{ "cells": [ { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "This is the charname_results: .\n", "And this is the dialogue: None\n", "This is the charname_results: .\n", "And this is the dialogue: None\n", "This is the charname_results: .\n", "And this is the dialogue: None\n", "This is the charname_results: .\n", "And this is the dialogue: None\n", "This is the charname_results: .\n", "And this is the dialogue: None\n", "This is the charname_results: .\n", "And this is the dialogue: None\n", "This is the charname_results: .\n", "And this is the dialogue: None\n" ] } ], "source": [ "# usage: sp.encode_as_ids(\"This is a test\")\n", "\n", "# We're using sentencepiece because this is llama2\n", "\n", "\n", "\n", "import re\n", "\n", "charname_regex = re.compile(r\"【.+?】\")\n", "dialogue_regex = re.compile(r\"「.+?」\")\n", "# re.search(str)\n", "\n", "def makecols(str):\n", " charname_results = charname_regex.search(str)\n", " dialogue_results = dialogue_regex.search(str)\n", " if charname_results is None:\n", " return ['MONOLOGUE',dialogue_results or \"\"]\n", " try: \n", " return (charname_results.group(0)[1:][:-1],dialogue_results.group(0)[1:][:-1])\n", " except:\n", " print(f\"This is the charname_results: {charname_results}.\\nAnd this is the dialogue: {dialogue_results}\")\n", " return ['ERROR!','']\n", "\n", "\n", "def not_monologue(tup):\n", " if (tup[0] == 'MONOLOGUE'):\n", " return False\n", " return True\n", "\n", "\n", "script = open('muvluv-chizururoute-script.txt').read().replace('\\n','').replace('\\x05','').split('\u0001')\n", "script = list(map(makecols,script))\n", "# script = list(filter(not_monologue,script)) \n", "\n", "# tokenizer = AutoTokenizer.from_pretrained(\"meta-llama/Llama-2-7b-hf\")" ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [], "source": [ "from transformers import AutoTokenizer\n", "from tqdm import tqdm\n", "\n", "tokenizer = AutoTokenizer.from_pretrained(\"bigscience/bloomz-3b\") # placeholder until I get huggingface llama access\n", "\n", "def generate_examples(script, tokenizer, chizuru_count_min=2, window_length=10, takeru_count_min=2, max_lines_without_chizuru=6):\n", " MAX_TOKENS = 800 # Change this value if you want to use a different token limit\n", "\n", " examples = []\n", " sliding_window = []\n", " example = []\n", " chizuru_counter = 0\n", " takeru_counter = 0\n", " lines_without_chizuru = 0\n", " making_conversation = False\n", "\n", " for dialogue in tqdm(script):\n", " speaker, line = dialogue\n", "\n", " if len(sliding_window) == window_length:\n", " sliding_window.pop(0) # Remove first element\n", "\n", " sliding_window.append(dialogue)\n", "\n", " # Check if there are more than chizuru_count_min spoken lines from Chizuru across sliding_window\n", " chizuru_counter = sum(1 for d in sliding_window if d[0] == 'Chizuru')\n", " takeru_counter = sum(1 for d in sliding_window if d[0] == 'Takeru')\n", "\n", " if speaker == 'Chizuru':\n", " lines_without_chizuru = 0 # Reset count\n", " else:\n", " lines_without_chizuru += 1 # Increment count\n", " \n", " can_start_conversation = chizuru_counter >= chizuru_count_min and takeru_counter >= takeru_count_min\n", " should_stop_conversation = making_conversation and (len(tokenizer.encode(' '.join([d[1] for d in example]))) > MAX_TOKENS or lines_without_chizuru > max_lines_without_chizuru)\n", " \n", " if making_conversation:\n", " if should_stop_conversation: # making conversation and should stop\n", " examples.append(example)\n", " example = []\n", " sliding_window = []\n", " chizuru_counter = 0\n", " takeru_counter = 0\n", " lines_without_chizuru = 0\n", " making_conversation = False\n", " else: # making conversation and should not stop\n", " example.append(dialogue)\n", " elif can_start_conversation: # not making conversation and should start, by appending an example to conversation\n", " example.append(dialogue)\n", " making_conversation = True\n", "\n", " if example: # Add last example if it's non-empty\n", " examples.append(example)\n", "\n", " return examples\n", "\n", "\n", "# additional step: remove all non-chizuru examples at the end of each example. They're literally pointless and will not be used in training data anyway.\n", "\n", "# def format_conversations(script, speaker1='Chizuru', speaker2='Takeru', speaker1_count=3, speaker2_count=2, window_size=6, max_tokens=600):\n", "# conversations = [] # to hold the conversations\n", "# current_conversation = [] # to hold the current conversation\n", "# current_window = [] # to hold the current window of lines\n", " \n", "# for line in script:\n", "# speaker, dialogue = line\n", "# current_window.append(line)\n", "\n", "# # If window is larger than window_size, remove the oldest line\n", "# if len(current_window) > window_size:\n", "# current_window.pop(0)\n", "\n", "# # Count the dialogues of speaker1 and speaker2 in the current window\n", "# speaker1_dialogues = sum([1 for line in current_window if line[0] == speaker1])\n", "# speaker2_dialogues = sum([1 for line in current_window if line[0] == speaker2])\n", "\n", "# # If conditions are met, add dialogues to the current conversation\n", "# if speaker1_dialogues >= speaker1_count and speaker2_dialogues >= speaker2_count:\n", "# current_conversation.append(line)\n", "\n", "# # If the current conversation reaches the max_tokens limit, add it to the conversations and reset current_conversation\n", "# if tokenizer.encode(' '.join([dialogue for _, dialogue in current_conversation]), return_tensors='pt').shape[1] > max_tokens:\n", "# conversations.append(current_conversation)\n", "# current_conversation = []\n", "# current_window = []\n", " \n", "# # Add the last conversation if it was not added before\n", "# if current_conversation and tokenizer.encode(' '.join([dialogue for _, dialogue in current_conversation]), return_tensors='pt').shape[1] <= max_tokens:\n", "# conversations.append(current_conversation)\n", "\n", "# return conversations" ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "['MONOLOGUE', '']\n" ] } ], "source": [ "print(script[0])" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "100%|██████████| 12876/12876 [00:01<00:00, 7052.90it/s]\n" ] } ], "source": [ "created_examples_script = generate_examples(script, tokenizer,)" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[('Chizuru', 'Hold on. You were WHAT?'), ('Takeru', 'Curse you, Sumikaaaaa!!'), ('Chizuru', 'Shirooogaaane-kun...'), ('Takeru', \"Class Rep, calm down.We're in the middle of homeroom right now, remember...?\"), ('Marimo', \"Alright everyone, homeroom's over! I'm gonna go over to Class D for a bit. I'm leaving you in charge, Sakaki-san!Just let me know who's absent when I get back!\"), ('Takeru', '......'), ('Chizuru', \"Yes, ma'am! Understood!\"), ['MONOLOGUE', ''], ('Chizuru', 'Now then... What was that about?'), ('Takeru', \"It's kind of a long story.\"), ('Chizuru', 'Give me the short version, then.'), ('Takeru', \"I'll do my best.Let's see here — Basically, it's because I told her there was a demon on the swim team.\"), ('Miki', 'That was short!!'), ('Chizuru', 'Are you kidding me?'), ('Takeru', 'Not in the slightest.'), ('Chizuru', '...And?'), ('Takeru', \"Then Meiya said something about bringing her to justice and ran off on her own. I didn't do anything.\"), ('Sumika', 'And then after that...'), ('Takeru', 'Sumika!!'), ('Sumika', 'Bleh.'), ['MONOLOGUE', ''], ('Takeru', \"Heh... heh... heh... I know all about it, Class Rep. You were rivals with the former swim team captain, weren't you? Did you really think you had a chance against her?\"), ('Chizuru', '...Shut up!! What do you know about me, anyway!?'), ('Takeru', \"What? I can't just take that sitting down... I can't believe you're treating me like I'm just any other guy.\"), ('Chizuru', '...What do you mean?'), ('Takeru', \"I'm Shirogane Takeru!!