Fix GM-Team bilingual title parsing

MAINTENANCE.md

@@ -50,7 +50,7 @@ uv run python train.py \
   --tokenizer char \
   --data-file datasets/AnimeName/dmhy_weak_char.jsonl \
   --vocab-file datasets/AnimeName/vocab.char.json \
-  --save-dir checkpoints/dmhy-char-full-relabel \
+  --save-dir checkpoints/dmhy-char-guoman-relabel \
   --init-model-dir . \
   --epochs 2 \
   --batch-size 256 \
@@ -59,7 +59,7 @@ uv run python train.py \
   --max-seq-length 128 \
   --checkpoint-steps 1000 \
   --parse-eval-limit 2048 \
-  --seed 48
+  --seed 52
 ```
 
 ## Publish a New Checkpoint
@@ -67,15 +67,15 @@ uv run python train.py \
 Copy the final checkpoint to the repository root:
 
 ```powershell
-Copy-Item checkpoints/dmhy-char-full-relabel/final/config.json . -Force
-Copy-Item checkpoints/dmhy-char-full-relabel/final/model.safetensors . -Force
-Copy-Item checkpoints/dmhy-char-full-relabel/final/tokenizer_config.json . -Force
-Copy-Item checkpoints/dmhy-char-full-relabel/final/training_args.bin . -Force
-Copy-Item checkpoints/dmhy-char-full-relabel/final/vocab.json . -Force
+Copy-Item checkpoints/dmhy-char-guoman-relabel/final/config.json . -Force
+Copy-Item checkpoints/dmhy-char-guoman-relabel/final/model.safetensors . -Force
+Copy-Item checkpoints/dmhy-char-guoman-relabel/final/tokenizer_config.json . -Force
+Copy-Item checkpoints/dmhy-char-guoman-relabel/final/training_args.bin . -Force
+Copy-Item checkpoints/dmhy-char-guoman-relabel/final/vocab.json . -Force
 Copy-Item datasets/AnimeName/vocab.char.json .\vocab.char.json -Force
-Copy-Item checkpoints/dmhy-char-full-relabel/final/run_metadata.json . -Force
-Copy-Item checkpoints/dmhy-char-full-relabel/final/trainer_eval_metrics.json . -Force
-Copy-Item checkpoints/dmhy-char-full-relabel/final/parse_eval_metrics.json . -Force
+Copy-Item checkpoints/dmhy-char-guoman-relabel/final/run_metadata.json . -Force
+Copy-Item checkpoints/dmhy-char-guoman-relabel/final/trainer_eval_metrics.json . -Force
+Copy-Item checkpoints/dmhy-char-guoman-relabel/final/parse_eval_metrics.json . -Force
 ```
 
 There is no tracked `model/` duplicate. The root checkpoint is the publishing

README.md

@@ -59,21 +59,21 @@ dataset relabeling and diagnostics, but the root checkpoint loads as `char`.
 ## Evaluation
 
 Final full-relabel char training (`632002` DMHY rows, 2 epochs, batch size 256,
-seed 48):
+seed 52):
 
 | Metric | Value |
 |--------|-------|
-| Eval loss | 0.0163 |
-| Entity precision | 0.9800 |
-| Entity recall | 0.9867 |
-| Entity F1 | 0.9833 |
-| Token accuracy | 0.9943 |
-| Held-out parse full match | 2008/2048 (0.9805) |
-| Fixed regression full match | 21/21 (1.0000) |
+| Eval loss | 0.0058 |
+| Entity precision | 0.9922 |
+| Entity recall | 0.9946 |
+| Entity F1 | 0.9934 |
+| Token accuracy | 0.9981 |
+| Held-out parse full match | 2029/2048 (0.9907) |
+| Fixed regression full match | 22/22 (1.0000) |
 
 The fixed regression set includes second-season aliases such as `Ni`,
-`Ni no Sara`, `貳`, and `弐ノ章`, plus long-running episode IDs and dense meta
-blocks.
+`Ni no Sara`, `貳`, and `弐ノ章`, plus GM-Team bilingual Chinese animation
+bracket layouts, long-running episode IDs, and dense meta blocks.
 
 ## Usage
 
@@ -121,13 +121,13 @@ uv run python convert_to_char_dataset.py \
 uv run python train.py --tokenizer char \
   --data-file datasets/AnimeName/dmhy_weak_char.jsonl \
   --vocab-file datasets/AnimeName/vocab.char.json \
-  --save-dir checkpoints/dmhy-char-full-relabel \
+  --save-dir checkpoints/dmhy-char-guoman-relabel \
   --init-model-dir . \
   --epochs 2 --batch-size 256 \
   --learning-rate 0.00008 --warmup-steps 300 \
   --checkpoint-steps 1000 --save-total-limit 3 \
   --parse-eval-limit 2048 \
-  --max-seq-length 128 --seed 48
+  --max-seq-length 128 --seed 52
 ```
 
 The converter keeps source metadata and adds `tokenizer_variant`, source token

build_repair_focus_dataset.py

@@ -88,6 +88,42 @@ def manual_cases() -> Iterable[dict]:
             ("FLAC", "SOURCE"),
         ],
     )
+    yield char_item(
+        "[GM-Team][国漫][逆天邪神 第2季][Against the Gods Ⅱ][2026][04][HEVC][GB][4K].mp4",
+        [
+            ("GM-Team", "GROUP"),
+            ("逆天邪神", "TITLE"),
+            ("第2季", "SEASON"),
+            ("04", "EPISODE"),
+            ("HEVC", "SOURCE"),
+            ("GB", "SOURCE"),
+            ("4K", "RESOLUTION"),
+        ],
+    )
+    yield char_item(
+        "[GM-Team][国漫][剑来 第2季][Sword of Coming Ⅱ][2025][04][HEVC][GB][4K]",
+        [
+            ("GM-Team", "GROUP"),
+            ("剑来", "TITLE"),
+            ("第2季", "SEASON"),
+            ("04", "EPISODE"),
+            ("HEVC", "SOURCE"),
+            ("GB", "SOURCE"),
+            ("4K", "RESOLUTION"),
+        ],
+    )
+    yield char_item(
+        "[GM-Team][国漫][大主宰 第2季][The Great Ruler Ⅱ][2026][04][HEVC][GB][4K]",
+        [
+            ("GM-Team", "GROUP"),
+            ("大主宰", "TITLE"),
+            ("第2季", "SEASON"),
+            ("04", "EPISODE"),
+            ("HEVC", "SOURCE"),
+            ("GB", "SOURCE"),
+            ("4K", "RESOLUTION"),
+        ],
+    )
 
