Spaces:
Running
Running
Update zeroshot.py
Browse files- zeroshot.py +9 -3
zeroshot.py
CHANGED
|
@@ -133,12 +133,16 @@ def process(
|
|
| 133 |
# print("after uroman")
|
| 134 |
# print("len lexicon", len(lexicon))
|
| 135 |
with tempfile.NamedTemporaryFile() as lexicon_file:
|
| 136 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 137 |
with open(lexicon_file.name, "w") as f:
|
| 138 |
idx = 10
|
| 139 |
for word, spelling in lexicon.items():
|
| 140 |
f.write(word + " " + spelling + "\n")
|
| 141 |
-
if idx %
|
| 142 |
print(word, spelling, flush=True)
|
| 143 |
idx += 1
|
| 144 |
|
|
@@ -150,11 +154,13 @@ def process(
|
|
| 150 |
)
|
| 151 |
if lmscore_usedefault:
|
| 152 |
lmscore = LM_SCORE_DEFAULT if lm_path is not None else 0
|
|
|
|
|
|
|
| 153 |
|
| 154 |
beam_search_decoder = ctc_decoder(
|
| 155 |
lexicon=lexicon_file.name,
|
| 156 |
tokens=token_file,
|
| 157 |
-
lm=
|
| 158 |
nbest=1,
|
| 159 |
beam_size=500,
|
| 160 |
beam_size_token=50,
|
|
|
|
| 133 |
# print("after uroman")
|
| 134 |
# print("len lexicon", len(lexicon))
|
| 135 |
with tempfile.NamedTemporaryFile() as lexicon_file:
|
| 136 |
+
print("lm_path before", lm_path)
|
| 137 |
+
if lm_path is not None and not lm_path.strip():
|
| 138 |
+
lm_path = None
|
| 139 |
+
print("lm_path after", lm_path)
|
| 140 |
+
|
| 141 |
with open(lexicon_file.name, "w") as f:
|
| 142 |
idx = 10
|
| 143 |
for word, spelling in lexicon.items():
|
| 144 |
f.write(word + " " + spelling + "\n")
|
| 145 |
+
if idx % 250 == 0:
|
| 146 |
print(word, spelling, flush=True)
|
| 147 |
idx += 1
|
| 148 |
|
|
|
|
| 154 |
)
|
| 155 |
if lmscore_usedefault:
|
| 156 |
lmscore = LM_SCORE_DEFAULT if lm_path is not None else 0
|
| 157 |
+
print("using word score", wscore)
|
| 158 |
+
print("using lm score", lmscore)
|
| 159 |
|
| 160 |
beam_search_decoder = ctc_decoder(
|
| 161 |
lexicon=lexicon_file.name,
|
| 162 |
tokens=token_file,
|
| 163 |
+
lm=lm_path,
|
| 164 |
nbest=1,
|
| 165 |
beam_size=500,
|
| 166 |
beam_size_token=50,
|