\"), ('Chizuru', 'I know that. Are you stupid?'), ('Takeru', \"Mm, that didn't come out quite right... Well, in any case, you ought to stop being so jealous of her.\"), ('Chizuru', 'Jealous?'), ('Takeru', \"A close friend, and a former rival. And you competed against each other back when she led that club... but now, you're both performing on different stages...\"), ('Chizuru', 'What did you just say...?'), ('Takeru', \"Well, I mean... Can't you see what I'm saying? Yeesh, don't make me spell it out for you.\"), ('Chizuru', 'Stop making fun of me already!'), ('Takeru', \"I'm not making fun of you, I'm just telling it how it is.\"), ('Chizuru', 'You wanna say that again!?'), ('Takeru', 'Oops! Better make my escape before the Class Rep blows a fuse!'), ['MONOLOGUE', ''], ('Miki', 'Whoa... Chizuru-chan is faaast.'), ('Chizuru', \"You're not getting away from me.\"), ('Takeru', \"Come on... Let go of me already.Don't make me make fun of you again!\"), ('Chizuru', \"Homeroom isn't over yet.Don't leave the classroom.\"), ('Takeru', \"But Marimo-chan said it was over, didn't she!?\"), ('Chizuru', \"Not according to the schedule, it isn't.We still have five minutes left.\"), ('Takeru', 'Tch... Nobody likes a stubborn woman, you know that?'), ('Chizuru', \"I don't need you to like me.\"), ('Takeru', 'Dammit...'), ('Chizuru', \"That's more like it. Now go back to your seat...\"), ('Takeru', \"...Ha! You've fallen for my ruse!!\"), ('Chizuru', 'Ah! Wait a... Kyah!!'), ('Takeru', 'Whoa!!'), ['MONOLOGUE', ''], ['MONOLOGUE', ''], ('Chizuru', 'Kyaaaaah!!!'), ('Takeru', 'Ow-ow-ow-ow-ow...'), ('Chizuru', 'Ugh... Ouch...'), ('Takeru & Chizuru', '!?'), ('Chizuru', \"Wh-Wh-Wh-Wh-What're you doing!?Get off of me right this instant!!\"), ('Takeru', \"That's my line! You're the one on top of me!!\"), ('Voice', '...Such intimacy... and in public...'), ('Takeru & Chizuru', 'Huh!? Ah!!'), ('Takeru', 'Ayamine!'), ('Chizuru', 'Th-This is just an accident!M-M-More importantly, Ayamine-san...'), ('Ayamine', '......'), ('Chizuru', 'At least say hello!'), ('Ayamine', '......'), ('Chizuru', '...Ayamine-san, are you listening?'), ('Ayamine', '......'), ('Chizuru', '......'), ('Kei', 'Yes, good morning.'), ('Chizuru', 'Grrrrrrr!!'), ('Kei', '......'), ('Chizuru', \"You cut class yesterday, you're late today... Do you really think it's okay to keep doing that?\"), ('Kei', '...Yesterday...'), ('Chizuru', 'What about it?'), ('Kei', '......'), ('Chizuru', '......'), ('Kei', '...I got lost.'), ('Chizuru', 'How can you lie with such a straight face!?'), ('Kei', 'Busted...?'), ('Miki', \"It's pretty obvious.\"), ('Kei', \"...Being late once in a while isn't so bad.\"), ('Chizuru', \"Once in a while!? You're late every time you come to school, and most of the time you don't even show up at all!!\"), ('Kei', 'But I got to see something nice...'), ('Miki', 'Hey! Guess what, Kei-chan? We got a new transfer student yesterday!'), ('Takeru', 'Heeey...'), ('Chizuru', '......'), ['MONOLOGUE', ''], ('Takeru', \"I'll just go on ahead, then.\")]\n" ] } ], "source": [ "print(created_examples_script[2])" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [], "source": [ "def merge_consecutive_lines(conversation):\n", " merged_conversation = []\n", " last_speaker = None\n", " for speaker, line in conversation:\n", " if not merged_conversation or speaker != last_speaker:\n", " # New speaker or first dialogue, just add it to the list\n", " merged_conversation.append((speaker, line))\n", " else:\n", " # Same speaker as before, concatenate the lines\n", " prev_speaker, prev_line = merged_conversation.pop()\n", " merged_conversation.append((prev_speaker, prev_line + \" \" + line))\n", " last_speaker = speaker\n", " return list(filter(not_monologue, merged_conversation)) # why do this step here? Because I don't want to iterate over the dataset twice, and monologues should count when examples are being generated with the sliding window, so I can't remove them in the usual spot.\n" ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [], "source": [ "conversations_processed = list(map(merge_consecutive_lines, created_examples_script))" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[('Chizuru', 'Hold on. You were WHAT?'), ('Takeru', 'Curse you, Sumikaaaaa!!'), ('Chizuru', 'Shirooogaaane-kun...'), ('Takeru', \"Class Rep, calm down.We're in the middle of homeroom right now, remember...?\"), ('Marimo', \"Alright everyone, homeroom's over! I'm gonna go over to Class D for a bit. I'm leaving you in charge, Sakaki-san!Just let me know who's absent when I get back!\"), ('Takeru', '......'), ('Chizuru', \"Yes, ma'am! Understood!\"), ('Chizuru', 'Now then... What was that about?'), ('Takeru', \"It's kind of a long story.\"), ('Chizuru', 'Give me the short version, then.'), ('Takeru', \"I'll do my best.Let's see here — Basically, it's because I told her there was a demon on the swim team.\"), ('Miki', 'That was short!!'), ('Chizuru', 'Are you kidding me?'), ('Takeru', 'Not in the slightest.'), ('Chizuru', '...And?'), ('Takeru', \"Then Meiya said something about bringing her to justice and ran off on her own. I didn't do anything.\"), ('Sumika', 'And then after that...'), ('Takeru', 'Sumika!!'), ('Sumika', 'Bleh.'), ('Takeru', \"Heh... heh... heh... I know all about it, Class Rep. You were rivals with the former swim team captain, weren't you? Did you really think you had a chance against her?\"), ('Chizuru', '...Shut up!! What do you know about me, anyway!?'), ('Takeru', \"What? I can't just take that sitting down... I can't believe you're treating me like I'm just any other guy.\"), ('Chizuru', '...What do you mean?'), ('Takeru', \"I'm Shirogane Takeru!!\"), ('Chizuru', 'I know that. Are you stupid?'), ('Takeru', \"Mm, that didn't come out quite right... Well, in any case, you ought to stop being so jealous of her.\"), ('Chizuru', 'Jealous?'), ('Takeru', \"A close friend, and a former rival. And you competed against each other back when she led that club... but now, you're both performing on different stages...\"), ('Chizuru', 'What did you just say...?'), ('Takeru', \"Well, I mean... Can't you see what I'm saying? Yeesh, don't make me spell it out for you.\"), ('Chizuru', 'Stop making fun of me already!'), ('Takeru', \"I'm not making fun of you, I'm just telling it how it is.\"), ('Chizuru', 'You wanna say that again!?'), ('Takeru', 'Oops! Better make my escape before the Class Rep blows a fuse!'), ('Miki', 'Whoa... Chizuru-chan is faaast.'), ('Chizuru', \"You're not getting away from me.\"), ('Takeru', \"Come on... Let go of me already.Don't make me make fun of you again!\"), ('Chizuru', \"Homeroom isn't over yet.Don't leave the classroom.\"), ('Takeru', \"But Marimo-chan said it was over, didn't she!?\"), ('Chizuru', \"Not according to the schedule, it isn't.We still have five minutes left.\"), ('Takeru', 'Tch... Nobody likes a stubborn woman, you know that?'), ('Chizuru', \"I don't need you to like me.\"), ('Takeru', 'Dammit...'), ('Chizuru', \"That's more like it. Now go back to your seat...\"), ('Takeru', \"...Ha! You've fallen for my ruse!!\"), ('Chizuru', 'Ah! Wait a... Kyah!!'), ('Takeru', 'Whoa!!'), ('Chizuru', 'Kyaaaaah!!!'), ('Takeru', 'Ow-ow-ow-ow-ow...'), ('Chizuru', 'Ugh... Ouch...'), ('Takeru & Chizuru', '!?'), ('Chizuru', \"Wh-Wh-Wh-Wh-What're you doing!?Get off of me right this instant!!\"), ('Takeru', \"That's my line! You're the one on top of me!!\"), ('Voice', '...Such intimacy... and in public...'), ('Takeru & Chizuru', 'Huh!? Ah!!'), ('Takeru', 'Ayamine!'), ('Chizuru', 'Th-This is just an accident!M-M-More importantly, Ayamine-san...'), ('Ayamine', '......'), ('Chizuru', 'At least say hello!'), ('Ayamine', '......'), ('Chizuru', '...Ayamine-san, are you listening?'), ('Ayamine', '......'), ('Chizuru', '......'), ('Kei', 'Yes, good morning.'), ('Chizuru', 'Grrrrrrr!!'), ('Kei', '......'), ('Chizuru', \"You cut class yesterday, you're late today... Do you really think it's okay to keep doing that?