 
 def main() -> None:

case_metrics.json

@@ -1,29 +1,29 @@
 {
   "model_dir": ".",
-  "case_file": "data\\parser_regression_cases.json",
+  "case_file": "data/parser_regression_cases.json",
   "tokenizer_variant": "char",
   "max_length": 128,
   "use_rules": true,
   "constrain_bio": true,
-  "case_count": 21,
-  "full_correct": 21,
+  "case_count": 22,
+  "full_correct": 22,
   "full_accuracy": 1.0,
   "field_correct": {
-    "group": 18,
-    "title": 21,
-    "episode": 21,
-    "resolution": 21,
-    "source": 14,
-    "season": 8,
+    "group": 19,
+    "title": 22,
+    "episode": 22,
+    "resolution": 22,
+    "source": 15,
+    "season": 9,
     "special": 1
   },
   "field_total": {
-    "group": 18,
-    "title": 21,
-    "episode": 21,
-    "resolution": 21,
-    "source": 14,
-    "season": 8,
+    "group": 19,
+    "title": 22,
+    "episode": 22,
+    "resolution": 22,
+    "source": 15,
+    "season": 9,
     "special": 1
   },
   "field_accuracy": {
@@ -454,6 +454,28 @@
         "season": 2,
         "title": "炎炎の消防隊"
       }
+    },
+    {
+      "id": "gm_team_guoman_bilingual_s2",
+      "filename": "[GM-Team][国漫][逆天邪神 第2季][Against the Gods Ⅱ][2026][04][HEVC][GB][4K].mp4",
+      "ok": true,
+      "errors": {},
+      "expected": {
+        "group": "GM-Team",
+        "title": "逆天邪神",
+        "season": 2,
+        "episode": 4,
+        "resolution": "4K",
+        "source": "GB"
+      },
+      "pred": {
+        "episode": 4,
+        "group": "GM-Team",
+        "resolution": "4K",
+        "season": 2,
+        "source": "GB",
+        "title": "逆天邪神"
+      }
     }
   ]
 }

data/parser_regression_cases.json

@@ -228,5 +228,17 @@
       "episode": 13,
       "resolution": "1920x1080"
     }
+  },
+  {
+    "id": "gm_team_guoman_bilingual_s2",
+    "filename": "[GM-Team][国漫][逆天邪神 第2季][Against the Gods Ⅱ][2026][04][HEVC][GB][4K].mp4",
+    "expected": {
+      "group": "GM-Team",
+      "title": "逆天邪神",
+      "season": 2,
+      "episode": 4,
+      "resolution": "4K",
+      "source": "GB"
+    }
   }
 ]

datasets/AnimeName

@@ -1 +1 @@
-Subproject commit 8d2b6c9e639fde6be0e428e5f34f56fccd5aa2ea
+Subproject commit 004a8c08628b6820fb2d1b59a80fdcfe925ef095

dmhy_dataset.py

@@ -35,6 +35,10 @@ NOISE_BRACKETS = {
     "tc", "sc", "gb", "big5", "cht", "chs", "jpn", "jp", "jap", "eng",
     "繁中", "简中", "繁日", "简日", "日语", "日文", "外挂", "内封", "字幕",
 }
+CATEGORY_BRACKETS = {
+    "国漫", "國漫", "国产", "國產", "国产动漫", "國產動漫", "国产动画", "國產動畫",
+    "国创", "國創", "中国动漫", "中國動漫", "中国动画", "中國動畫",
+}
 
 SPECIAL_RE = re.compile(r"^(?:ova\d*|oad\d*|sp\d*|movie|the\s*movie|op|ed|pv|cm|ncop|nced|剧场版|劇場版|特别篇|特別篇)$", re.I)
 SPECIAL_SEARCH_RE = re.compile(r"^(?:檢索|检索|搜索|搜寻|搜尋|别名|別名|alias|search|keyword)\s*[:：].+", re.I)
@@ -186,7 +190,8 @@ def is_source(token: str) -> bool:
         return True
     if has_wrapping_brackets(token):
         parts = [part for part in re.split(r"[\s&+/,._-]+", clean) if part]
-        return bool(parts) and all(SOURCE_RE.match(part) or is_noise_bracket(part) for part in parts)
+        has_source_part = any(SOURCE_RE.match(part) for part in parts)
+        return has_source_part and all(SOURCE_RE.match(part) or is_noise_bracket(part) for part in parts)
     return False
 
 
@@ -195,6 +200,11 @@ def is_special(token: str) -> bool:
     return bool(SPECIAL_RE.match(clean) or SPECIAL_SEARCH_RE.match(clean))
 
 
+def is_category_bracket(token: str) -> bool:
+    clean = re.sub(r"[\s._-]+", "", clean_bracket(token))
+    return has_wrapping_brackets(token) and clean in CATEGORY_BRACKETS
+
+
 def is_noise_bracket(token: str) -> bool:
     clean = clean_bracket(token)
     if not clean:
@@ -202,6 +212,8 @@ def is_noise_bracket(token: str) -> bool:
     normalized = re.sub(r"[\s._-]+", "", clean).lower()
     if normalized in NOISE_BRACKETS:
         return True
+    if is_category_bracket(token):
+        return True
     if DATE_RE.match(clean) or HASH_RE.match(clean):
         return True
     return False
@@ -335,6 +347,42 @@ def label_context_season_tokens(
             categories[idx] = "season"
 
