Spaces:
Build error
Build error
Commit ยท
85c2795
1
Parent(s): 40607dc
change to conditional kogpt trintiy project
Browse files
app.py
CHANGED
|
@@ -47,12 +47,13 @@ def infer(input_ids, max_length, temperature, top_k, top_p):
|
|
| 47 |
|
| 48 |
|
| 49 |
# prompts
|
| 50 |
-
st.title("
|
| 51 |
-
st.write("
|
|
|
|
| 52 |
|
| 53 |
# text and sidebars
|
| 54 |
-
default_value = "
|
| 55 |
-
sent = st.text_area("Text", default_value, max_chars=
|
| 56 |
max_length = st.sidebar.slider("์์ฑ ๋ฌธ์ฅ ๊ธธ์ด๋ฅผ ์ ํํด์ฃผ์ธ์!", min_value=42, max_value=64)
|
| 57 |
temperature = st.sidebar.slider(
|
| 58 |
"Temperature", value=1.0, min_value=0.0, max_value=1.0, step=0.05
|
|
@@ -93,67 +94,25 @@ def infer_sentence(
|
|
| 93 |
generated_sequence = output_sequences[0]
|
| 94 |
print(generated_sequence)
|
| 95 |
|
| 96 |
-
# print(f"=== GENERATED SEQUENCE {generated_sequence_idx + 1} ===")
|
| 97 |
-
# generated_sequences = generated_sequence.tolist()
|
| 98 |
# Decode text
|
| 99 |
text = tokenizer.decode(generated_sequence, clean_up_tokenization_spaces=True)
|
| 100 |
print(text)
|
| 101 |
-
|
|
|
|
| 102 |
stop_token = tokenizer.pad_token
|
| 103 |
print(stop_token)
|
| 104 |
text = text[: text.find(stop_token) if stop_token else None]
|
| 105 |
print(text)
|
| 106 |
-
|
|
|
|
| 107 |
condition_index = find_nth(text, "๋ฌธ์ฅ์ด๋ค", 2)
|
| 108 |
text = text[condition_index + 5 :]
|
| 109 |
text = text.strip()
|
| 110 |
return text
|
| 111 |
|
| 112 |
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
list_samhaengshi = []
|
| 116 |
-
|
| 117 |
-
# initializing text and index for iteration purpose
|
| 118 |
-
index = 0
|
| 119 |
-
|
| 120 |
-
# iterating over the input letter string
|
| 121 |
-
for index, letter_item in enumerate(input_letter):
|
| 122 |
-
# initializing the input_letter
|
| 123 |
-
if index == 0:
|
| 124 |
-
residual_text = letter_item
|
| 125 |
-
# print('residual_text:', residual_text)
|
| 126 |
-
|
| 127 |
-
# infer and add to the output
|
| 128 |
-
conditional_input = f"{condition_sentence} {residual_text}"
|
| 129 |
-
inferred_sentence = infer_sentence(conditional_input, tokenizer)
|
| 130 |
-
if index != 0:
|
| 131 |
-
# remove previous sentence from the output
|
| 132 |
-
print("inferred_sentence:", inferred_sentence)
|
| 133 |
-
inferred_sentence = inferred_sentence.replace(
|
| 134 |
-
list_samhaengshi[index - 1], ""
|
| 135 |
-
).strip()
|
| 136 |
-
else:
|
| 137 |
-
pass
|
| 138 |
-
list_samhaengshi.append(inferred_sentence)
|
| 139 |
-
|
| 140 |
-
# until the end of the input_letter, give the previous residual_text to the next iteration
|
| 141 |
-
if index < len(input_letter) - 1:
|
| 142 |
-
residual_sentence = list_samhaengshi[index]
|
| 143 |
-
next_letter = input_letter[index + 1]
|
| 144 |
-
residual_text = (
|
| 145 |
-
f"{residual_sentence} {next_letter}" # previous sentence + next letter
|
| 146 |
-
)
|
| 147 |
-
print("residual_text", residual_text)
|
| 148 |
-
|
| 149 |
-
elif index == len(input_letter) - 1: # end of the input_letter
|
| 150 |
-
# Concatenate strings in the list without intersection
|
| 151 |
-
|
| 152 |
-
return list_samhaengshi
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
return_text = make_residual_conditional_samhaengshi(
|
| 156 |
-
input_letter=sent, condition_sentence=condition_sentence
|
| 157 |
)
|
| 158 |
|
| 159 |
print(return_text)
|
|
|
|
| 47 |
|
| 48 |
|
| 49 |
# prompts
|
| 50 |
+
st.title("์ฃผ์ด์ง ๊ฐ์ ์ ๋ง๊ฒ ๋ฌธ์ฅ์ ๋ง๋๋ KoGPT์
๋๋ค ๐ฆ")
|
| 51 |
+
st.write("์ข์ธก์ ๊ฐ์ ์ํ์ ๋ณํ๋ฅผ ์ค๋ณด์ธ์.")
|
| 52 |
+
st.write("์
๋ ฅํ๊ณ ๋์ CTRL+Enter(CMD+Enter)๋ฅผ ๋๋ฅด์ธ์ ๐ค")
|
| 53 |
|
| 54 |
# text and sidebars
|
| 55 |
+
default_value = "์์ํ ๋ฐค๋ค์ด ๊ณ์๋๋ ๋ ์ธ์ ๊ฐ๋ถํฐ ๋๋"
|
| 56 |
+
sent = st.text_area("Text", default_value, max_chars=30, height=275)
|
| 57 |
max_length = st.sidebar.slider("์์ฑ ๋ฌธ์ฅ ๊ธธ์ด๋ฅผ ์ ํํด์ฃผ์ธ์!", min_value=42, max_value=64)
|
| 58 |
temperature = st.sidebar.slider(
|
| 59 |
"Temperature", value=1.0, min_value=0.0, max_value=1.0, step=0.05
|
|
|
|
| 94 |
generated_sequence = output_sequences[0]
|
| 95 |
print(generated_sequence)
|
| 96 |
|
|
|
|
|
|
|
| 97 |
# Decode text
|
| 98 |
text = tokenizer.decode(generated_sequence, clean_up_tokenization_spaces=True)
|
| 99 |
print(text)
|
| 100 |
+
|
| 101 |
+
# Remove all text after the pad token
|
| 102 |
stop_token = tokenizer.pad_token
|
| 103 |
print(stop_token)
|
| 104 |
text = text[: text.find(stop_token) if stop_token else None]
|
| 105 |
print(text)
|
| 106 |
+
|
| 107 |
+
# Remove condition sentence
|
| 108 |
condition_index = find_nth(text, "๋ฌธ์ฅ์ด๋ค", 2)
|
| 109 |
text = text[condition_index + 5 :]
|
| 110 |
text = text.strip()
|
| 111 |
return text
|
| 112 |
|
| 113 |
|
| 114 |
+
return_text = infer_sentence(
|
| 115 |
+
condition_plus_input=condition_plus_input, tokenizer=tokenizer
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 116 |
)
|
| 117 |
|
| 118 |
print(return_text)
|