\"), ('Kei', '...Yesterday...'), ('Chizuru', 'What about it?'), ('Kei', '......'), ('Chizuru', '......'), ('Kei', '...I got lost.'), ('Chizuru', 'How can you lie with such a straight face!?'), ('Kei', 'Busted...?'), ('Miki', \"It's pretty obvious.\"), ('Kei', \"...Being late once in a while isn't so bad.\"), ('Chizuru', \"Once in a while!? You're late every time you come to school, and most of the time you don't even show up at all!!\"), ('Kei', 'But I got to see something nice...'), ('Miki', 'Hey! Guess what, Kei-chan? We got a new transfer student yesterday!'), ('Takeru', 'Heeey...'), ('Chizuru', '......'), ('Takeru', \"I'll just go on ahead, then.\")]\n" ] } ], "source": [ "print(conversations_processed[2])" ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [], "source": [ "def add_space_after_punctuation(conversations):\n", " corrected_conversations = []\n", " for conversation in conversations:\n", " corrected_conversation = []\n", " for speaker, line in conversation:\n", " # Add a space wherever there is a punctuation mark followed by a letter, excluding ellipsis\n", " corrected_line = re.sub(r'([.,!?])(\\w)', r'\\1 \\2', line)\n", " corrected_conversation.append((speaker, corrected_line))\n", " corrected_conversations.append(corrected_conversation)\n", " return corrected_conversations" ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [ { "ename": "NameError", "evalue": "name 'add_space_after_punctuation' is not defined", "output_type": "error", "traceback": [ "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", "\u001b[0;31mNameError\u001b[0m Traceback (most recent call last)", "Cell \u001b[0;32mIn[1], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m conversations_processed_whitespacefix \u001b[39m=\u001b[39m add_space_after_punctuation(conversations_processed)\n", "\u001b[0;31mNameError\u001b[0m: name 'add_space_after_punctuation' is not defined" ] } ], "source": [ "conversations_processed_whitespacefix = add_space_after_punctuation(conversations_processed)" ] }, { "cell_type": "code", "execution_count": 11, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[('Chizuru', 'Hold on. You were WHAT?'), ('Takeru', 'Curse you, Sumikaaaaa!!'), ('Chizuru', 'Shirooogaaane-kun...'), ('Takeru', \"Class Rep, calm down. We're in the middle of homeroom right now, remember...?\"), ('Marimo', \"Alright everyone, homeroom's over! I'm gonna go over to Class D for a bit. I'm leaving you in charge, Sakaki-san! Just let me know who's absent when I get back!\"), ('Takeru', '......'), ('Chizuru', \"Yes, ma'am! Understood!\"), ('Chizuru', 'Now then... What was that about?'), ('Takeru', \"It's kind of a long story.\"), ('Chizuru', 'Give me the short version, then.'), ('Takeru', \"I'll do my best. Let's see here — Basically, it's because I told her there was a demon on the swim team.\"), ('Miki', 'That was short!!'), ('Chizuru', 'Are you kidding me?'), ('Takeru', 'Not in the slightest.'), ('Chizuru', '... And?'), ('Takeru', \"Then Meiya said something about bringing her to justice and ran off on her own. I didn't do anything.\"), ('Sumika', 'And then after that...'), ('Takeru', 'Sumika!!'), ('Sumika', 'Bleh.'), ('Takeru', \"Heh... heh... heh... I know all about it, Class Rep. You were rivals with the former swim team captain, weren't you? Did you really think you had a chance against her?\"), ('Chizuru', '... Shut up!! What do you know about me, anyway!?'), ('Takeru', \"What? I can't just take that sitting down... I can't believe you're treating me like I'm just any other guy.\"), ('Chizuru', '... What do you mean?'), ('Takeru', \"I'm Shirogane Takeru!!\"), ('Chizuru', 'I know that. Are you stupid?'), ('Takeru', \"Mm, that didn't come out quite right... Well, in any case, you ought to stop being so jealous of her.\"), ('Chizuru', 'Jealous?'), ('Takeru', \"A close friend, and a former rival. And you competed against each other back when she led that club... but now, you're both performing on different stages...\"), ('Chizuru', 'What did you just say...?'), ('Takeru', \"Well, I mean... Can't you see what I'm saying? Yeesh, don't make me spell it out for you.\"), ('Chizuru', 'Stop making fun of me already!'), ('Takeru', \"I'm not making fun of you, I'm just telling it how it is.\"), ('Chizuru', 'You wanna say that again!?'), ('Takeru', 'Oops! Better make my escape before the Class Rep blows a fuse!'), ('Miki', 'Whoa... Chizuru-chan is faaast.'), ('Chizuru', \"You're not getting away from me.\"), ('Takeru', \"Come on... Let go of me already. Don't make me make fun of you again!\"), ('Chizuru', \"Homeroom isn't over yet. Don't leave the classroom.\"), ('Takeru', \"But Marimo-chan said it was over, didn't she!?\"), ('Chizuru', \"Not according to the schedule, it isn't. We still have five minutes left.\"), ('Takeru', 'Tch... Nobody likes a stubborn woman, you know that?'), ('Chizuru', \"I don't need you to like me.\"), ('Takeru', 'Dammit...'), ('Chizuru', \"That's more like it. Now go back to your seat...\"), ('Takeru', \"... Ha! You've fallen for my ruse!!\"), ('Chizuru', 'Ah! Wait a... Kyah!!'), ('Takeru', 'Whoa!!'), ('Chizuru', 'Kyaaaaah!!!'), ('Takeru', 'Ow-ow-ow-ow-ow...'), ('Chizuru', 'Ugh... Ouch...'), ('Takeru & Chizuru', '!?'), ('Chizuru', \"Wh-Wh-Wh-Wh-What're you doing!? Get off of me right this instant!!\"), ('Takeru', \"That's my line! You're the one on top of me!!\"), ('Voice', '... Such intimacy... and in public...'), ('Takeru & Chizuru', 'Huh!? Ah!!'), ('Takeru', 'Ayamine!'), ('Chizuru', 'Th-This is just an accident! M-M-More importantly, Ayamine-san...'), ('Ayamine', '......'), ('Chizuru', 'At least say hello!'), ('Ayamine', '......'), ('Chizuru', '... Ayamine-san, are you listening?'), ('Ayamine', '......'), ('Chizuru', '......'), ('Kei', 'Yes, good morning.'), ('Chizuru', 'Grrrrrrr!!'), ('Kei', '......'), ('Chizuru', \"You cut class yesterday, you're late today... Do you really think it's okay to keep doing that?\"), ('Kei', '... Yesterday...'), ('Chizuru', 'What about it?'), ('Kei', '......'), ('Chizuru', '......'), ('Kei', '... I got lost.'), ('Chizuru', 'How can you lie with such a straight face!?'), ('Kei', 'Busted...?'), ('Miki', \"It's pretty obvious.\"), ('Kei', \"... Being late once in a while isn't so bad.\"), ('Chizuru', \"Once in a while!? You're late every time you come to school, and most of the time you don't even show up at all!!\"), ('Kei', 'But I got to see something nice...'), ('Miki', 'Hey! Guess what, Kei-chan? We got a new transfer student yesterday!'), ('Takeru', 'Heeey...'), ('Chizuru', '......'), ('Takeru', \"I'll just go on ahead, then.\")]\n" ] } ], "source": [ "print(conversations_processed_whitespacefix[2])" ] }, { "cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [], "source": [ "def generate_training_examples(conversation):\n", " training_examples = []\n", " temp_dialogue = []\n", " for idx, dialogue in enumerate(conversation):\n", " speaker, _ = dialogue\n", " temp_dialogue.append(dialogue)\n", " if speaker == 'Chizuru' and idx != 0:\n", " training_examples.append(temp_dialogue.copy()) # Add up to and including current line\n", " return training_examples" ] }, { "cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "[[('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\")],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\")],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\")],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\")],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\")],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?')],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\")],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\"),\n", " ('Kei', 'End of the line...'),\n", " ('Sumika', \"Hey, what're you doing? Takeru-chan, let's go home!