 
+def repair_structured_bracket_title_aliases(
+    tokens: Sequence[str],
+    categories: List[str],
+    episode_idx: int,
+) -> None:
+    """Keep the primary title in category-prefixed bracket series.
+
+    GM-Team-style rows often look like:
+    [GROUP][国漫][中文标题 第2季][English Alias Ⅱ][2026][04][meta]
+    The category, alias, and year brackets are metadata for parsing purposes;
+    the first real title bracket after the category is the canonical title.
+    """
+    if not any(is_category_bracket(tokens[idx]) for idx in range(min(episode_idx, len(tokens)))):
+        return
+
+    title_candidates = [
+        idx
+        for idx in range(episode_idx)
+        if categories[idx] == "title"
+        and has_wrapping_brackets(tokens[idx])
+        and is_title_token(tokens[idx])
+    ]
+    if not title_candidates:
+        return
+
+    primary_idx = title_candidates[0]
+    for idx in title_candidates[1:]:
+        categories[idx] = "sep"
+
+    for idx in range(episode_idx):
+        if idx == primary_idx:
+            continue
+        if is_category_bracket(tokens[idx]) or DATE_RE.match(clean_bracket(tokens[idx])):
+            categories[idx] = "sep"
+
+
 def embedded_bracket_episode(token: str) -> Optional[tuple[str, str, str]]:
     """Split malformed tokens such as '[Group}Title[658]' into title + episode."""
     if episode_number(token) is not None:
@@ -390,6 +438,10 @@ def finalize_weak_sample(
             continue
         if is_explicit_season(token):
             expanded_categories[idx] = "season"
+            prev_idx = idx - 1
+            while prev_idx >= 0 and is_separator_token(expanded_tokens[prev_idx]) and expanded_categories[prev_idx] == "title":
+                expanded_categories[prev_idx] = "sep"
+                prev_idx -= 1
 
     labels = assign_iob2(expanded_categories)
     if len(expanded_tokens) != len(labels):
@@ -699,7 +751,9 @@ def weak_label_filename(filename: str, tokenizer: AnimeTokenizer) -> Optional[di
     for idx, token in enumerate(tokens):
         if categories[idx] == "group":
             continue
-        if is_resolution(token):
+        if is_category_bracket(token):
+            categories[idx] = "sep"
+        elif is_resolution(token):
             categories[idx] = "resolution"
         elif is_source(token):
             categories[idx] = "source"
@@ -715,6 +769,7 @@ def weak_label_filename(filename: str, tokenizer: AnimeTokenizer) -> Optional[di
         return fallback_embedded_episode_sample(tokens, tokenizer) or fallback_no_episode_sample(tokens, tokenizer)
     categories[episode_idx] = "episode"
     label_context_season_tokens(tokens, categories, episode_idx)
+    repair_structured_bracket_title_aliases(tokens, categories, episode_idx)
 
     # S01E07 is tokenized as S01 + E07 after tokenizer changes. If an older
     # token slips through, expand_tokens_and_categories will split it.

exports/anime_filename_parser.metadata.json

@@ -8,5 +8,5 @@
     128,
     15
   ],
-  "max_abs_diff": 3.3855438232421875e-05
+  "max_abs_diff": 5.65648078918457e-05
 }

exports/anime_filename_parser.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f9b874fbd4217a190487f512dcc6dd7ce2f0e610147703ca0cddcc0db44fb1c7
+oid sha256:6d967c5c2305e6737c9e791956a174655deebef2cfa477e081890ebddd56e004
 size 19633926

inference.py

@@ -330,6 +330,11 @@ NOISE_META_RE = re.compile(
     r"Opus|ASS.*|CHS|CHT|BIG5|GB|JPN?|MP4|MKV|繁中|简中|内封|外挂)$",
     re.I,
 )
+DATE_RE = re.compile(r"^(?:19|20)\d{2}(?:[.\-_年]?(?:0?[1-9]|1[0-2]))?(?:[.\-_月]?(?:0?[1-9]|[12]\d|3[01]))?日?$")
+CATEGORY_BRACKETS = {
+    "国漫", "國漫", "国产", "國產", "国产动漫", "國產動漫", "国产动画", "國產動畫",
+    "国创", "國創", "中国动漫", "中國動漫", "中国动画", "中國動畫",
+}
 
 
 def cn_number_to_int(text: str) -> Optional[int]:
@@ -372,8 +377,11 @@ def looks_like_episode_or_meta(text: str) -> bool:
     if not text:
         return False
     clean = text.strip()
+    normalized = re.sub(r"[\s._-]+", "", clean)
     return bool(
         re.fullmatch(r"(?:EP?|#)?\d{1,4}(?:v\d+)?", clean, re.I)
+        or DATE_RE.fullmatch(clean)
+        or normalized in CATEGORY_BRACKETS
         or RESOLUTION_RE.search(clean)
         or SOURCE_TAG_RE.fullmatch(clean)
         or SOURCE_RE.search(clean)
@@ -492,6 +500,10 @@ def apply_rule_assists(filename: str, result: Dict) -> Dict:
         if repaired_title:
             repaired["title"] = repaired_title
 
+    structured_title = infer_structured_bracket_title(filename, group, repaired.get("episode"))
+    if structured_title:
+        repaired["title"] = structured_title
+
     if repaired.get("title") and repaired.get("season") is not None:
         repaired["title"] = strip_trailing_season_from_title(repaired["title"], repaired["season"])
 
@@ -584,6 +596,56 @@ def source_candidates(filename: str) -> List[str]:
     return [value for _priority, _neg_start, value in sorted(deduped.values(), reverse=True)]
 
 
+def is_category_text(text: str) -> bool:
+    normalized = re.sub(r"[\s._-]+", "", text.strip())
+    return normalized in CATEGORY_BRACKETS
+
+
+def infer_structured_bracket_title(
+    filename: str,
+    group: Optional[str],
+    episode: Optional[int],
+) -> Optional[str]:
+    """Pick the primary title from [group][category][title][alias][year][episode] rows."""
+    brackets = bracket_parts(filename)
+    if len(brackets) < 4 or episode is None:
+        return None
+
+    start_index = 0
+    if group and brackets and brackets[0][0] == group:
+        start_index = 1
+
+    search = brackets[start_index:]
+    if not search or not any(is_category_text(text) for text, _start, _end in search[:2]):
+        return None
+
+    episode_index = None
+    for idx, (text, _start, _end) in enumerate(brackets):
+        if re.fullmatch(rf"(?:EP?|#)?0*{episode}(?:v\d+)?", text.strip(), re.I):
+            episode_index = idx
+            break
+    if episode_index is None:
+        return None
+
+    candidates: List[Tuple[int, str]] = []
+    for idx in range(start_index, episode_index):
+        text = brackets[idx][0].strip()
+        if not text or looks_like_episode_or_meta(text):
+            continue
+        score = 0
+        if SEASON_RE.search(text) or TRAILING_SEQUEL_MARKER_RE.search(text):
+            score += 50
+        if re.search(r"[\u3400-\u9fff]", text):
+            score += 20
+        if idx > start_index:
+            score += 10
+        candidates.append((score, text))
+
+    if not candidates:
+        return None
+    return max(candidates, key=lambda item: item[0])[1]
+
+
 def best_structural_episode(filename: str) -> Optional[int]:
     priorities = {
         "season_episode": 1000,
@@ -635,6 +697,7 @@ def strip_trailing_season_from_title(title: str, season: int) -> str:
         rf"\s+[Ss]0*{season_text}$",
         rf"\s+Season\s*0*{season_text}$",
         rf"\s+0*{season_text}$",
+        rf"\s+第(?:0*{season_text}|{season_text})[季期部章]$",
     ]
     cleaned = title
     for pattern in patterns:

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:697d7491b83ef615994e02f11f0f65362c400f5eb6b4be8f43f02435ad43173f
+oid sha256:347b2f619fd63a71804c4742a069b20acd0cde870fc03cc2ac0f175b06586b72
 size 19142604

parse_eval_metrics.json

@@ -1,22 +1,22 @@
 {
   "sample_count": 2048,
   "field_accuracy": {
-    "group": 1.0,
-    "title": 0.99658203125,
-    "season": 0.994140625,
-    "episode": 0.99609375,
-    "resolution": 0.998046875,
-    "source": 0.99365234375,
-    "special": 0.998046875
+    "group": 0.99951171875,
+    "title": 0.99755859375,
+    "season": 0.99609375,
+    "episode": 0.998046875,
+    "resolution": 1.0,
+    "source": 0.99853515625,
+    "special": 0.9990234375
   },
   "field_correct": {
-    "group": 2048,
-    "title": 2041,
-    "season": 2036,
-    "episode": 2040,
-    "resolution": 2044,
-    "source": 2035,
-    "special": 2044
+    "group": 2047,
+    "title": 2043,
+    "season": 2040,
+    "episode": 2044,
+    "resolution": 2048,
+    "source": 2045,
+    "special": 2046
   },
   "field_total": {
     "group": 2048,
@@ -27,487 +27,487 @@
     "source": 2048,
     "special": 2048
   },
-  "full_match_accuracy": 0.98046875,
-  "full_match_correct": 2008,
+  "full_match_accuracy": 0.99072265625,
+  "full_match_correct": 2029,
   "full_match_total": 2048,
   "failures": [
     {
-      "filename": "[DBD-Raws][Boruto Naruto Next Generations][menu][S13][D2][02][1080P][BDRip][HEVC-10bit][FLAC]",
+      "filename": "[ig]Itai no wa Iya nano de Bougyoryoku ni Kyokufuri Shitai to Omoimasu[WebRip 1920x1080 AVC YUV420 8Bit 1080p  AAC].03.TC",
       "errors": {
-        "season": {
-          "gold": null,
-          "pred": "13"
+        "episode": {
+          "gold": "3",
+          "pred": null
         }
       },
       "gold": {
-        "group": "DBD-Raws",
-        "title": "Boruto Naruto Next Generations",
+        "group": "ig",
+        "title": "Itai no wa Iya nano de Bougyoryoku ni Kyokufuri Shitai to Omoimasu",
         "season": null,
-        "episode": 2,
-        "resolution": "1080P",
-        "source": "BDRip",
+        "episode": 3,
+        "resolution": "1080p",
+        "source": "WebRip",
         "special": null
       },
       "pred": {
-        "group": "DBD-Raws",
-        "title": "Boruto Naruto Next Generations",
-        "season": 13,
-        "episode": 2,
-        "resolution": "1080P",
-        "source": "BDRip",
+        "group": "ig",
+        "title": "Itai no wa Iya nano de Bougyoryoku ni Kyokufuri Shitai to Omoimasu",
+        "season": null,
+        "episode": null,
+        "resolution": "1080p",
+        "source": "WebRip",
         "special": null
       }
     },
     {
-      "filename": "[アニメ BD] ギャラクシーエンジェル 第1期(無印) 第09話「ロストテクノロジーのローストビーフ」(1424x1072 HEVC 10bit FLAC softSub(chi+eng) chap)",
+      "filename": "[YYDM-11FANS][Nanana's Buried Treasure][preview][09][BDrip][720P][X264-10bit_AAC][34D29ED6]",
       "errors": {
-        "season": {
-          "gold": null,
-          "pred": "1"
+        "special": {
+          "gold": "ed",
+          "pred": null
         }
       },
       "gold": {
-        "group": "アニメ BD",
-        "title": "ギャラクシーエンジェル 第1期(無印) 第09話「ロストテクノロジーのローストビーフ」",
+        "group": "YYDM-11FANS",
+        "title": "Nanana's Buried Treasure",
         "season": null,
         "episode": 9,
-        "resolution": "1424x1072",
-        "source": "BD",
-        "special": null
+        "resolution": "720P",
+        "source": "BDrip",
+        "special": "ED"
       },
       "pred": {
-        "group": "アニメ BD",
-        "title": "ギャラクシーエンジェル 第1期(無印) 第09話「ロストテクノロジーのローストビーフ」",
-        "season": 1,
+        "group": "YYDM-11FANS",
+        "title": "Nanana's Buried Treasure",
+        "season": null,
         "episode": 9,
-        "resolution": "1424x1072",
-        "source": "BD",
+        "resolution": "720P",
+        "source": "BDrip",
         "special": null
       }
     },
     {
-      "filename": "コメットさん☆ 第11話 「バトンの力」(DVD DivX4.12 QB95 640x480 24f) [CRC32_C09E1AB0]",
+      "filename": "[Moozzi2] Madou King Granzort Saigo no Magical Taisen OVA - 01 [ 1990 ] (BD 1440x1080 x.264 Flac)",
       "errors": {
-        "source": {
-          "gold": "cr",
-          "pred": "dvd"
+        "title": {
+          "gold": "madou king granzort saigo no magical taisen ova",
+          "pred": "madou king granzort saigo no magical taisen ova - 01 [ 1990"
+        },
+        "episode": {
+          "gold": "1",
+          "pred": "1990"
         }
       },
       "gold": {
-        "group": null,
-        "title": "コメットさん☆",
+        "group": "Moozzi2",
+        "title": "Madou King Granzort Saigo no Magical Taisen OVA",
         "season": null,
-        "episode": 11,
-        "resolution": "640x480",
-        "source": "CR",
-        "special": null
+        "episode": 1,
+        "resolution": "1440x1080",
+        "source": "BD",
+        "special": "OVA"
       },
       "pred": {
-        "group": null,
-        "title": "コメットさん☆",
+        "group": "Moozzi2",
+        "title": "Madou King Granzort Saigo no Magical Taisen OVA - 01 [ 1990 ",
         "season": null,
-        "episode": 11,
-        "resolution": "640x480",
-        "source": "DVD",
-        "special": null
+        "episode": 1990,
+        "resolution": "1440x1080",
+        "source": "BD",
+        "special": "OVA"
       }
     },
     {
-      "filename": "[Kamigami&Mabors&VCB-Studio] Saenai Heroine no Sodatekata Flat [07][Ma10p_1080p][x265_2aac]",
+      "filename": "[64bitsub][Tensui no Sakuna-hime][08][BDRIP_1920x1080][AVC_FLAC_SUP]",
       "errors": {
         "source": {
-          "gold": "aac",
-          "pred": "x265-2aac"
+          "gold": "flac",
+          "pred": "avc-flac"
         }
       },
       "gold": {
-        "group": "Kamigami&Mabors&VCB-Studio",
-        "title": "Saenai Heroine no Sodatekata Flat",
+        "group": "64bitsub",
+        "title": "Tensui no Sakuna-hime",
         "season": null,
-        "episode": 7,
-        "resolution": "1080p",
-        "source": "aac",
+        "episode": 8,
+        "resolution": "1920x1080",
+        "source": "FLAC",
         "special": null
       },
       "pred": {
-        "group": "Kamigami&Mabors&VCB-Studio",
-        "title": "Saenai Heroine no Sodatekata Flat",
+        "group": "64bitsub",
+        "title": "Tensui no Sakuna-hime",
