Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -74,10 +74,14 @@ else:
|
|
74 |
|
75 |
@st.cache_resource # 👈 Add the caching decorator
|
76 |
def load_model(selected_language, model_name=None, entity_set=None):
|
77 |
-
#
|
|
|
|
|
|
|
|
|
|
|
78 |
with warnings.catch_warnings():
|
79 |
warnings.simplefilter("ignore")
|
80 |
-
|
81 |
try:
|
82 |
# This block handles the spaCy models for German and English
|
83 |
if selected_language == "German":
|
@@ -87,12 +91,9 @@ def load_model(selected_language, model_name=None, entity_set=None):
|
|
87 |
st.info("Downloading German language model... This may take a moment.")
|
88 |
spacy.cli.download("de_core_news_lg")
|
89 |
nlp_model_de = spacy.load("de_core_news_lg")
|
90 |
-
|
91 |
if "entityfishing" not in nlp_model_de.pipe_names:
|
92 |
-
try:
|
93 |
-
|
94 |
-
except Exception as e:
|
95 |
-
st.warning(f"Entity-fishing not available, using basic NER only: {e}")
|
96 |
return nlp_model_de
|
97 |
|
98 |
elif selected_language == "English - spaCy":
|
@@ -102,52 +103,58 @@ def load_model(selected_language, model_name=None, entity_set=None):
|
|
102 |
st.info("Downloading English language model... This may take a moment.")
|
103 |
spacy.cli.download("en_core_web_sm")
|
104 |
nlp_model_en = spacy.load("en_core_web_sm")
|
105 |
-
|
106 |
if "entityfishing" not in nlp_model_en.pipe_names:
|
107 |
-
try:
|
108 |
-
|
109 |
-
except Exception as e:
|
110 |
-
st.warning(f"Entity-fishing not available, using basic NER only: {e}")
|
111 |
return nlp_model_en
|
112 |
|
113 |
# This block handles the ReFinED model and the "add_special_tokens" error
|
114 |
else:
|
115 |
try:
|
116 |
-
# First, attempt to load the model as usual
|
117 |
return Refined.from_pretrained(model_name=model_name, entity_set=entity_set)
|
118 |
|
119 |
except Exception as e:
|
120 |
-
# If the specific "add_special_tokens" error occurs, apply the fix
|
121 |
if "add_special_tokens" in str(e):
|
122 |
-
st.warning("Conflict detected. Applying fix by
|
123 |
-
|
124 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
local_model_path = f"./{model_name}-{entity_set}-fixed"
|
126 |
|
127 |
-
# Download tokenizer
|
128 |
-
|
129 |
-
tokenizer.
|
|
|
130 |
|
|
|
|
|
|
|
|
|
|
|
|
|
131 |
config_path = os.path.join(local_model_path, "tokenizer_config.json")
|
132 |
with open(config_path, "r") as f:
|
133 |
config_data = json.load(f)
|
134 |
|
135 |
-
# Remove the conflicting
|
136 |
-
config_data.pop("add_special_tokens", None)
|
137 |
|
138 |
with open(config_path, "w") as f:
|
139 |
json.dump(config_data, f, indent=2)
|
140 |
|
141 |
-
#
|
142 |
-
st.success("
|
143 |
return Refined.from_pretrained(model_name=local_model_path, entity_set=entity_set)
|
144 |
|
145 |
else:
|
146 |
-
# If it's a different error,
|
147 |
-
raise e
|
148 |
|
149 |
except Exception as e:
|
150 |
-
st.error(f"
|
151 |
return None
|
152 |
|
153 |
# Use the cached model
|
|
|
74 |
|
75 |
@st.cache_resource # 👈 Add the caching decorator
|
76 |
def load_model(selected_language, model_name=None, entity_set=None):
|
77 |
+
# This dictionary maps the easy names to their full Hugging Face Hub IDs
|
78 |
+
model_mapping = {
|
79 |
+
"aida_model": "amazon-science/ReFinED-aida-model",
|
80 |
+
"wikipedia_model_with_numbers": "amazon-science/ReFinED-wikipedia-model"
|
81 |
+
}
|
82 |
+
|
83 |
with warnings.catch_warnings():
|
84 |
warnings.simplefilter("ignore")
|
|
|
85 |
try:
|
86 |
# This block handles the spaCy models for German and English
|
87 |
if selected_language == "German":
|
|
|
91 |
st.info("Downloading German language model... This may take a moment.")