\"),\n", " ('Kei', 'Here, Kagami. Good luck with the cleaning. Bye now.'),\n", " ('Sumika', 'Huh? Oh, okay. Bye-bye.'),\n", " ('Chizuru', 'W-Wait just a minute!')],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\"),\n", " ('Kei', 'End of the line...'),\n", " ('Sumika', \"Hey, what're you doing? Takeru-chan, let's go home!\"),\n", " ('Kei', 'Here, Kagami. Good luck with the cleaning. Bye now.'),\n", " ('Sumika', 'Huh? Oh, okay. Bye-bye.'),\n", " ('Chizuru', 'W-Wait just a minute!'),\n", " ('Kei', '... Her problem now.'),\n", " ('Chizuru', \"No it isn't!!\")],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\"),\n", " ('Kei', 'End of the line...'),\n", " ('Sumika', \"Hey, what're you doing? Takeru-chan, let's go home!\"),\n", " ('Kei', 'Here, Kagami. Good luck with the cleaning. Bye now.'),\n", " ('Sumika', 'Huh? Oh, okay. Bye-bye.'),\n", " ('Chizuru', 'W-Wait just a minute!'),\n", " ('Kei', '... Her problem now.'),\n", " ('Chizuru', \"No it isn't!!\"),\n", " ('Meiya', 'What is it? Are you not going home?'),\n", " ('Kei', \"She'll help too.\"),\n", " ('Meiya', 'What do you mean?'),\n", " ('Chizuru', 'Would you just... Ah, wait!!')],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\"),\n", " ('Kei', 'End of the line...'),\n", " ('Sumika', \"Hey, what're you doing? Takeru-chan, let's go home!\"),\n", " ('Kei', 'Here, Kagami. Good luck with the cleaning. Bye now.'),\n", " ('Sumika', 'Huh? Oh, okay. Bye-bye.'),\n", " ('Chizuru', 'W-Wait just a minute!'),\n", " ('Kei', '... Her problem now.'),\n", " ('Chizuru', \"No it isn't!!\"),\n", " ('Meiya', 'What is it? Are you not going home?'),\n", " ('Kei', \"She'll help too.\"),\n", " ('Meiya', 'What do you mean?'),\n", " ('Chizuru', 'Would you just... Ah, wait!!'),\n", " ('Sumika', 'Hey, Takeru-chan, over here! Help me move the desks!'),\n", " ('Takeru', \"Are you an idiot!? Don't just go with the flow!\"),\n", " ('Chizuru', 'Huh? ... A-Ah...! Ugh, forget it!! What a jerk!!')],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\"),\n", " ('Kei', 'End of the line...'),\n", " ('Sumika', \"Hey, what're you doing? Takeru-chan, let's go home!\"),\n", " ('Kei', 'Here, Kagami. Good luck with the cleaning. Bye now.'),\n", " ('Sumika', 'Huh? Oh, okay. Bye-bye.'),\n", " ('Chizuru', 'W-Wait just a minute!'),\n", " ('Kei', '... Her problem now.'),\n", " ('Chizuru', \"No it isn't!!\"),\n", " ('Meiya', 'What is it? Are you not going home?'),\n", " ('Kei', \"She'll help too.\"),\n", " ('Meiya', 'What do you mean?'),\n", " ('Chizuru', 'Would you just... Ah, wait!!'),\n", " ('Sumika', 'Hey, Takeru-chan, over here! Help me move the desks!'),\n", " ('Takeru', \"Are you an idiot!? Don't just go with the flow!\"),\n", " ('Chizuru', 'Huh? ... A-Ah...! Ugh, forget it!! What a jerk!!'),\n", " ('Takeru', 'Huh?'),\n", " ('Meiya', 'Is something about to commence?'),\n", " ('Sumika', \"We're gonna clean the room.\"),\n", " ('Meiya',\n", " 'I see... so you yourselves clean that which you have dirtied... A novel and admirable approach. I shall assist.'),\n", " ('Chizuru', 'Shirogane-kun, quit slacking off. Come on, just get to work.')],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\"),\n", " ('Kei', 'End of the line...'),\n", " ('Sumika', \"Hey, what're you doing? Takeru-chan, let's go home!\"),\n", " ('Kei', 'Here, Kagami. Good luck with the cleaning. Bye now.'),\n", " ('Sumika', 'Huh? Oh, okay. Bye-bye.'),\n", " ('Chizuru', 'W-Wait just a minute!'),\n", " ('Kei', '... Her problem now.'),\n", " ('Chizuru', \"No it isn't!!\"),\n", " ('Meiya', 'What is it? Are you not going home?'),\n", " ('Kei', \"She'll help too.\"),\n", " ('Meiya', 'What do you mean?'),\n", " ('Chizuru', 'Would you just... Ah, wait!!'),\n", " ('Sumika', 'Hey, Takeru-chan, over here! Help me move the desks!'),\n", " ('Takeru', \"Are you an idiot!? Don't just go with the flow!\"),\n", " ('Chizuru', 'Huh? ... A-Ah...! Ugh, forget it!! What a jerk!!'),\n", " ('Takeru', 'Huh?'),\n", " ('Meiya', 'Is something about to commence?'),\n", " ('Sumika', \"We're gonna clean the room.\"),\n", " ('Meiya',\n", " 'I see... so you yourselves clean that which you have dirtied... A novel and admirable approach. I shall assist.'),\n", " ('Chizuru', 'Shirogane-kun, quit slacking off. Come on, just get to work.'),\n", " ('Takeru',\n", " 'Alright, I get it... Jeez, you sure know how to kill my motivation.'),\n", " ('Chizuru',\n", " \"There are some things you just have to do, whether you're motivated or not.\")],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\"),\n", " ('Kei', 'End of the line...'),\n", " ('Sumika', \"Hey, what're you doing? Takeru-chan, let's go home!\"),\n", " ('Kei', 'Here, Kagami. Good luck with the cleaning. Bye now.'),\n", " ('Sumika', 'Huh? Oh, okay. Bye-bye.'),\n", " ('Chizuru', 'W-Wait just a minute!'),\n", " ('Kei', '... Her problem now.'),\n", " ('Chizuru', \"No it isn't!!\"),\n", " ('Meiya', 'What is it? Are you not going home?'),\n", " ('Kei', \"She'll help too.\"),\n", " ('Meiya', 'What do you mean?'),\n", " ('Chizuru', 'Would you just... Ah, wait!!'),\n", " ('Sumika', 'Hey, Takeru-chan, over here! Help me move the desks!'),\n", " ('Takeru', \"Are you an idiot!? Don't just go with the flow!\"),\n", " ('Chizuru', 'Huh? ... A-Ah...! Ugh, forget it!! What a jerk!!'),\n", " ('Takeru', 'Huh?'),\n", " ('Meiya', 'Is something about to commence?'),\n", " ('Sumika', \"We're gonna clean the room.\"),\n", " ('Meiya',\n", " 'I see... so you yourselves clean that which you have dirtied... A novel and admirable approach. I shall assist.'),\n", " ('Chizuru', 'Shirogane-kun, quit slacking off. Come on, just get to work.'),\n", " ('Takeru',\n", " 'Alright, I get it... Jeez, you sure know how to kill my motivation.'),\n", " ('Chizuru',\n", " \"There are some things you just have to do, whether you're motivated or not.\"),\n", " ('Takeru', \"Yeah, yeah... We're done! We're done!\"),\n", " ('Sumika', \"Alright, let's go home.\"),\n", " ('Meiya', 'Yes.'),\n", " ('Takeru', 'Class Rep, you going home, too?'),\n", " ('Chizuru', 'Huh? No, I have club activities.')],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\"),\n", " ('Kei', 'End of the line...'),\n", " ('Sumika', \"Hey, what're you doing? Takeru-chan, let's go home!\"),\n", " ('Kei', 'Here, Kagami. Good luck with the cleaning. Bye now.'),\n", " ('Sumika', 'Huh? Oh, okay. Bye-bye.'),\n", " ('Chizuru', 'W-Wait just a minute!'),\n", " ('Kei', '... Her problem now.'),\n", " ('Chizuru', \"No it isn't!!\"),\n", " ('Meiya', 'What is it? Are you not going home?'),\n", " ('Kei', \"She'll help too.\"),\n", " ('Meiya', 'What do you mean?'),\n", " ('Chizuru', 'Would you just... Ah, wait!!'),\n", " ('Sumika', 'Hey, Takeru-chan, over here! Help me move the desks!'),\n", " ('Takeru', \"Are you an idiot!? Don't just go with the flow!\"),\n", " ('Chizuru', 'Huh? ... A-Ah...! Ugh, forget it!! What a jerk!!'),\n", " ('Takeru', 'Huh?'),\n", " ('Meiya', 'Is something about to commence?'),\n", " ('Sumika', \"We're gonna clean the room.\"),\n", " ('Meiya',\n", " 'I see... so you yourselves clean that which you have dirtied... A novel and admirable approach. I shall assist.'),\n", " ('Chizuru', 'Shirogane-kun, quit slacking off. Come on, just get to work.'),\n", " ('Takeru',\n", " 'Alright, I get it... Jeez, you sure know how to kill my motivation.'),\n", " ('Chizuru',\n", " \"There are some things you just have to do, whether you're motivated or not.\"),\n", " ('Takeru', \"Yeah, yeah... We're done! We're done!\"),\n", " ('Sumika', \"Alright, let's go home.\"),\n", " ('Meiya', 'Yes.'),\n", " ('Takeru', 'Class Rep, you going home, too?'),\n", " ('Chizuru', 'Huh? No, I have club activities.'),\n", " ('Takeru', 'Oh. I see.'),\n", " ('Chizuru', '... Are you implying something?')],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\"),\n", " ('Kei', 'End of the line...'),\n", " ('Sumika', \"Hey, what're you doing? Takeru-chan, let's go home!