         "season": null,
-        "episode": 7,
-        "resolution": "1080p",
-        "source": "x265_2aac",
+        "episode": 8,
+        "resolution": "1920x1080",
+        "source": "AVC_FLAC",
         "special": null
       }
     },
     {
-      "filename": "[Liuyun&VCB-Studio] Hanasaku Iroha [07][Hi10p_1080p][x264_flac_ac3]",
+      "filename": "[VCB-Studio] Shingeki no Kyojin Movie 3 Kakusei no Houkou [Teaser_S3][Ma10p_1080p][x265_flac]",
       "errors": {
-        "source": {
-          "gold": "flac",
-          "pred": "x264-flac"
+        "season": {
+          "gold": null,
+          "pred": "3"
         }
       },
       "gold": {
-        "group": "Liuyun&VCB-Studio",
-        "title": "Hanasaku Iroha",
+        "group": "VCB-Studio",
+        "title": "Shingeki no Kyojin Movie 3 Kakusei no Houkou [Teaser_S3",
         "season": null,
-        "episode": 7,
+        "episode": 3,
         "resolution": "1080p",
-        "source": "flac",
-        "special": null
+        "source": "x265_flac",
+        "special": "Movie"
       },
       "pred": {
-        "group": "Liuyun&VCB-Studio",
-        "title": "Hanasaku Iroha",
-        "season": null,
-        "episode": 7,
+        "group": "VCB-Studio",
+        "title": "Shingeki no Kyojin Movie 3 Kakusei no Houkou [Teaser_S3",
+        "season": 3,
+        "episode": 3,
         "resolution": "1080p",
-        "source": "x264_flac",
-        "special": null
+        "source": "x265_flac",
+        "special": "Movie"
       }
     },
     {
-      "filename": "小新外传4[EP02][2017.06.07]出动！妖怪克星",
+      "filename": "FF：U ファイナルファンタジー：アンリミテッド ～異界の章～ #15 「ジェーン～うごきだすうみパズル」(DVD 640x480 DivX5 QB98 120fps lameVBR)[CRC_5FA44899]",
       "errors": {
-        "title": {
-          "gold": "小新外传4 ep02 2017 06",
-          "pred": "小新外传 ep02 2"
-        },
-        "season": {
-          "gold": null,
-          "pred": "4"
-        },
-        "episode": {
-          "gold": "7",
-          "pred": "2"
+        "source": {
+          "gold": "cr",
+          "pred": "dvd"
         }
       },
       "gold": {
         "group": null,
-        "title": "小新外传4 EP02 2017 06",
+        "title": "FF：U ファイナルファンタジー：アンリミテッド ～異界の章～",
         "season": null,
-        "episode": 7,
-        "resolution": null,
-        "source": null,
+        "episode": 15,
+        "resolution": "640x480",
+        "source": "CR",
         "special": null
       },
       "pred": {
         "group": null,
-        "title": "小新外传 EP02 2",
-        "season": 4,
-        "episode": 2,
-        "resolution": null,
-        "source": null,
+        "title": "FF：U ファイナルファンタジー：アンリミテッド ～異界の章～",
+        "season": null,
+        "episode": 15,
+        "resolution": "640x480",
+        "source": "DVD",
         "special": null
       }
     },
     {
-      "filename": "[GM-Team][国漫][异常生物见闻录][The Record of Unusual Creatures][2019][12][HEVC][GB][3840×2160]",
+      "filename": "[OVA]GALLFORCE ガルフォース2 宇宙章 vol2 [DESTRUCTION]",
       "errors": {
-        "resolution": {
-          "gold": "3840×2160",
-          "pred": "3840×"
+        "title": {
+          "gold": "gallforce ガルフォース2 宇宙章 vol",
+          "pred": "gallforce ガルフォース2 宇宙"
         }
       },
       "gold": {
-        "group": "GM-Team",
-        "title": "国漫",
+        "group": "OVA",
+        "title": "GALLFORCE ガルフォース2 宇宙章 vol",
         "season": null,
-        "episode": 12,
-        "resolution": "3840×2160",
-        "source": "GB",
-        "special": null
+        "episode": 2,
+        "resolution": null,
+        "source": null,
+        "special": "OVA"
       },
       "pred": {
-        "group": "GM-Team",
-        "title": "国漫",
+        "group": "OVA",
+        "title": "GALLFORCE ガルフォース2 宇宙",
         "season": null,
-        "episode": 12,
-        "resolution": "3840×",
-        "source": "GB",
-        "special": null
+        "episode": 2,
+        "resolution": null,
+        "source": null,
+        "special": "OVA"
       }
     },
     {
-      "filename": "Ⅱ 116 第108次鐘聲已經敲過了嗎？",
+      "filename": "[病毒].[Fosky_Fansub][Virus_Buster_Serge][DVDrip][12][H264_AAC][640x480][GB&BIG5][F77551D0](ED2000.COM)",
       "errors": {
-        "title": {
-          "gold": "ⅱ 116 第",
-          "pred": "第"
+        "special": {
+          "gold": "ed",
+          "pred": "e"
         }
       },
       "gold": {
-        "group": null,
-        "title": "Ⅱ 116 第",
+        "group": "病毒",
+        "title": "Fosky_Fansub",
         "season": null,
-        "episode": 116,
-        "resolution": null,
-        "source": null,
-        "special": null
+        "episode": 12,
+        "resolution": "640x480",
+        "source": "DVDrip",
+        "special": "ED"
       },
       "pred": {
-        "group": null,
-        "title": "第",
+        "group": "病毒",
+        "title": "Fosky_Fansub",
         "season": null,
-        "episode": 116,
-        "resolution": null,
-        "source": null,
-        "special": null
+        "episode": 12,
+        "resolution": "640x480",
+        "source": "DVDrip",
+        "special": "E"
       }
     },
     {
-      "filename": "EP08 & EP11 NCED",
+      "filename": "[DBD-Raws][Shadows House S1][Gekijou][18][1080P][BDRip][HEVC-10bit][FLAC]",
       "errors": {
-        "title": {
-          "gold": "&",
-          "pred": "ep"
+        "season": {
+          "gold": null,
+          "pred": "1"
         }
       },
       "gold": {
-        "group": null,
-        "title": "&",
+        "group": "DBD-Raws",
+        "title": "Shadows House",
         "season": null,
-        "episode": 11,
-        "resolution": null,
-        "source": null,
-        "special": "NCED"
+        "episode": 18,
+        "resolution": "1080P",
+        "source": "BDRip",
+        "special": null
       },
       "pred": {
-        "group": null,
-        "title": "EP",
-        "season": null,
-        "episode": 11,
-        "resolution": null,
-        "source": null,
-        "special": "NCED"
+        "group": "DBD-Raws",
+        "title": "Shadows House",
+        "season": 1,
+        "episode": 18,
+        "resolution": "1080P",
+        "source": "BDRip",
+        "special": null
       }
     },
     {
-      "filename": "[S1YURICON] Necronomico no Cosmic Horror Show[06][1080p][WebRip][HEVC_AAC][CHS]",
+      "filename": "Girls und Panzer - 10.5 (BD 1280x720 AVC AACx2)",
