File size: 7,251 Bytes
551358b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
{
  "generated_at": "2026-08-16T06:07:50+00:00",
  "device": "mps",
  "geometry": {
    "geometry_index": 1,
    "window_frames": 12,
    "stride_frames": 10,
    "pad_left_frames": 0,
    "pad_right_frames": 2,
    "encoder_look_back": 1,
    "lookahead_ms": 120.0,
    "training_config": {
      "chunk_size": [
        8,
        12,
        16
      ],
      "stride": [
        6,
        10,
        14
      ],
      "pad_left": [
        0,
        0,
        0
      ],
      "encoder_att_look_back_factor": [
        1,
        1,
        1
      ]
    }
  },
  "scoring": {
    "cer": "corpus-level Levenshtein over NFKC-normalised characters",
    "punctuation": "stripped",
    "whitespace": "stripped",
    "rich_tags": "stripped",
    "wer_note": "whitespace tokens; near-meaningless for unsegmented Japanese and Chinese - read CER"
  },
  "models": {
    "base": {
      "dir": "/Users/nelmo-ux/Desktop/SenceVoice/models/SenseVoiceSmall",
      "checkpoint": null
    },
    "finetuned": {
      "dir": "/Users/nelmo-ux/Desktop/SenceVoice/models/SenseVoiceSmall",
      "checkpoint": null
    }
  },
  "japanese": {
    "note": "the selection axis: checkpoint choice keys off these numbers only",
    "val": {
      "jsonl": "/Users/nelmo-ux/Desktop/SenceVoice/data/vn/val.jsonl",
      "num_clips": 772,
      "metrics": {
        "per_model": {
          "base": {
            "chunk": {
              "cer": 0.4058795637743006,
              "mean_cer": 0.4907838660061337,
              "wer": 0.9991586032814472,
              "num_clips": 772,
              "ref_chars": 18981,
              "char_edits": 7704,
              "ref_words": 2377,
              "word_edits": 2375
            },
            "chunk_last_partial": {
              "cer": 0.43016700911437755,
              "mean_cer": 0.5131696714023274,
              "wer": 0.9974758098443416,
              "num_clips": 772,
              "ref_chars": 18981,
              "char_edits": 8165,
              "ref_words": 2377,
              "word_edits": 2371
            },
            "full": {
              "cer": 0.16253095200463621,
              "mean_cer": 0.24466634683089186,
              "wer": 0.9882204459402608,
              "num_clips": 772,
              "ref_chars": 18981,
              "char_edits": 3085,
              "ref_words": 2377,
              "word_edits": 2349
            },
            "chunk_minus_full_cer": 0.2433486117696644,
            "chunk_vs_full_cer": 0.3289047310434219
          }
        }
      }
    },
    "clips": [
      {
        "key": "ja.mp3",
        "path": "/Users/nelmo-ux/.cache/huggingface/hub/models--FunAudioLLM--SenseVoiceSmall/snapshots/3847d57b6bdf2dd8875cb1508d2af43d80a16bf7/example/ja.mp3",
        "language": "ja",
        "reference_source": "base model full-attention decode (no ground truth)",
        "per_model": {
          "base": {
            "full": "<|ja|><|NEUTRAL|><|Speech|><|woitn|>うちの中学は弁当制で持っていきない場合は50円の学校販売のパンを買う",
            "chunk": "ううちの中学は弁当等制で持っていききない場合は五0円の学校行販売ののパンを買う買う",
            "chunk_last_partial": "ううちの中学は弁当等制で持っていききない場合は五0円の学校行販売ののパンを買う買う",
            "chunk_vs_full_cer": 0.23529411764705882
          }
        },
        "drift_cer_vs_base_full": 0.0,
        "gap_delta": null
      }
    ]
  },
  "reference_only": {
    "note": "reference only, not a selection criterion: this finetune is Japanese-specialised and Chinese retention is not a maintained property",
    "clips": [
      {
        "key": "zh_sample.wav",
        "path": "/Users/nelmo-ux/Desktop/SenceVoice/runtime/llama.cpp/tests/sample.wav",
        "language": "auto",
        "reference_source": "base model full-attention decode (no ground truth)",
        "per_model": {
          "base": {
            "full": "<|zh|><|NEUTRAL|><|Speech|><|woitn|>我想问我在滨海新区有房",
            "chunk": "我想想问我在滨海心新区有有房",
            "chunk_last_partial": "我想想问我在滨海心新区有",
            "chunk_vs_full_cer": 0.2727272727272727
          }
        },
        "drift_cer_vs_base_full": 0.0,
        "gap_delta": null
      }
    ]
  },
  "selection": {
    "metric": "ja_val_chunk_cer",
    "value": 0.4058795637743006,
    "model": "base",
    "note": "corpus CER of the complete chunk-mode decode on the held-out Japanese val set; lower is better"
  },
  "examples": [
    {
      "key": "GIGA_Ai_Kiss_2__うめき声__a1040039",
      "scope": "japanese",
      "reference": "――うぅぉぁあ……",
      "base_full": "<|nospeech|><|EMO_UNKNOWN|><|Event_UNK|><|woitn|>",
      "base_chunk": "うんうん",
      "base_chunk_last_partial": "うんうん"
    },
    {
      "key": "GIGA_Ai_Kiss_2__アヤメ・ヒナタ__06900000",
      "scope": "japanese",
      "reference": "蓮くん、デートしよ♪",
      "base_full": "<|ja|><|EMO_UNKNOWN|><|Speech|><|woitn|>レくんデートしよう",
      "base_chunk": "レんくくんでデートしよよ",
      "base_chunk_last_partial": "レんくくんで"
    },
    {
      "key": "GIGA_Ai_Kiss_2__クラスメイトの男子__a1050000",
      "scope": "japanese",
      "reference": "――みんな気になるのさ。あんな可愛い双子の子に挟まれてるあいつは何者なのか、ってね",
      "base_full": "<|ja|><|ANGRY|><|Speech|><|woitn|>みんな気になるのさあんな可愛い双子の子に挟まれてるあいつは何者なのかってね",
      "base_chunk": "みんなき気になるのさあんな顔かい子の子に挟まれてるあいつつは何何者なのかってねてね",
      "base_chunk_last_partial": "みんなき気になるのさあんな顔かい子の子に挟まれてるあいつつは何何者なのかってね"
    },
    {
      "key": "GIGA_Ai_Kiss_2__マネージャー__01900000",
      "scope": "japanese",
      "reference": "分かりました、気をつけて下さい。委員のキミもごくろう、くれぐれもよろしく頼むよ",
      "base_full": "<|ja|><|ANGRY|><|Speech|><|woitn|>わかりました気をつけてください委員の君もご苦労くれぐれもよろしく頼むよ",
      "base_chunk": "わかりましした気をつけてくださいい委員の君もご苦労くれぐれれもよろしく頼むよ",
      "base_chunk_last_partial": "わかりましした気をつけてくださいい委員の君もご苦労くれぐれれもよろしく頼"
    },
    {
      "key": "GIGA_Ai_Kiss_2__事故にあった女子__02030027",
      "scope": "japanese",
      "reference": "――ねえ、私の腕。見つからないの、どこにあるか知らない?",
      "base_full": "<|ja|><|EMO_UNKNOWN|><|Speech|><|woitn|>ねえ私の腕見つからないのどこにあるか知らない",
      "base_chunk": "ねえね私の腕見つからないのどこにあるか知らなない",
      "base_chunk_last_partial": "ねえね私の腕見つからないのどこにあるか知らな"
    }
  ],
  "warnings": []
}