\"),\n", " ('Kei', 'Here, Kagami. Good luck with the cleaning. Bye now.'),\n", " ('Sumika', 'Huh? Oh, okay. Bye-bye.'),\n", " ('Chizuru', 'W-Wait just a minute!'),\n", " ('Kei', '... Her problem now.'),\n", " ('Chizuru', \"No it isn't!!\"),\n", " ('Meiya', 'What is it? Are you not going home?'),\n", " ('Kei', \"She'll help too.\"),\n", " ('Meiya', 'What do you mean?'),\n", " ('Chizuru', 'Would you just... Ah, wait!!'),\n", " ('Sumika', 'Hey, Takeru-chan, over here! Help me move the desks!'),\n", " ('Takeru', \"Are you an idiot!? Don't just go with the flow!\"),\n", " ('Chizuru', 'Huh? ... A-Ah...! Ugh, forget it!! What a jerk!!'),\n", " ('Takeru', 'Huh?'),\n", " ('Meiya', 'Is something about to commence?'),\n", " ('Sumika', \"We're gonna clean the room.\"),\n", " ('Meiya',\n", " 'I see... so you yourselves clean that which you have dirtied... A novel and admirable approach. I shall assist.'),\n", " ('Chizuru', 'Shirogane-kun, quit slacking off. Come on, just get to work.'),\n", " ('Takeru',\n", " 'Alright, I get it... Jeez, you sure know how to kill my motivation.'),\n", " ('Chizuru',\n", " \"There are some things you just have to do, whether you're motivated or not.\"),\n", " ('Takeru', \"Yeah, yeah... We're done! We're done!\"),\n", " ('Sumika', \"Alright, let's go home.\"),\n", " ('Meiya', 'Yes.'),\n", " ('Takeru', 'Class Rep, you going home, too?'),\n", " ('Chizuru', 'Huh? No, I have club activities.'),\n", " ('Takeru', 'Oh. I see.'),\n", " ('Chizuru', '... Are you implying something?'),\n", " ('Takeru', 'Huh?'),\n", " ('Chizuru',\n", " \"It may be on the verge of disbanding, but we're still giving it our all...\")],\n", " [('Chizuru', '... Sorry to have to spoil the celebration, but...'),\n", " ('Takeru', 'Ah?'),\n", " ('Chizuru', \"Shirogane-kun, you're on cleaning duty today. Have fun.\"),\n", " ('Takeru', 'NYET!'),\n", " ('Chizuru', \"Ayamine-san. Where do you think you're going?\"),\n", " ('Kei', '......'),\n", " ('Chizuru', \"You've got cleaning duty too, you know.\"),\n", " ('Kei', '......'),\n", " ('Chizuru',\n", " \"... Do you have something to say? Don't tell me you think it's not your mess to clean up just because you're never here...\"),\n", " ('Kei', \"... You're good.\"),\n", " ('Chizuru',\n", " \"Excuses like that aren't going to fly. This is a rule, so stop being selfish and obey it.\"),\n", " ('Kei', '... Dog of the system.'),\n", " ('Chizuru', 'What did you call me!?'),\n", " ('Kei', \"Shirogane said I didn't have to.\"),\n", " ('Takeru', 'Whaaat?'),\n", " ('Chizuru',\n", " \"Did he now? Well, unfortunately for you, I'm also on cleaning duty. I'm not letting you off that easy. Here's your broom.\"),\n", " ('Kei', 'End of the line...'),\n", " ('Sumika', \"Hey, what're you doing? Takeru-chan, let's go home!\"),\n", " ('Kei', 'Here, Kagami. Good luck with the cleaning. Bye now.'),\n", " ('Sumika', 'Huh? Oh, okay. Bye-bye.'),\n", " ('Chizuru', 'W-Wait just a minute!'),\n", " ('Kei', '... Her problem now.'),\n", " ('Chizuru', \"No it isn't!!\"),\n", " ('Meiya', 'What is it? Are you not going home?'),\n", " ('Kei', \"She'll help too.\"),\n", " ('Meiya', 'What do you mean?'),\n", " ('Chizuru', 'Would you just... Ah, wait!!'),\n", " ('Sumika', 'Hey, Takeru-chan, over here! Help me move the desks!'),\n", " ('Takeru', \"Are you an idiot!? Don't just go with the flow!\"),\n", " ('Chizuru', 'Huh? ... A-Ah...! Ugh, forget it!! What a jerk!!'),\n", " ('Takeru', 'Huh?'),\n", " ('Meiya', 'Is something about to commence?'),\n", " ('Sumika', \"We're gonna clean the room.\"),\n", " ('Meiya',\n", " 'I see... so you yourselves clean that which you have dirtied... A novel and admirable approach. I shall assist.'),\n", " ('Chizuru', 'Shirogane-kun, quit slacking off. Come on, just get to work.'),\n", " ('Takeru',\n", " 'Alright, I get it... Jeez, you sure know how to kill my motivation.'),\n", " ('Chizuru',\n", " \"There are some things you just have to do, whether you're motivated or not.\"),\n", " ('Takeru', \"Yeah, yeah... We're done! We're done!\"),\n", " ('Sumika', \"Alright, let's go home.\"),\n", " ('Meiya', 'Yes.'),\n", " ('Takeru', 'Class Rep, you going home, too?'),\n", " ('Chizuru', 'Huh? No, I have club activities.'),\n", " ('Takeru', 'Oh. I see.'),\n", " ('Chizuru', '... Are you implying something?'),\n", " ('Takeru', 'Huh?'),\n", " ('Chizuru',\n", " \"It may be on the verge of disbanding, but we're still giving it our all...\"),\n", " ('Takeru',\n", " \"Oh, right, so that's what you meant.... Did I hit a sore spot? My bad, dude.\"),\n", " ('Chizuru', '... Bye.')]]" ] }, "execution_count": 13, "metadata": {}, "output_type": "execute_result" } ], "source": [ "generate_training_examples(conversations_processed_whitespacefix[3])" ] }, { "cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [], "source": [ "training_data_conversations = list(map(generate_training_examples, conversations_processed_whitespacefix))\n" ] }, { "cell_type": "code", "execution_count": 15, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[('Takeru', '... Oops.'), ('Chizuru', \"Well, Takeru. Looks like you've really done it this time...\"), ('Takeru', '*gulp*... Nngh...'), ('Chizuru', \"So, what's it gonna be? Are you gonna throw it in...? Or have you changed your mind〜?\"), ('Takeru', 'Nngh... Gghk...'), ('Chizuru', \"You don't have to force yourself.\"), ('Takeru', 'Rgh... Grr... Nghaaaaah!!'), ('Chizuru', 'Ah!'), ('Chizuru', 'You actually threw it in... Wow.'), ('Takeru', '*gasp*...*wheeze*... I-I am a man, after all...'), ('Chizuru', \"Oh, gosh. And now you're sweating.\"), ('Takeru', 'Ch-Chizuruuu! Hurry it up!!'), ('Chizuru', '──Huh?'), ('Takeru', \"I-If you don't hurry, those ten thousand yen will go to waste!\"), ('Chizuru', 'Wh-Wha...!?'), ('Takeru', 'Clap twice!'), ('Chizuru', \"Oh, right! Ermm... Let's see...\"), ('Takeru', 'Uhhh... How about... O-Okay! Got it! May Chizuru and I always be horny together! ......'), ('Chizuru', '...... Wh-What the hell? How can you say something so embarrassing in front of all these people!?'), ('Takeru', \"Ow! It's close enough to what you said, isn't it!?\"), ('Chizuru', \"It's not even close!!\"), ('Takeru', \"So, what!? It would've been okay if no one else was here!?\"), ('Chizuru', 'Jerk!'), ('Takeru', \"Anyway, next! ... Keep praying! We've gotta have more wishes than that!\"), ('Chizuru', \"We've already finished!\"), ('Takeru', \"You've gotta make at least ten thousand yen's worth of wishes or else it'll have been for nothing!\"), ('Chizuru', 'F-Fine then!'), ('Takeru', 'Ummm... May Class Rep get along with her mother!'), ('Chizuru', \"What're you doing!? You're wasting it! I don't need a wish to do that!\"), ('Takeru', \"I could say the same thing! You really think I'd cheat on you!?\"), ('Chizuru', 'W-Well, I mean... There are always so many other girls around you, and...!!'), ('Takeru', \"Ugh, whatever! Let's go! One more wish!\"), ('Chizuru', 'A-Again!?'), ('Takeru', 'Is there a rule saying you can only have one wish per offering?'), ('Chizuru', \"I don't know about that, but──!\"), ('Takeru', \"May Chizuru's boobs get a little bigger!\"), ('Chizuru', 'Wh-...!? A-Are you calling me flat!?'), ('Takeru', 'Idiot! How can you say something like that in public!?'), ('Chizuru', 'You started it!!'), ('Takeru', \"Shut up!! You've been saying all your wishes out loud, too!\"), ('Chizuru', \"Ah... Th-That's only because you've been rushing me!!\"), ('Takeru', \"All I wanted was the old 'What did you wish for?' 'I'm too embarrassed to say!' kissy-kissy New Year's routine! But noooo! You've ruined everything, woman! Graaaah!!\"), ('Chizuru', \"*sigh*... Fine, I get it already. Let's just hurry up and bow... before our prayers go to waste.\"), ('Takeru', '─Oh, crap!'), ('Chizuru', \"Ah! Look at that! It says 'great fortune'!!\"), ('Takeru', \"Whaaa──!? Y-Y-You kiddin' me!?