Hugging Face's logo

Voltline
/
vimeml-tiny-zh-v1

Text Generation
Core ML
PyTorch
Chinese
chinese
pinyin
input-method
causal-lm
contextual-reranking
tiny-gpt
rmsnorm
swiglu
Model card Files
xet
 Community
vimeml-tiny-zh-v1
File size: 4,767 Bytes
81cffe7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
{
  "format": "vimeml_public_evaluation_zh_v1",
  "model_version": "zh-v1-step844730",
  "checkpoint_sha256": "44c423a1e7cb0e32161afc94d6167cabb439a3d25fd9de6d1fbb150a927542d4",
  "training": {
    "arm": "12k-fresh-s1",
    "epochs": 1,
    "updates": 844730,
    "seed": 1591996171,
    "supervised_targets": 14519248976,
    "web_characters": 21280000000,
    "chat_share_of_supervised_tokens": 0.05,
    "validation_bpc_fp32": {
      "web": 4.354450029201291,
      "chat": 5.690730282652201
    },
    "selection": "final checkpoint of the release arm, chosen by the preregistered eight-arm rule before blind use",
    "blind_used_for_selection": false
  },
  "ime_standard_zh_v1_r1": {
    "labels": "AI-reviewed, not native-speaker gold",
    "development": {
      "full": {
        "items": 1456,
        "top1": 1363
      },
      "abbrev": {
        "items": 990,
        "top1": 337
      }
    },
    "blind_confirmation": {
      "full": {
        "items": 1445,
        "top1": 1355,
        "top1_rate": 0.938,
        "top3_rate": 0.972,
        "mrr": 0.954
      },
      "abbrev": {
        "items": 964,
        "top1_rate": 0.35
      },
      "rime_ice_first_candidate_top1_rate": 0.723,
      "vs_rime": "+329/-19, McNemar p=3.6e-74",
      "passed": true
    },
    "baselines_blind_full_top1_rate": {
      "vimeml-tiny-zh-v1": 0.938,
      "Qwen2.5-0.5B": 0.918,
      "Qwen3-0.6B-Base": 0.892,
      "rime-ice + wanxiang grammar": 0.822,
      "rime-ice": 0.722,
      "luna_pinyin_simp": 0.689
    }
  },
  "coreml": {
    "development_rankings": {
      "cuda_fp32": {
        "full": {
          "items": 1456,
          "top1": 1363
        },
        "abbrev": {
          "items": 990,
          "top1": 337
        }
      },
      "coreml_fp32": {
        "full": {
          "items": 1456,
          "top1": 1363,
          "top3": 1416
        },
        "abbrev": {
          "items": 990,
          "top1": 337,
          "top3": 389
        }
      },
      "coreml_int8": {
        "full": {
          "items": 1456,
          "top1": 1364,
          "top3": 1416
        },
        "abbrev": {
          "items": 990,
          "top1": 335,
          "top3": 389
        }
      },
      "kvchunk_fp32": {
        "full": {
          "items": 1456,
          "top1": 1363,
          "top3": 1416
        },
        "abbrev": {
          "items": 990,
          "top1": 337,
          "top3": 389
        }
      },
      "kvchunk_int8": {
        "full": {
          "items": 1456,
          "top1": 1364,
          "top3": 1416
        },
        "abbrev": {
          "items": 990,
          "top1": 335,
          "top3": 389
        }
      }
    },
    "int8_vs_fp32_top1_changes": {
      "full": "+2/-1",
      "abbrev": "+0/-2"
    },
    "fp32_alignment_max_abs": {
      "stateless": 3.91e-05,
      "kvchunk": 4.63e-05
    },
    "int8_strict_alignment": {
      "passed": false,
      "max_abs": 0.392,
      "note": "expected lossy quantization drift"
    },
    "kvchunk_int8_vs_stateless_int8_max_abs": 4.15e-05,
    "package_bytes": {
      "fp32": 80367198,
      "int8": 22792958,
      "kvchunk_int8": 22830199
    }
  },
  "measurements": {
    "mac": {
      "source_sha256": "713bb2d455235bf3348828591ae60b6ffcd7a2e637c718b27f416e2edcde6465",
      "hardware": "Apple M3, macOS 27.3, CPU_ONLY, Python host process",
      "zh_v1_int8": {
        "package_bytes": 22792958,
        "load_ms_p50": 28.388874998199753,
        "warm_p50_of_process_p50_ms": {
          "T1": 1.652375000048778,
          "T8": 2.4276039985124953,
          "T32": 3.84631300039473,
          "T64": 5.521499999304069,
          "T128": 8.48720900103217
        },
        "warm_p50_of_process_p95_ms": {
          "T1": 2.1141144019566127,
          "T8": 2.8757605994542246,
          "T32": 4.280841699255689,
          "T64": 5.88047934925271,
          "T128": 8.882592001282319
        }
      },
      "models": [
        "v42-fp32",
        "v42-int8",
        "v42-kvchunk-fp32",
        "v42-kvchunk-int8",
        "zhv1-fp32",
        "zhv1-int8",
        "zhv1-kvchunk-fp32",
        "zhv1-kvchunk-int8"
      ]
    },
    "iphone": null,
    "keyboard_extension": null
  },
  "deployment_status": "Release for iOS deployment in Vime. Core ML packages pass the Mac and simulator functional gates; iPhone and keyboard-extension measurements have not been run yet.",
  "scope": "Aggregate metrics only. No corpus, benchmark text, candidate pools, per-case scores, case identifiers, user inputs or raw device traces are distributed."
}