": 5,
-      "t": 11,
-      "emitted_tokens": 0,
-      "probs_shape": [
-        1025
-      ],
-      "token": 1024,
-      "token_text": "<blk>",
-      "is_blank": true,
-      "step_prediction": 4,
-      "top_5_tokens": [
-        {
-          "token": 1024,
-          "text": "<blk>",
-          "prob": 41.378902
-        },
-        {
-          "token": 34,
-          "text": "▁I",
-          "prob": 32.861183
-        },
-        {
-          "token": 123,
-          "text": "▁And",
-          "prob": 32.594582
-        },
-        {
-          "token": 155,
-          "text": "▁The",
-          "prob": 32.279743
-        },
-        {
-          "token": 819,
-          "text": "▁",
-          "prob": 32.114971
-        }
-      ],
-      "top_5_probs": [],
-      "action": "blank",
-      "advance": 4,
-      "advance_reason": "duration_step"
-    },
-    {
-      "step": 6,
-      "t": 15,
-      "emitted_tokens": 0,
-      "probs_shape": [
-        1025
-      ],
-      "token": 1024,
-      "token_text": "<blk>",
-      "is_blank": true,
-      "step_prediction": 1,
-      "top_5_tokens": [
-        {
-          "token": 1024,
-          "text": "<blk>",
-          "prob": 48.968941
-        },
-        {
-          "token": 34,
-          "text": "▁I",
-          "prob": 43.725601
-        },
-        {
-          "token": 74,
-          "text": "▁Th",
-          "prob": 41.064816
-        },
-        {
-          "token": 219,
-          "text": "▁G",
-          "prob": 41.000469
-        },
-        {
-          "token": 260,
-          "text": "▁E",
-          "prob": 40.559464
-        }
-      ],
-      "top_5_probs": [],
-      "action": "blank",
-      "advance": 1,
-      "advance_reason": "duration_step"
-    }
-  ],
-  "preprocessing": {
-    "features_shape": [
-      1,
-      1101,
-      128
-    ],
-    "features_lens": [
-      1101
-    ],
-    "features_sample": [
-      6.473908,
-      3.58362,
-      -2.028411,
-      -2.028411,
-      -2.028409,
-      -2.028214,
-      -1.430135,
-      -0.635826,
-      -1.805648,
-      -1.180914,
-      0.139405,
-      0.423509,
-      -0.254901,
-      -1.161156,
-      0.343323,
-      -0.877018,
-      -1.558379,
-      -0.60808,
-      -1.13381,
-      -1.102631,
-      0.779868,
-      0.108186,
-      -1.602304,
-      -0.413008,
-      0.014313,
-      0.368889,
-      -0.184027,
-      -1.061025,
-      -1.77546,
-      -0.943931,
-      -0.283598,
-      -1.202594,
-      -1.094526,
-      0.152471,
-      0.48408,
-      -0.438624,
-      -0.192583,
-      -1.456899,
-      -1.78464,
-      -1.120869,
-      -1.801295,
-      -1.852787,
-      -1.885587,
-      -1.48906,
-      -1.306157,
-      0.219165,
-      0.023559,
-      -0.32646,
-      -0.66996,
-      -1.473523
-    ],
-    "features_dtype": "float32"
-  },
-  "encoding": {
-    "encoder_out_shape": [
-      1,
-      1024,
-      16
-    ],
-    "encoder_out_lens": [
-      16
-    ],
-    "encoder_out_sample": [
-      0.007837,
-      0.003748,
-      -0.05033,
-      -0.034014,
-      -0.022349,
-      -0.017078,
-      -0.019876,
-      -0.016203,
-      -0.025393,
-      -0.017256,
-      -0.012161,
-      -0.008812,
-      -0.066451,
-      -0.089987,
-      -0.047875,
-      -0.176737,
-      -0.027524,
-      -0.042996,
-      -0.059277,
-      -0.056243,
-      -0.033654,
-      -0.024435,
-      -0.029944,
-      -0.036262,
-      -0.026254,
-      -0.038526,
-      -0.050476,
-      -0.067926,
-      -0.054781,
-      -0.094117,
-      -0.106241,
-      -0.050649,
-      0.057139,
-      0.009718,
-      0.009076,
-      -0.003998,
-      -0.007633,
-      -0.003716,
-      0.005142,
-      0.010054,
-      -0.001808,
-      0.015909,
-      0.026671,
-      0.036184,
-      0.033018,
-      0.03269,
-      0.044188,
-      0.083504,
-      0.009676,
-      0.009832
-    ],
-    "encoder_out_dtype": "float32"
-  },
-  "final_results": {
-    "tokens": [],
-    "timestamps": [],
-    "decoded_text": "",
-    "total_steps": 7,
-    "final_t": 16,
-    "encodings_len": 16
-  },
-  "full_transcription": "<unk>"
-}