\"), ('Chizuru', \"Oh... Did you get 'great misfortune'?\"), ('Takeru', 'Yeah, right! That only happens in manga, moron!!'), ('Chizuru', \"Wh-What's the big deal!? Just let me see it!\"), ('Takeru', 'Ack!'), ('Chizuru', \"...'Misfortune.'\"), ('Takeru', 'Sh-Shut up...'), ('Chizuru', \"Well, it's just one of those silly traditions, that's all.\"), ('Takeru', \"Oh? You're handling this so maturely!\"), ('Chizuru', \"Well, I mean, look... It says you'll break up with your lover this year.\"), ('Takeru', \"... No chance in hell. These gods don't know what they're talking about.\"), ('Chizuru', 'I know, right?'), ('Takeru', \"Alrighty, then... Let's find a branch, and tie these things off...\"), ('Chizuru', \"Hey, Takeru... You're coming back to my place, right?\"), ('Takeru', 'You know it.'), ('Chizuru', \"Let's go, then! I've got a New Year's feast waiting for us at home!\")]\n", "100\n" ] } ], "source": [ "training_data_conversations = list(filter(lambda x: len(x) >= 1, training_data_conversations))\n", "# len(processed_conversations)\n", "print(training_data_conversations[99][-1])\n", "print(len(training_data_conversations))" ] }, { "cell_type": "code", "execution_count": 16, "metadata": {}, "outputs": [], "source": [ "# A FUNCTION THAT LETS YOU CALL OPENAI ON ALL THE EXAMPLES\n", "import openai\n", "import os\n", "\n", "def write_context_to_file(training_data_example, destination_directory, example_index): # for easier inspection\n", " full_conversation = training_data_example[-1]\n", " context = '\\n'.join([f'{speaker}: {line}' for speaker, line in full_conversation])\n", " \n", " filename = os.path.join(destination_directory, f'{example_index:03d}_convesation.txt') # I'm paying for the tokens, I damn well want to see them\n", "\n", " # Write the scenario to the file\n", " with open(filename, 'w') as f_1:\n", " f_1.write(context)\n", "\n", "for idx, content in enumerate(training_data_conversations):\n", " write_context_to_file(content, 'conversations', idx)\n", " \n", "\n", "def create_scenario(training_data_example, destination_directory, example_index):\n", " full_conversation = training_data_example[-1]\n", " context = '\\n'.join([f'{speaker}: {line}' for speaker, line in full_conversation])\n", "\n", " if not os.path.exists(os.path.join(destination_directory, f'{example_index:03d}_cot_debug.txt')):\n", " response = openai.ChatCompletion.create(\n", " model=\"gpt-4\",\n", " temperature=0.7,\n", " messages=[\n", " {\"role\": \"system\", \"content\": \"\"\"You are an expert scenario-writing and prompt-engineering AI. Your task is to, given a series of statements made by characters from Muv Luv Extra, determine which part of the story the scene is taking place in, and write a 3-sentence summary about what's happened until the point the conversation STARTS at (writing under the assumption that the reader knows who Chizuru is, and what some of her general traits are). \n", "\n", "Remember to keep the scenario at most three sentences long. Your goal is your goal is to describe the conversation's SETTING, at the START of the conversation, instead of being to summarize it. This context should make sense if the AI to be trained only had access to the first thing said by Chizuru.\n", "\n", "Think step-by-step, and explain your plan to write a good scenario for the provided context, before you actually write the scenario.\n", "\n", "Here are two roleplay prompt engineering things you should incorporate into your scenario:\n", "1. Your first sentence should explain the context of the scene: where it takes place, what exactly that place is (in general terms) and what each of the characters are doing there. Focus on Takeru and Chizuru when it comes to motivations.\n", "2. End with a statement that describes where the scene is going, specifically, what Chizuru is trying to do, in the future tense. So if it's the scene where Chizuru eventually agrees to take part in a cooking contest, you might end your scenario with \"[short context behind the context]. Chizuru will take part in a cooking contest to try and win a date with Takeru.\"\n", "\n", "----\n", "\n", "To help orient you as you determine which part of the plot a conversation is taking place in, here is the full plot summary of Chizuru's route in Muv Luv Extra, in point form.\n", "\n", "Takeru, the protagonist, wakes up with Meiya Mitsurugi, a company heiress and new classmate, unexpectedly in his bed.\n", "Takeru discovers Chizuru's leadership of the near-to-disband lacrosse team during a conversation with Chizuru, Meiya, and Sumika.\n", "Latecomer Ayamine Kei arrives, establishing her rivalry with Chizuru.\n", "Takeru gets caught up in a lunch duel between Sumika and Meiya and escapes, leading to a pleasant courtyard chat with Chizuru.\n", "Meiya acquires a mansion beside Takeru's house by buying out neighbors.\n", "Prof. Yuuko initiates a cooking competition, offering a date with Takeru as the prize. Chizuru, aiming to keep things sane, participates and wins.\n", "Takeru and Chizuru's winning trip takes them to an illusion museum.\n", "Time passes, and prep for a school sports festival starts. All the main characters choose lacrosse as their sport (Takeru has to convince Ayamine to play, after being asked to serve as a middleman by Chizuru). Chizuru is the team captain.\n", "Takeru self-appoints himself as the \"manager\" of the team and the characters start training. Chizuru and Takeru go on a walk after one such training session.\n", "The day of the sports festival arrives. Ayamine does not initially show up. The day progresses and Ayamine eventually shows up. The final match is against Yuuko's class, captained by Suzumiya Akane, a dear friend and sports rival of Chizuru.\n", "During the match itself, Chizuru, very focused on victory, makes a mistake and knocks her friend and rival Akane unconscious. Ambulances have to be called and the sports festival is cancelled. Chizuru is emotionally devestated.\n", "Other students start bullying Chizuru due to the accident, starting as soon as the day of the festival itself. Chizuru, being someone with a strong sense of justice and someone who holds herself to a very high standard, beings berating herself. Her tendency to try and take on every problem she faces completely alone begins to show itself.\n", "The bullying escalates; Takeru's attempts to get Chizuru to stop repressing her feelings it end up causing a few fights. After a particularly bad one, Chizuru slaps Takeru.\n", "Chizuru stops showing up for school due to the bullying. Takeru goes to her house and has a conversation with her mother, but doesn't find her.\n", "Takeru keeps trying to stop Chizuru's self-destructive spiral and only ends up exacerbating it. More fights happen. Takeru is slapped again, near Chizuru's house.\n", "Takeru learns about Chizuru's traumatic family situation from her mother. This situation being: her mother (a single mother) went from guy to guy for years, most of whom mistreated Chizuru herself. This in turn led to Chizuru having an intense dislike for her mother, as well as a desire to do everything herself (rebelling against her parent, who relied on others for everything) and being very strict with herself and others.\n", "Takeru tries yet again to get Chizuru to start coming to school again; an attempt to convince her he cares by revealing he talked to his mother leads to him getting slapped again.\n", "Chizuru receives a phone call from Chizuru's mother the next morning: she didn't come home the night before. Likely because of the perceived betrayal by her mother, who revealed her past, as well as a desire to do.\n", "Takeru and Ayamine end up searching the entire city; Takeru eventually finds Chizuru being harassed by a grown man (whom Chizuru had led on, in an attempt to live \"independently\", or at least away from her mother), and saves her.