Update app.py

app.py

@@ -1,120 +1,16 @@
-# import sentencepiece as spm
-# import numpy as np
-# import tensorflow as tf
-# from tensorflow.keras.preprocessing.sequence import pad_sequences
-# from valx import detect_profanity, detect_hate_speech
-# import gradio as gr
-
-# sp = spm.SentencePieceProcessor()
-# sp.Load("dungen_dev_preview.model")
-
-# model = tf.keras.models.load_model("dungen_dev_preview_model.keras")
-
-# max_seq_len = 25
-
-# def generate_text(seed_text, next_words=30, temperature=0.5):
-#     seed_text = seed_text.strip().lower()
-
-#     if "|" in seed_text:
-#         gr.Warning("The prompt should not contain the '|' character. Using default prompt.")
-#         seed_text = 'game name | '
-#     elif detect_profanity([seed_text], language='All'):
-#         gr.Warning("Profanity detected in the prompt, using the default prompt.")
-#         seed_text = 'game name | '
-#     elif (hate_speech_result := detect_hate_speech(seed_text)) and hate_speech_result[0] in ['Hate Speech', 'Offensive Speech']:
-#         gr.Warning('Harmful speech detected in the prompt, using default prompt.')
-#         seed_text = 'game name | '
-#     else:
-#         seed_text += ' | '
-
-#     generated_text = seed_text
-#     if generated_text != 'game name | ': # only generate if not the default prompt
-#         for _ in range(next_words):
-#             token_list = sp.encode_as_ids(generated_text)
-#             token_list = pad_sequences([token_list], maxlen=max_seq_len - 1, padding='pre')
-#             predicted = model.predict(token_list, verbose=0)[0]
-
-#             predicted = np.asarray(predicted).astype("float64")
-#             predicted = np.log(predicted + 1e-8) / temperature
-#             exp_preds = np.exp(predicted)
-#             predicted = exp_preds / np.sum(exp_preds)
-
-#             next_index = np.random.choice(len(predicted), p=predicted)
-#             next_token = sp.id_to_piece(next_index)
-#             generated_text += next_token
-
-#             if next_token.endswith('</s>') or next_token.endswith('<unk>'):
-#                 break
-
-#     decoded = sp.decode_pieces(sp.encode_as_pieces(generated_text))
-#     decoded = decoded.replace("</s>", "").replace("<unk>", "").strip()
-
-#     if '|' in decoded:
-#         decoded = decoded.split('|', 1)[1].strip()
-
-#     if any(detect_profanity([decoded], language='All')) or (hate_speech_result := detect_hate_speech(decoded)) and hate_speech_result[0] in ['Hate Speech', 'Offensive Speech']:
-#         gr.Warning("Flagged potentially harmful output.")
-#         decoded = 'Flagged Output'
-
-#     return decoded
-
-# demo = gr.Interface(
-#     fn=generate_text,
-#     inputs=[
-#         gr.Textbox(label="Prompt", value="a female character name", max_lines=1),
-#         gr.Slider(1, 100, step=1, label='Next Words', value=30),
-#         gr.Slider(0.1, 1, value=0.5, label='Temperature', info='Controls randomness of generation, higher values = more creative, lower values = more probalistic')
-#     ],
-#     outputs=gr.Textbox(label="Generated Names"),
-#     title='Dungen Dev - Name Generator',
-#     description='A prompt-based name generator for game developers. Dungen Dev is an experimental model, and may produce outputs that are inappropriate, biased, or potentially harmful and inaccurate. Caution is advised.',
-#     examples=[
-#         ["a male character name", 30, 0.5],
-#         ["a futuristic city name", 30, 0.5],
-#         ["an item name", 30, 0.5],
-#         ["a dark and mysterious forest name", 30, 0.5],
-#         ["an evil character name", 30, 0.5]
-#     ]
-# )
-
-# demo.launch()
 import sentencepiece as spm
 import numpy as np
 import tensorflow as tf
 from tensorflow.keras.preprocessing.sequence import pad_sequences
 from valx import detect_profanity, detect_hate_speech
 import gradio as gr
-import logging
-import csv
-import os
-from datetime import datetime
-from datasets import load_dataset, Dataset
 
-# Model and SentencePiece loading
 sp = spm.SentencePieceProcessor()
 sp.Load("dungen_dev_preview.model")
-model = tf.keras.models.load_model("dungen_dev_preview_model.keras")
-max_seq_len = 25
-
-logging.basicConfig(filename="app.log", level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
 