|
92 |
spacy.cli.download("de_core_news_lg")
|
93 |
nlp_model_de = spacy.load("de_core_news_lg")
|
|
|
94 |
if "entityfishing" not in nlp_model_de.pipe_names:
|
95 |
+
try: nlp_model_de.add_pipe("entityfishing")
|
96 |
+
except Exception as e: st.warning(f"Entity-fishing not available: {e}")
|
|
|
|
|
97 |
return nlp_model_de
|
98 |
|
99 |
elif selected_language == "English - spaCy":
|
|
|
103 |
st.info("Downloading English language model... This may take a moment.")
|
104 |
spacy.cli.download("en_core_web_sm")
|
105 |
nlp_model_en = spacy.load("en_core_web_sm")
|
|
|
106 |
if "entityfishing" not in nlp_model_en.pipe_names:
|
107 |
+
try: nlp_model_en.add_pipe("entityfishing")
|
108 |
+
except Exception as e: st.warning(f"Entity-fishing not available: {e}")
|
|
|
|
|
109 |
return nlp_model_en
|
110 |
|
111 |
# This block handles the ReFinED model and the "add_special_tokens" error
|
112 |
else:
|
113 |
try:
|
|
|
114 |
return Refined.from_pretrained(model_name=model_name, entity_set=entity_set)
|
115 |
|
116 |
except Exception as e:
|
|
|
117 |
if "add_special_tokens" in str(e):
|
118 |
+
st.warning("Conflict detected. Applying fix by downloading and patching model...")
|
119 |
+
|
120 |
+
# 1. Get the REAL model name from our mapping
|
121 |
+
real_model_name = model_mapping.get(model_name)
|
122 |
+
if not real_model_name:
|
123 |
+
st.error(f"Unknown model alias: {model_name}")
|
124 |
+
return None
|
125 |
+
|
126 |
+
# 2. Define a local path to save the fixed model
|
127 |
local_model_path = f"./{model_name}-{entity_set}-fixed"
|
128 |
|
129 |
+
# 3. Download the tokenizer and the model using the REAL name
|
130 |
+
st.info(f"Downloading model files for {real_model_name}...")
|
131 |
+
tokenizer = AutoTokenizer.from_pretrained(real_model_name)
|
132 |
+
model_files = AutoModelForSeq2SeqLM.from_pretrained(real_model_name)
|
133 |
|
134 |
+
# 4. Save them to the local directory
|
135 |
+
tokenizer.save_pretrained(local_model_path)
|
136 |
+
model_files.save_pretrained(local_model_path)
|
137 |
+
st.info("Model files downloaded.")
|
138 |
+
|
139 |
+
# 5. Patch the tokenizer config file
|
140 |
config_path = os.path.join(local_model_path, "tokenizer_config.json")
|
141 |
with open(config_path, "r") as f:
|
142 |
config_data = json.load(f)
|
143 |
|
144 |
+
config_data.pop("add_special_tokens", None) # Remove the conflicting key
|
|
|
145 |
|
146 |
with open(config_path, "w") as f:
|
147 |
json.dump(config_data, f, indent=2)
|
148 |
|
149 |
+
# 6. Load the model from the local, fixed path
|
150 |
+
st.success("Patch applied. Loading model from local cache...")
|
151 |
return Refined.from_pretrained(model_name=local_model_path, entity_set=entity_set)
|
152 |
|
153 |
else:
|
154 |
+
raise e # If it's a different error, we still want to see it
|
|
|
155 |
|
156 |
except Exception as e:
|
157 |
+
st.error(f"Failed to load model. Error: {e}")
|
158 |
return None
|
159 |
|
160 |
# Use the cached model
|