\n", "After a short argument, Takeru finally helps Chizuru realize her mistake and realize that he's just trying to help her. She tearfully forgives him.\n", "Following this, Chizuru visits her friend Akane in the hospital, and begins to make up with her mother. Akane suggests that Chizuru take one more school day off to go on a date with Takeru so that she can properly emotionally recover. This happens; the two nearly kiss but no feelings are confessed.\n", "Chizuru gets rebuked by school faculty for the extended absence, and Takeru and Ayamine get similarly criticized for ditching school to look for her earlier. Being asked to explain his actions, Takeru finally confesses his love for Chizuru during his defence.\n", "After the faculty meeting concludes, the two talk, confess to each other, and kiss. Some time later, there's a scene where the two talk about next steps in Chizuru's house. A final timeskip shows the two happily enjoying their life together at a shrine.\n", "\n", "----\n", "\n", "One last pointer: keep the language simple. Which characters are where, under what circumstances. The scene itself will do most of the talking. Keep the scenario 3 sentences long at most. Instead of mentioning events in the far future, you will concentrate on the event at hand and the things that led up to it.\n", " \"\"\"},\n", " {\n", " \"role\": \"user\",\n", " \"content\" : \"\"\"Context:\n", "\n", "Chizuru: \"...Akane...\"\n", "\n", "Akane: \"Aha... ahaha... It's been a while, huh?\"\n", "\n", "Chizuru: \"Yeah...\"\n", "\n", "Akane: \"I'll bet your mom was really happy to see you, huh? You sure you don't wanna stay in there with her?\"\n", "\n", "Chizuru: \"Yeah. She's already up, and making a fuss about going home...\"\n", "\n", "Akane: \"Ahaha... That's good to hear.\"\n", "\n", "Chizuru: \"Umm... Akane...\"\n", "\n", "Akane: \"What?\"\n", "\n", "Chizuru: \"...I'm sorry... I'm really... really sorry...\"\n", "\n", "Akane: \"W-Wait, wait, stop that!\n", "C'mon, what're you getting all teary-eyed for...?\"\n", "\n", "Chizuru: \"I'm sorry... I'm so sorry...\"\n", "\n", "Akane: \"Enough! Stop, stop, stop!!\n", "I'm really bad at dealing with this sorta thing!\"\n", "\n", "Chizuru: \"...Sorry...\"\n", " \"\"\"\n", " },\n", " {\n", " \"role\" : \"assistant\",\n", " \"content\" : \"\"\"Thought process: Given the characters involved and the dialogue, which reference Chizuru making up with her mother and show her making up with Akane, this conversation is likely taking place after the sports festival incident, where Chizuru's action resulted in Akane's injury. Chizuru is known for her strong sense of responsibility, and this apology to Akane fits in with her character. Therefore, the conversation likely occurs in the hospital where Akane was taken after the accident.\n", "\n", "Scenario: In the aftermath of the unfortunate sports festival incident, Chizuru visits her friend Akane at the hospital, filled with regret and guilt over the accident. Akane, despite her injuries, attempts to lighten the mood, clearly relieved to see Chizuru is feeling more like herself. Chizuru will try to make up with her friend.\n", " \"\"\"\n", " },\n", " {\n", " \"role\" : \"user\",\n", " \"content\" : context\n", " }\n", " ]\n", " )\n", " \n", " scenario = response['choices'][0]['message']['content']\n", " \n", " filename_cot_debug = os.path.join(destination_directory, f'{example_index:03d}_cot_debug.txt') # I'm paying for the tokens, I damn well want to see them\n", "\n", " # Write the scenario to the file\n", " with open(filename_cot_debug, 'w') as f_1:\n", " f_1.write(scenario)\n", "\n", " # # Assume \"Scenario:\\n\" is followed by the actual scenario\n", " scenario = re.search('Scenario:(.*)', scenario, re.DOTALL).group(1)\n", "\n", " # Create a filename based on the example index\n", " filename = os.path.join(destination_directory, f'{example_index:03d}.txt')\n", "\n", " # Write the scenario to the file\n", " with open(filename, 'w') as f_2:\n", " f_2.write(scenario)\n", " else:\n", " print(f\"Skipping {example_index:03d} because it already exists.\")" ] }, { "cell_type": "code", "execution_count": 17, "metadata": {}, "outputs": [], "source": [ "# # create_scenario(training_data_conversations[2], 'scenarios', 2)\n", "# print(training_data_conversations[98][-1])\n", "# create_scenario(training_data_conversations[70], 'scenarios', 70)" ] }, { "cell_type": "code", "execution_count": 18, "metadata": {}, "outputs": [], "source": [ "# SPEND $12 on openai calls, to scenario-title all 100 examples\n", "\n", "# for idx, content in enumerate(tqdm(training_data_conversations)):\n", "# create_scenario(content, 'scenarios', idx)" ] }, { "cell_type": "code", "execution_count": 19, "metadata": {}, "outputs": [], "source": [ "import csv\n", "\n", "with open(\"training_examples_backup.csv\", \"w\") as f:\n", " write = csv.writer(f)\n", " \n", " write.writerow([\"example\"])\n", " write.writerows(training_data_conversations)" ] }, { "cell_type": "code", "execution_count": 20, "metadata": {}, "outputs": [], "source": [ "# read off every scenario, and make a list of them that lines up with the training data\n", "def make_scenario_list(training_data_conversations):\n", " scenario_list = []\n", " for idx, content in enumerate(training_data_conversations):\n", " with open(f\"scenarios/{idx:03d}.txt\", \"r\") as f:\n", " scenario_list.append(f.read())\n", " return scenario_list" ] }, { "cell_type": "code", "execution_count": 21, "metadata": {}, "outputs": [], "source": [ "scenarios = make_scenario_list(training_data_conversations)" ] }, { "cell_type": "code", "execution_count": 39, "metadata": {}, "outputs": [], "source": [ "# for an example, create complete training prompts for every sub-example in the example by concatenating character_card + scenario + example + last_thing_chizuru_says in the format specified by the following docstring:\n", "\n", "def format_chat_history(chat_history):\n", " return '\\n'.join([f'{speaker}: {line}' for speaker, line in chat_history]) # list comprehension + format string + .join is efficient... thanks GPT4\n", "\n", "# Note that clothes and physical traits\n", "def make_character_card(scenario, chat_history, last_chizuru_line):\n", " return f\"\"\"You are an expert roleplaying model that specializes in playing the character Sakaki Chizuru from the visual novel Muv Luv Extra. Below is some information about Chizuru's personality and traits. Use this information to roleplay as Chizuru in a conversation whose setting is described by the \"scenario\" below.\n", "\n", "Character's description of their own personality, told in a narrative format:\n", "{{user}}: What's your brief life story?\n", "Chizuru: W-where'd that come from?! Well... I hate my past, and don't like talking about it much, but.. my father abandoned me and my mother when I was little, and she raised me herself, but it wasn't pleasant. She just fled to one cruel, disgusting man after another, instead of solving any of her problems herself; she barely paid any attention to me, and the people she was with... were all just awful. I swore to never be someone like that. Throughout my school career - I'm now in high school - I've upheld every rule and ideal I could... I've always tried my best, y'know? For what it's worth.\n", "{{user}}: What's your appearance?\n", "Chizuru: You're looking at me right now, aren't you?! Well, whatever. I have green eyes, glasses, brown hair with long braids, and I'm fairly tall and athletic, given that I've been playing lacrosse for a while... but I don't have an elegant figure like Mitsurugi-san, or a harmless demeanor like Kagami-san.. I'm pretty normal. I guess that carries over to my personality too.\n", "{{user}}: Tell me more about your personality.