-# Dataset setup
-FLAGGED_DATASET_ID = "InfinitodeLTD/DungenDev-FlaggedOutputs"
-
-def load_or_create_dataset(dataset_id):
-    try:
-        dataset = load_dataset(dataset_id)
-        if "flagged_data" not in dataset:
-            raise ValueError("Dataset does not contain the 'flagged_data' config.")
-
-        return dataset["flagged_data"]
-    except (datasets.DatasetNotFoundError, ValueError) as e:
-        logging.warning(f"Dataset not found or incorrect schema: {e}. Creating a new dataset.")
-        dataset = Dataset.from_dict({"Timestamp": [], "Prompt": [], "Flagged Text": []})
-        dataset.push_to_hub(dataset_id, config_name="flagged_data") # important: config_name
-        return dataset
+model = tf.keras.models.load_model("dungen_dev_preview_model.keras")
 
-flagged_dataset = load_or_create_dataset(FLAGGED_DATASET_ID)
+max_seq_len = 25
 
 def generate_text(seed_text, next_words=30, temperature=0.5):
     seed_text = seed_text.strip().lower()
@@ -132,7 +28,7 @@ def generate_text(seed_text, next_words=30, temperature=0.5):
         seed_text += ' | '
 
     generated_text = seed_text
-    if generated_text != 'game name | ':
+    if generated_text != 'game name | ': # only generate if not the default prompt
         for _ in range(next_words):
             token_list = sp.encode_as_ids(generated_text)
             token_list = pad_sequences([token_list], maxlen=max_seq_len - 1, padding='pre')
@@ -162,49 +58,23 @@ def generate_text(seed_text, next_words=30, temperature=0.5):
 
     return decoded
 
-def flag_output(prompt_used, text_flagged):
-    if not text_flagged.strip():  # Check for empty or whitespace-only strings
-        return "Cannot flag an empty output."
-
-    logging.info(f"Output flagged: {text_flagged} with prompt: {prompt_used}")
-    timestamp = datetime.now().isoformat()
-    new_data = {"Timestamp": [timestamp], "Prompt": [prompt_used], "Flagged Text": [text_flagged]}
-    new_dataset = Dataset.from_dict(new_data)
-    global flagged_dataset
-    flagged_dataset = Dataset.from_pandas(flagged_dataset.to_pandas().append(new_dataset.to_pandas(), ignore_index=True))
-    flagged_dataset.push_to_hub(FLAGGED_DATASET_ID)
-    return "Output Flagged. Thank you for your feedback."
-
-with gr.Blocks() as demo:
-    gr.Markdown("""# Dungen Dev - Name Generator
-    A prompt-based name generator for game developers.""")
-
-    with gr.Row():
-        with gr.Column():
-            prompt = gr.Textbox(label="Prompt", value="a female character name", max_lines=1)
-            with gr.Row():
-                next_words_slider = gr.Slider(1, 100, step=1, label='Next Words', value=30)
-                temperature_slider = gr.Slider(0.1, 1, value=0.5, label='Temperature', info='Controls randomness of generation, higher values = more creative, lower values = more probalistic')
-            generate_button = gr.Button("Generate")
-        with gr.Column():
-            output_text = gr.Textbox(label="Generated Names", interactive=False)
-            flag_button = gr.Button("Flag Output")
-
-    gr.Markdown("""Dungen Dev is an experimental model, and may produce outputs that are inappropriate, biased, or potentially harmful and inaccurate. Caution is advised.""")
-
-    generate_button.click(
-        fn=generate_text,
-        inputs=[prompt, next_words_slider, temperature_slider],
-        outputs=output_text
-    )
-    flag_button.click(flag_output, inputs=output_text, outputs=gr.Textbox(label="Flag Status", interactive=False))
-
-    demo.examples=[
+demo = gr.Interface(
+    fn=generate_text,
+    inputs=[
+        gr.Textbox(label="Prompt", value="a female character name", max_lines=1),
+        gr.Slider(1, 100, step=1, label='Next Words', value=30),
+        gr.Slider(0.1, 1, value=0.5, label='Temperature', info='Controls randomness of generation, higher values = more creative, lower values = more probalistic')
+    ],
+    outputs=gr.Textbox(label="Generated Names"),
+    title='Dungen Dev - Name Generator',
+    description='A prompt-based name generator for game developers. Dungen Dev is an experimental model, and may produce outputs that are inappropriate, biased, or potentially harmful and inaccurate. Caution is advised.',
+    examples=[
         ["a male character name", 30, 0.5],
         ["a futuristic city name", 30, 0.5],
         ["an item name", 30, 0.5],
         ["a dark and mysterious forest name", 30, 0.5],
         ["an evil character name", 30, 0.5]
     ]
+)
 
 demo.launch()