       "errors": {
         "season": {
-          "gold": null,
+          "gold": "10",
           "pred": "1"
         }
       },
       "gold": {
-        "group": "S1YURICON",
-        "title": "Necronomico no Cosmic Horror Show",
-        "season": null,
-        "episode": 6,
-        "resolution": "1080p",
-        "source": "WebRip",
+        "group": null,
+        "title": "Girls und Panzer - 10.5",
+        "season": 10,
+        "episode": 5,
+        "resolution": "1280x720",
+        "source": "BD",
         "special": null
       },
       "pred": {
-        "group": "S1YURICON",
-        "title": "Necronomico no Cosmic Horror Show",
+        "group": null,
+        "title": "Girls und Panzer - 10.5",
         "season": 1,
-        "episode": 6,
-        "resolution": "1080p",
-        "source": "WebRip",
+        "episode": 5,
+        "resolution": "1280x720",
+        "source": "BD",
         "special": null
       }
     },
     {
-      "filename": "[FZsub]Gate - Jieitai Kanochi nite, Kaku Tatakaeri 2 - 02(14) (MX 1280x720 x264 AAC)_x264",
+      "filename": "[POPGO&SumiSora&TxxZ] Ginga Eiyuu Densetsu Die Neue These - Seiran 14 (BDRip 1080P X265 Main10p TrueHDX2 Chap)[A4E18C32]",
       "errors": {
-        "title": {
-          "gold": "gate - jieitai kanochi nite, kaku tatakaeri 2",
-          "pred": "gate - jieitai kanochi nite, kaku tatakaeri 2 - 02"
+        "group": {
+          "gold": null,
+          "pred": "popgo&sumisora&txxz"
         },
-        "season": {
-          "gold": "2",
-          "pred": null
+        "title": {
+          "gold": "popgo&sumisora&txxz",
+          "pred": "ginga eiyuu densetsu die neue these - seiran 14"
         }
       },
       "gold": {
-        "group": "FZsub",
-        "title": "Gate - Jieitai Kanochi nite, Kaku Tatakaeri 2",
-        "season": 2,
+        "group": null,
+        "title": "POPGO&SumiSora&TxxZ",
+        "season": null,
         "episode": 14,
-        "resolution": "1280x720",
-        "source": "x264",
+        "resolution": "1080P",
+        "source": "BDRip",
         "special": null
       },
       "pred": {
-        "group": "FZsub",
-        "title": "Gate - Jieitai Kanochi nite, Kaku Tatakaeri 2 - 02",
+        "group": "POPGO&SumiSora&TxxZ",
+        "title": "Ginga Eiyuu Densetsu Die Neue These - Seiran 14",
         "season": null,
         "episode": 14,
-        "resolution": "1280x720",
-        "source": "x264",
+        "resolution": "1080P",
+        "source": "BDRip",
         "special": null
       }
     },
     {
-      "filename": "Mobile Suit Gundam SEED Destiny - HD Remaster 2013 Anime Music Clip2 - Life Goes On [BD 1248x702 23.976fps AVC-yuv420p10 FLAC] v2 - yan04000985",
+      "filename": "[アニメ BD] Serial Experiments Lain 映像特典 「trailer 01」 （1440x1080 x264 AAC 2ch）",
       "errors": {
+        "title": {
+          "gold": "serial experiments lain 映像特典 「trailer 01」",
+          "pred": "serial experiments lain 映像特典 「trailer"
+        },
         "episode": {
-          "gold": null,
-          "pred": "23"
+          "gold": "2",
+          "pred": "1"
         }
       },
       "gold": {
-        "group": null,
-        "title": "Mobile Suit Gundam SEED Destiny - HD Remaster 2013 Anime Music Clip2 - Life Goes On",
+        "group": "アニメ BD",
+        "title": "Serial Experiments Lain 映像特典 「trailer 01」",
         "season": null,
-        "episode": null,
-        "resolution": "1248x702",
+        "episode": 2,
+        "resolution": "1440x1080",
         "source": "BD",
         "special": null
       },
       "pred": {
-        "group": null,
-        "title": "Mobile Suit Gundam SEED Destiny - HD Remaster 2013 Anime Music Clip2 - Life Goes On",
+        "group": "アニメ BD",
+        "title": "Serial Experiments Lain 映像特典 「trailer",
         "season": null,
-        "episode": 23,
-        "resolution": "1248x702",
+        "episode": 1,
+        "resolution": "1440x1080",
         "source": "BD",
         "special": null
       }
     },
     {
-      "filename": "Mary_E_Il_Giardino_Segreto_-_07_-_Camilla_[DvdMUX_by_Magic_©2008]",
+      "filename": "[AJZ＆BLU][God Eater][05][BIG5][v2] (2)",
       "errors": {
-        "source": {
-          "gold": null,
-          "pred": "dvd"
+        "episode": {
+          "gold": "2",
+          "pred": "5"
         }
       },
       "gold": {
-        "group": null,
-        "title": "Mary_E_Il_Giardino_Segreto",
+        "group": "AJZ＆BLU",
+        "title": "God Eater",
         "season": null,
-        "episode": 7,
+        "episode": 2,
         "resolution": null,
-        "source": null,
+        "source": "BIG5",
         "special": null
       },
       "pred": {
-        "group": null,
-        "title": "Mary_E_Il_Giardino_Segreto",
+        "group": "AJZ＆BLU",
+        "title": "God Eater",
         "season": null,
-        "episode": 7,
+        "episode": 5,
         "resolution": null,
-        "source": "Dvd",
+        "source": "BIG5",
         "special": null
       }
     },
     {
-      "filename": "(アニメ) アイドル伝説えり子 第24話 「心をつなぐ輪舞曲」 (DVD 640x480DivX5.02QB93 48kHz128kbps)",
+      "filename": "(アニメ) YAT安心！宇宙旅行 第1期 第07話 「サバイバル！野生のカネア」 (LD 640x480 WMV9 QB90 24fps)",
       "errors": {
-        "resolution": {
+        "season": {
           "gold": null,
-          "pred": "640x480"
+          "pred": "1"
         }
       },
       "gold": {
         "group": "アニメ",
-        "title": "アイドル伝説えり子",
+        "title": "YAT安心！宇宙旅行",
         "season": null,
-        "episode": 24,
-        "resolution": null,
-        "source": "DVD",
+        "episode": 7,
+        "resolution": "640x480",
+        "source": null,
         "special": null
       },
       "pred": {
         "group": "アニメ",
-        "title": "アイドル伝説えり子",
-        "season": null,
-        "episode": 24,
+        "title": "YAT安心！宇宙旅行",
+        "season": 1,
+        "episode": 7,
         "resolution": "640x480",
-        "source": "DVD",
+        "source": null,
         "special": null
       }
     },
     {
-      "filename": "[DMG] 東京レイヴンズ 第06話「days in nest -休日-」 [BDRip][AVC_AAC][720P][CHS](A8161323)",
+      "filename": "Lord El-Melloi II-sei no Jikenbo 06 [1AAC021C]",
       "errors": {
-        "episode": {
-          "gold": "1323",
-          "pred": "6"
+        "source": {
+          "gold": "aac",
+          "pred": null
