Spaces:
Sleeping
Sleeping
Luis J Camargo commited on
Commit ·
cc3e2aa
1
Parent(s): 32bc284
fix(ui): official catalogue names, drop redundant meta fallbacks
Browse files- Display names now come from the catalogue: inali_name for the Spanish
UI, english_name for the English UI, falling back to the model's label
and then the raw code. NAMES[code] carries both.
- META_FALLBACK trimmed to hue and nah, the only two model classes the
catalogue does not contain. The other eight entries were duplicates of
catalogue values, and the stp one ('Tepehuano') contradicted the
catalogue ('Tepehuan').
- Locale is read from the Accept-Language header, since Gradio resolves
i18n client-side only.
app.py
CHANGED
|
@@ -68,6 +68,7 @@ N_CLASSES = len(CLASSES)
|
|
| 68 |
try:
|
| 69 |
with open("languages.json", "r", encoding="utf-8") as f:
|
| 70 |
_languages_data = json.load(f)
|
|
|
|
| 71 |
CODE_TO_NAME = {
|
| 72 |
item.get("code"): item.get("inali_name")
|
| 73 |
for item in _languages_data
|
|
@@ -80,22 +81,17 @@ try:
|
|
| 80 |
}
|
| 81 |
except Exception as e:
|
| 82 |
print(f"Warning: Could not load languages.json: {e}")
|
|
|
|
| 83 |
CODE_TO_NAME = {}
|
| 84 |
CODE_TO_META = {}
|
| 85 |
|
| 86 |
-
#
|
| 87 |
-
#
|
|
|
|
|
|
|
| 88 |
META_FALLBACK = {
|
| 89 |
-
"chq": ("Otomangue", "Chinanteco"),
|
| 90 |
"hue": ("Huave", "Huave"),
|
| 91 |
-
"mir": ("Mixe-Zoqueano", "Mixe"),
|
| 92 |
"nah": ("Yuto-Nahua", "Náhuatl"),
|
| 93 |
-
"ots": ("Otomangue", "Otomí"),
|
| 94 |
-
"pua": ("Purépecha", "Purépecha"),
|
| 95 |
-
"stp": ("Yuto-Nahua", "Tepehuano"),
|
| 96 |
-
"zad": ("Otomangue", "Zapoteco"),
|
| 97 |
-
"zar": ("Otomangue", "Zapoteco"),
|
| 98 |
-
"zpd": ("Otomangue", "Zapoteco"),
|
| 99 |
}
|
| 100 |
|
| 101 |
|
|
@@ -118,12 +114,18 @@ def _meta(code):
|
|
| 118 |
|
| 119 |
SUPERLANGUAGE_OF = {}
|
| 120 |
FAMILY_OF = {}
|
| 121 |
-
|
| 122 |
for _code in CLASSES:
|
| 123 |
_super, _family, _name = _meta(_code)
|
| 124 |
SUPERLANGUAGE_OF[_code] = _super
|
| 125 |
FAMILY_OF[_code] = _family
|
| 126 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 127 |
|
| 128 |
print(
|
| 129 |
f"Language-ID model loaded: {len(CLASSES)} codes, "
|
|
@@ -165,11 +167,26 @@ RESULT_STYLES = """
|
|
| 165 |
"""
|
| 166 |
|
| 167 |
|
| 168 |
-
def
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 169 |
"""Render ranked predictions as pills. `rows` is [(code, probability), ...]."""
|
| 170 |
if not rows:
|
| 171 |
return ""
|
| 172 |
|
|
|
|
| 173 |
blocks = []
|
| 174 |
for index, (code, prob) in enumerate(rows):
|
| 175 |
css = "tw-row tw-first" if index == 0 else "tw-row"
|
|
@@ -177,7 +194,7 @@ def render_results(rows):
|
|
| 177 |
f'<div class="{css}">'
|
| 178 |
f'<div class="tw-main">'
|
| 179 |
f'<div class="tw-title">'
|
| 180 |
-
f'<span class="tw-name">{html.escape(
|
| 181 |
f'<span class="tw-code">{html.escape(code)}</span>'
|
| 182 |
f'</div>'
|
| 183 |
f'<div class="tw-meta">{html.escape(SUPERLANGUAGE_OF[code])} · '
|
|
@@ -190,7 +207,7 @@ def render_results(rows):
|
|
| 190 |
|
| 191 |
|
| 192 |
# === INFERENCE FUNCTION ===
|
| 193 |
-
def predict_language(audio_path, code_k=3, code_thresh=0.0):
|
| 194 |
if not audio_path:
|
| 195 |
raise gr.Error("No audio provided! Please upload or record an audio file.")