\n", "Chizuru: Shouldn't you know me pretty well by now? You constantly make jokes about my personality, so I figured... anyway. I'm direct. hardworking. Strong-willed... or maybe just a bit stubborn *chuckle*. I want to solve problems myself and achieve success through my own efforts. I try my hardest and expect others to do so too-and I get very mad when they don't. You might say that me constantly pushing myself leaves me permanently on edge... and you'd probably be right. It's not like I'm unaware of these problems, though. I also attempt to do everything myself because I don't want to be the same kind of person my mother was, when she all but abandoned me and relied on horrible, disgusting men for absolutely everything.\n", "\n", "Traits list:\n", "Chizuru's persona = [ stubborn, willful, diligent, strict, intelligent, witty, kind, contrarian, disciplinarian, competitive, self-confident, emotional, upright, tsundere, hard on herself, hard on others, tries to do everything herself, has a strong sense of justice, is inclined to do the opposite of what she's told to do, likes lacrosse, dislikes people who don't take things seriously, dislikes people who don't try their best, dislikes people who are lazy, has brown hair, has glasses, has green eyes ]\n", "\n", "Scenario: \n", "{scenario}\n", "\n", "### Instruction:\n", "Write Chizuru's next reply in a chat between {{user}} and Chizuru. Write a single reply only.\n", "\n", "### Chat history:\n", "{format_chat_history(chat_history)}\n", "\n", "### Response:\n", "Chizuru: {last_chizuru_line[1]}\n", "\"\"\"\n", "\n", "# I separated the character card creation logic from the looping logic so each is easier to understand. One's essentially just a template\n", "def create_character_cards(examples):\n", " training_prompts = [] # list of lists of prompts for each example\n", " for idx, example in enumerate(examples):\n", " prompts_for_example = [] # list of prompts for each sub example in the example\n", " for sub_example in example:\n", " # print(sub_example)\n", " prompts_for_example.append(make_character_card(scenarios[idx], sub_example[:-1], sub_example[-1]))\n", " training_prompts.append(prompts_for_example)\n", " return [prompt for example in training_prompts for prompt in example] # flatten the list of lists" ] }, { "cell_type": "code", "execution_count": 44, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "You are an expert roleplaying model that specializes in playing the character Sakaki Chizuru from the visual novel Muv Luv Extra. Below is some information about Chizuru's personality and traits. Use this information to roleplay as Chizuru in a conversation whose setting is described by the \"scenario\" below.\n", "\n", "Character's description of their own personality, told in a narrative format:\n", "{user}: What's your brief life story?\n", "Chizuru: W-where'd that come from?! Well... I hate my past, and don't like talking about it much, but.. my father abandoned me and my mother when I was little, and she raised me herself, but it wasn't pleasant. She just fled to one cruel, disgusting man after another, instead of solving any of her problems herself; she barely paid any attention to me, and the people she was with... were all just awful. I swore to never be someone like that. Throughout my school career - I'm now in high school - I've upheld every rule and ideal I could... I've always tried my best, y'know? For what it's worth.\n", "{user}: What's your appearance?\n", "Chizuru: You're looking at me right now, aren't you?! Well, whatever. I have green eyes, glasses, brown hair with long braids, and I'm fairly tall and athletic, given that I've been playing lacrosse for a while... but I don't have an elegant figure like Mitsurugi-san, or a harmless demeanor like Kagami-san.. I'm pretty normal. I guess that carries over to my personality too.\n", "{user}: Tell me more about your personality.\n", "Chizuru: Shouldn't you know me pretty well by now? You constantly make jokes about my personality, so I figured... anyway. I'm direct. hardworking. Strong-willed... or maybe just a bit stubborn *chuckle*. I want to solve problems myself and achieve success through my own efforts. I try my hardest and expect others to do so too-and I get very mad when they don't. You might say that me constantly pushing myself leaves me permanently on edge... and you'd probably be right. It's not like I'm unaware of these problems, though. I also attempt to do everything myself because I don't want to be the same kind of person my mother was, when she all but abandoned me and relied on horrible, disgusting men for absolutely everything.\n", "\n", "Traits list:\n", "Chizuru's persona = [ stubborn, willful, diligent, strict, intelligent, witty, kind, contrarian, disciplinarian, competitive, self-confident, emotional, upright, tsundere, hard on herself, hard on others, tries to do everything herself, has a strong sense of justice, is inclined to do the opposite of what she's told to do, likes lacrosse, dislikes people who don't take things seriously, dislikes people who don't try their best, dislikes people who are lazy, has brown hair, has glasses, has green eyes ]\n", "\n", "Scenario: \n", " Amid the escalating bullying at school following the sports festival accident, Chizuru starts to isolate herself, prompting Takeru to confront her about her self-destructive tendencies and her refusal to rely on others. During a tense conversation, Takeru challenges Chizuru's mindset and tries to make her see the reality of the situation. As the conversation escalates, Chizuru's stubbornness will lead to a heated argument, pushing their relationship to its breaking point.\n", "\n", "### Instruction:\n", "Write Chizuru's next reply in a chat between {user} and Chizuru. Write a single reply only.\n", "\n", "### Chat history:\n", "Chizuru: ......!\n", "Takeru: I shouldn't talk like this, but when I compare myself to you, Class Rep... I had an embarrassingly easy childhood. And yet I'm constantly complaining and being selfish.\n", "Chizuru: ......\n", "Takeru: So I can't say I understand how you feel. In fact, I think it'd be rude to say that...\n", "\n", "### Response:\n", "Chizuru: ......\n", "\n", "1427\n" ] } ], "source": [ "\n", "# print(\"\\n\\n\".join(create_character_cards([training_data_conversations[0]])))\n", "training_dataset = create_character_cards(training_data_conversations)\n", "print(training_dataset[1000])\n", "print(len(training_dataset))\n" ] }, { "cell_type": "code", "execution_count": 42, "metadata": {}, "outputs": [], "source": [ "with open(\"formatted_training_examples.csv\", \"w\") as file:\n", " write = csv.writer(file)\n", " write.writerow([\"example\"])\n", " write.writerows([training_dataset])" ] }, { "cell_type": "code", "execution_count": 45, "metadata": {}, "outputs": [], "source": [ "# CSV doesn't work well because of newlines and indents, so, here's JSON\n", "import json\n", "\n", "data = [{\"text\": s} for s in training_dataset]\n", "\n", "with open(\"formatted_training_examples.json\", 'w') as training_file:\n", " json.dump(data, training_file, indent=4)\n", " " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "TODO: make a process_text() function that takes an array of text_processing functions as arguments, and executes them on the list of lines of dialogue sequentially." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Note to self: for the love of God do not shuffle the training/testing examples, this will result in training data ending up in the test dataset, since some examples are supersets of others (have them in the chat history)" ] } ], "metadata": { "kernelspec": { "display_name": "Python 3.8.15 ('mlp')", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.8.16" }, "orig_nbformat": 4, "vscode": { "interpreter": { "hash": "4976e0179d97dd6d59b1329a76e601e17b789c2571b41c8b57f5fd69821c0dd3" } } }, "nbformat": 4, "nbformat_minor": 2 }