         }
       },
       "gold": {
-        "group": "DMG",
-        "title": "東京レイヴンズ 第06話「days in nest -休日-」",
+        "group": null,
+        "title": "Lord El-Melloi II-sei no Jikenbo",
         "season": null,
-        "episode": 1323,
-        "resolution": "720P",
-        "source": "BDRip",
+        "episode": 6,
+        "resolution": null,
+        "source": "AAC",
         "special": null
       },
       "pred": {
-        "group": "DMG",
-        "title": "東京レイヴン��� 第06話「days in nest -休日-」",
+        "group": null,
+        "title": "Lord El-Melloi II-sei no Jikenbo",
         "season": null,
         "episode": 6,
-        "resolution": "720P",
-        "source": "BDRip",
+        "resolution": null,
+        "source": null,
         "special": null
       }
     },
     {
-      "filename": "[S1YURICON] Necronomico no Cosmic Horror Show[05v2][1080p][WebRip][AVC_AAC][CHS]",
+      "filename": "[Skymoon-Raws] Mashle 2nd Season - 01(13) [ViuTV][WEB-DL][1080p][AVC AAC]",
       "errors": {
+        "title": {
+          "gold": "mashle 2nd season - 01",
+          "pred": "mashle 2nd season"
+        },
         "season": {
-          "gold": null,
+          "gold": "2",
           "pred": "1"
         }
       },
       "gold": {
-        "group": "S1YURICON",
-        "title": "Necronomico no Cosmic Horror Show",
-        "season": null,
-        "episode": 5,
+        "group": "Skymoon-Raws",
+        "title": "Mashle 2nd Season - 01",
+        "season": 2,
+        "episode": 13,
         "resolution": "1080p",
-        "source": "WebRip",
+        "source": "WEB-DL",
         "special": null
       },
       "pred": {
-        "group": "S1YURICON",
-        "title": "Necronomico no Cosmic Horror Show",
+        "group": "Skymoon-Raws",
+        "title": "Mashle 2nd Season",
         "season": 1,
-        "episode": 5,
+        "episode": 13,
         "resolution": "1080p",
-        "source": "WebRip",
+        "source": "WEB-DL",
         "special": null
       }
     },
     {
-      "filename": "Cardcaptor Sakura - 17 [x264-AAC-BD1440x1080p][Sakura][C-W][E2B50799]",
+      "filename": "【CXRAW】【S17】【Power Rangers RPM】【30】【End Game】【x264 Hi10p AAC】【MP4】",
       "errors": {
-        "resolution": {
-          "gold": null,
-          "pred": "1080p"
-        },
-        "source": {
+        "season": {
           "gold": null,
-          "pred": "e2b50799"
+          "pred": "17"
         }
       },
       "gold": {
-        "group": null,
-        "title": "Cardcaptor Sakura",
+        "group": "CXRAW",
+        "title": "S17",
         "season": null,
-        "episode": 17,
+        "episode": 30,
         "resolution": null,
-        "source": null,
+        "source": "AAC",
         "special": null
       },
       "pred": {
-        "group": null,
-        "title": "Cardcaptor Sakura",
-        "season": null,
-        "episode": 17,
-        "resolution": "1080p",
-        "source": "E2B50799",
+        "group": "CXRAW",
+        "title": "S17",
+        "season": 17,
+        "episode": 30,
+        "resolution": null,
+        "source": "AAC",
         "special": null
       }
     },
     {
-      "filename": "[Xspitfire911] Tate no Yuusha no Nariagari S01E20 BDRIP 1080p X265 10bit VOSTFR",
+      "filename": "(アニメ) YAT安心！宇宙旅行 第1期 第24話 「モーレツ！かあちゃん珍道中」 (LD 640x480 WMV9 QB90 24fps)",
       "errors": {
         "season": {
           "gold": null,
@@ -515,80 +515,49 @@
         }
       },
       "gold": {
-        "group": "Xspitfire911",
-        "title": "Tate no Yuusha no Nariagari",
+        "group": "アニメ",
+        "title": "YAT安心！宇宙旅行",
         "season": null,
-        "episode": 20,
-        "resolution": "1080p",
-        "source": "BDRIP",
+        "episode": 24,
+        "resolution": "640x480",
+        "source": null,
         "special": null
       },
       "pred": {
-        "group": "Xspitfire911",
-        "title": "Tate no Yuusha no Nariagari",
+        "group": "アニメ",
+        "title": "YAT安心！宇宙旅行",
         "season": 1,
-        "episode": 20,
-        "resolution": "1080p",
-        "source": "BDRIP",
+        "episode": 24,
+        "resolution": "640x480",
+        "source": null,
         "special": null
       }
     },
     {
-      "filename": "[KTXP][Dungeon ni Deai wo Motomeru no wa Machigatteiru Darou ka IV][13][BIG5][720P][MP4]",
+      "filename": "[Snow-Raws] アイドルマスター シンデレラガールズ劇場 第2期 SP17 (DVD 1280x720 HEVC-YUV420P10 FLAC)",
       "errors": {
-        "title": {
-          "gold": "dungeon ni deai wo motomeru no wa machigatteiru darou ka",
-          "pred": "dungeon ni deai wo motomeru no wa machigatteiru darou ka iv"
-        },
         "season": {
-          "gold": "4",
-          "pred": null
-        }
-      },
-      "gold": {
-        "group": "KTXP",
-        "title": "Dungeon ni Deai wo Motomeru no wa Machigatteiru Darou ka",
-        "season": 4,
-        "episode": 13,
-        "resolution": "720P",
-        "source": "BIG5",
-        "special": null
-      },
-      "pred": {
-        "group": "KTXP",
-        "title": "Dungeon ni Deai wo Motomeru no wa Machigatteiru Darou ka IV",
-        "season": null,
-        "episode": 13,
-        "resolution": "720P",
-        "source": "BIG5",
-        "special": null
-      }
-    },
-    {
-      "filename": "[JyFanSub][Fate_Apocrypha][15][GB][1080]p",
-      "errors": {
-        "episode": {
-          "gold": "1080",
-          "pred": "15"
+          "gold": null,
+          "pred": "2"
         }
       },
       "gold": {
-        "group": "JyFanSub",
-        "title": "Fate_Apocrypha",
+        "group": "Snow-Raws",
+        "title": "アイドルマスター シンデレラガールズ劇場 第2期 SP17",
         "season": null,
-        "episode": 1080,
-        "resolution": null,
-        "source": "GB",
-        "special": null
+        "episode": 17,
+        "resolution": "1280x720",
+        "source": "DVD",
+        "special": "SP"
       },
       "pred": {
-        "group": "JyFanSub",
-        "title": "Fate_Apocrypha",
-        "season": null,
-        "episode": 15,
-        "resolution": null,
-        "source": "GB",
-        "special": null
+        "group": "Snow-Raws",
+        "title": "アイドルマスター シンデレラガールズ劇場 第2期 SP17",
+        "season": 2,
+        "episode": 17,
+        "resolution": "1280x720",
+        "source": "DVD",
+        "special": "SP"
       }
     }
   ]