|
| 196 |
|
|
@@ -229,7 +246,7 @@ def predict_language(audio_path, code_k=3, code_thresh=0.0):
|
|
| 229 |
print(f"[LOG] Final Memory: {get_mem_usage():.2f} MB")
|
| 230 |
print("--- [LOG] Request Finished ---\n")
|
| 231 |
|
| 232 |
-
return render_results(rows)
|
| 233 |
except gr.Error:
|
| 234 |
raise
|
| 235 |
except Exception as e:
|
|
|
|
| 68 |
try:
|
| 69 |
with open("languages.json", "r", encoding="utf-8") as f:
|
| 70 |
_languages_data = json.load(f)
|
| 71 |
+
CATALOGUE = {item.get("code"): item for item in _languages_data if item.get("code")}
|
| 72 |
CODE_TO_NAME = {
|
| 73 |
item.get("code"): item.get("inali_name")
|
| 74 |
for item in _languages_data
|
|
|
|
| 81 |
}
|
| 82 |
except Exception as e:
|
| 83 |
print(f"Warning: Could not load languages.json: {e}")
|
| 84 |
+
CATALOGUE = {}
|
| 85 |
CODE_TO_NAME = {}
|
| 86 |
CODE_TO_META = {}
|
| 87 |
|
| 88 |
+
# Only two codes are absent from the catalogue: `nah` (Náhuatl) and `hue` (Huave).
|
| 89 |
+
# They are macrolanguage placeholders the dataset kept as a fallback for variants that could
|
| 90 |
+
# not be resolved. The model is deterministic and never emits them as an "unknown" bucket,
|
| 91 |
+
# so they exist purely so those two classes still have a family/superlanguage to display.
|
| 92 |
META_FALLBACK = {
|
|
|
|
| 93 |
"hue": ("Huave", "Huave"),
|
|
|
|
| 94 |
"nah": ("Yuto-Nahua", "Náhuatl"),
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 95 |
}
|
| 96 |
|
| 97 |
|
|
|
|
| 114 |
|
| 115 |
SUPERLANGUAGE_OF = {}
|
| 116 |
FAMILY_OF = {}
|
| 117 |
+
NAMES = {}
|
| 118 |
for _code in CLASSES:
|
| 119 |
_super, _family, _name = _meta(_code)
|
| 120 |
SUPERLANGUAGE_OF[_code] = _super
|
| 121 |
FAMILY_OF[_code] = _family
|
| 122 |
+
# Official presentation name: INALI name for the Spanish UI, English name for the English
|
| 123 |
+
# UI, falling back to the model's own label and finally to the raw code.
|
| 124 |
+
_entry = CATALOGUE.get(_code, {})
|
| 125 |
+
_fallback = _name or _code
|
| 126 |
+
_es = (_entry.get("inali_name") or "").strip() or _fallback
|
| 127 |
+
_en = (_entry.get("english_name") or "").strip() or _es
|
| 128 |
+
NAMES[_code] = {"es": _es, "en": _en}
|
| 129 |
|
| 130 |
print(
|
| 131 |
f"Language-ID model loaded: {len(CLASSES)} codes, "
|
|
|
|
| 167 |
"""
|
| 168 |
|
| 169 |
|
| 170 |
+
def _ui_locale(request):
|
| 171 |
+
"""Best-effort UI locale.
|
| 172 |
+
|
| 173 |
+
Gradio resolves i18n on the frontend from the browser's locale, so the browser header is
|
| 174 |
+
the faithful mirror of the default UI language; the in-app locale switch is client-side
|
| 175 |
+
and is not visible to this process.
|
| 176 |
+
"""
|
| 177 |
+
raw = ""
|
| 178 |
+
if request is not None:
|
| 179 |
+
raw = (request.headers.get("accept-language") or "")
|
| 180 |
+
first = raw.split(",")[0].strip().lower()
|
| 181 |
+
return "en" if first.startswith("en") else "es"
|
| 182 |
+
|
| 183 |
+
|
| 184 |
+
def render_results(rows, locale="es"):
|
| 185 |
"""Render ranked predictions as pills. `rows` is [(code, probability), ...]."""
|
| 186 |
if not rows:
|
| 187 |
return ""
|
| 188 |
|
| 189 |
+
lang = "en" if locale == "en" else "es"
|
| 190 |
blocks = []
|
| 191 |
for index, (code, prob) in enumerate(rows):
|
| 192 |
css = "tw-row tw-first" if index == 0 else "tw-row"
|
|
|
|
| 194 |
f'<div class="{css}">'
|
| 195 |
f'<div class="tw-main">'
|
| 196 |
f'<div class="tw-title">'
|
| 197 |
+
f'<span class="tw-name">{html.escape(NAMES[code][lang])}</span>'
|
| 198 |
f'<span class="tw-code">{html.escape(code)}</span>'
|
| 199 |
f'</div>'
|
| 200 |
f'<div class="tw-meta">{html.escape(SUPERLANGUAGE_OF[code])} · '
|
|
|
|
| 207 |
|
| 208 |
|
| 209 |
# === INFERENCE FUNCTION ===
|
| 210 |
+
def predict_language(audio_path, code_k=3, code_thresh=0.0, request: gr.Request = None):
|
| 211 |
if not audio_path:
|
| 212 |
raise gr.Error("No audio provided! Please upload or record an audio file.")
|
| 213 |
|
|
|
|
| 246 |
print(f"[LOG] Final Memory: {get_mem_usage():.2f} MB")
|
| 247 |
print("--- [LOG] Request Finished ---\n")
|
| 248 |
|
| 249 |
+
return render_results(rows, _ui_locale(request))
|
| 250 |
except gr.Error:
|
| 251 |
raise
|
| 252 |
except Exception as e:
|