run_metadata.json

@@ -1,5 +1,5 @@
 {
-  "experiment_name": "dmhy-char-full-relabel",
+  "experiment_name": "dmhy-char-guoman-relabel",
   "data_file": "datasets/AnimeName/dmhy_weak_char.jsonl",
   "tokenizer_variant": "char",
   "vocab_file": "datasets/AnimeName/vocab.char.json",
@@ -15,7 +15,7 @@
   "batch_size": 256,
   "learning_rate": 8e-05,
   "warmup_steps": 300,
-  "seed": 48,
+  "seed": 52,
   "device": "cuda",
   "fp16": true,
   "gradient_accumulation_steps": 1,

trainer_eval_metrics.json

@@ -1,11 +1,11 @@
 {
-  "eval_loss": 0.01631847210228443,
-  "eval_precision": 0.9799749533444652,
-  "eval_recall": 0.986698478236683,
-  "eval_f1": 0.9833252228334185,
-  "eval_accuracy": 0.9943065860243627,
-  "eval_runtime": 39.3604,
-  "eval_samples_per_second": 321.161,
-  "eval_steps_per_second": 1.27,
+  "eval_loss": 0.005763721186667681,
+  "eval_precision": 0.9921522239605195,
+  "eval_recall": 0.9946191314105016,
+  "eval_f1": 0.9933841461473317,
+  "eval_accuracy": 0.9980711558885925,
+  "eval_runtime": 45.558,
+  "eval_samples_per_second": 277.471,
+  "eval_steps_per_second": 1.098,
   "epoch": 2.0
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b5aa0df615ce731796aa9934b0505e00a685611be134c071d7b2487d8112dde1
+oid sha256:f01503ec029ec161063c2d78a00732c80072525b8d258c7c717b2e21f4f55d93
 size 5265