--- language: - en license: wtfpl library_name: peft pipeline_tag: text-generation base_model: tiiuae/falcon-7b --- ## Model description The tiiuae/falcon-7b model finetuned for Paraphrasing, Changing the Tone of the input sentence(to casual/professional/witty), Summary and Topic generation from a dialogue. Data for Paraphrasing and Changing the Tone was generated using gpt-35-turbo and a sample of roughly 1000 data points from the [Dialogsum](https://github.com/cylnlp/dialogsum) dataset was used for Summary and Topic generation. Look at the repo [llm-toys](https://github.com/kuutsav/llm-toys) for usage and other details. Try in colab (you might need the pro version): Open In Colab ## Installation ```bash pip install llm-toys ``` ```python from llm_toys.tasks import GeneralTaskAssitant from llm_toys.config import TaskType gta = GeneralTaskAssitant() gta.complete(TaskType.PARAPHRASE_TONE, "Hey, can yuo hepl me cancel my last order?") # "Could you assist me in canceling my previous order?" gta.complete(TaskType.PARAPHRASE_TONE, "Hey, can yuo hepl me cancel my last order?", tone="casual") # "Hey, can you help me cancel my last order?" gta.complete(TaskType.PARAPHRASE_TONE, "Hey, can yuo hepl me cancel my last order?", tone="professional") # "I would appreciate if you could assist me in canceling my previous order." gta.complete(TaskType.PARAPHRASE_TONE, "Hey, can yuo hepl me cancel my last order?", tone="witty") # "Oops! Looks like I got a little carried away with my shopping spree. Can you help me cancel my last order?" chat = """ #Person1#: I'm so excited for the premiere of the latest Studio Ghibli movie! #Person2#: What's got you so hyped? #Person1#: Studio Ghibli movies are pure magic! The animation, storytelling, everything is incredible. #Person2#: Which movie is it? #Person1#: It's called "Whisper of the Wind." It's about a girl on a magical journey to save her village. #Person2#: Sounds amazing! I'm in for the premiere. #Person1#: Great! We're in for a visual masterpiece and a heartfelt story. #Person2#: Can't wait to be transported to their world. #Person1#: It'll be an unforgettable experience, for sure! """.strip() gta.complete(TaskType.DIALOGUE_SUMMARY_TOPIC, chat) # {"summary": "#Person1# tells #Person2# about the upcoming Studio Ghibli movie. # #Person1# thinks it's magical and #Person2#'s excited to watch it.", # "topic": "Movie premiere"} ``` ## Sample training data ```json [ { "original": "If you have any further questions, feel free to ask.", "casual": "Got more questions? Feel free to ask away. I'm here to help!", "professional": "Should you have any additional inquiries, please don't hesitate to ask.", "witty": "Curiosity is always in style! If you have more mysteries to solve, I'm all ears!", "paraphrase": "Don't hesitate to ask if you have any more questions." }, { "fname": "dev_473", "dialogue": "#Person1#: Did you enjoy your weekend at the highland hotel? I heard it's and excellent place to stay and has good facilities.\n#Person2#: I had a wonderful time. The rooms are not very big, but they are well furnished. The restaurant is excellent and reasonably priced. There's a sauna and a Jacuzzi.\n#Person1#: Do they have a swimming pool?\n#Person2#: No, they don't. they have a beauty parlor, but I didn't go there.\n#Person1#: What's the service like?\n#Person2#: It's very good. Check in and check out at the reception only took a few minutes. The wait staff is very good. A waiter recommended their baked fish, which tasted wonderful. The hotel was quite full, so I'd suggest making a reservation if you intend to go there. The hotel offers a discount at the weekends.\n#Person1#: It sounds perfect. Did you have any complaints at all?\n#Person2#: There was a problem with the internet access, so I couldn't check my email, but I didn't complain about it to the management.\n#Person1#: I suppose you were happy to forget about the outside world.\n#Person2#: Yes, I was. Here's their business card.\n#Person1#: Thanks. Was there a mina bar in the room?\n#Person2#: No, there wasn't. There is a bar on the ground floor and of course you can buy drinks in the restaurant to go with your meal.\n#Person1#: One of the things I dislike about hotels is that everyone expects tips.\n#Person2#: I know. At the inland hotel, they have an interesting policy. When you check out, you put some money in a special box at reception. Each evening, the money in the box is shared equally by the hotel staff.", "summary": "#Person2# enjoys #Person2#'s weekend at the highland hotel because of the hotel's excellent and reasonably priced restaurant and good service. #Person2# introduces the hotel's facilities, weekend discount, and its interesting tip policy and suggests #Person1# make a reservation in advance.", "topic": "Experience in hotel" } ] ``` ## Training params ```json { "batch_size": 1, "eval_ratio": 0.05, "eval_steps": 100, "gradient_accumulation_steps": 4, "learning_rate": 0.0001, "logging_steps": 100, "lora_alpha": 32, "lora_dropout": 0.05, "lora_r": 16, "max_length": 1024, "model_name": "tiiuae/falcon-7b", "num_train_epochs": 3, "seed": 10, "task_type": "paraphrase_tone,dialogue_summary_topic", "use_aim": True } ``` ## Training curve ![train_eval_loss](falcon-7b-paraphrase-tone-dialogue-summary-topic.jpeg) ## Training procedure The following `bitsandbytes` quantization config was used during training: - load_in_8bit: False - load_in_4bit: True - llm_int8_threshold: 6.0 - llm_int8_skip_modules: None - llm_int8_enable_fp32_cpu_offload: False - llm_int8_has_fp16_weight: False - bnb_4bit_quant_type: nf4 - bnb_4bit_use_double_quant: True - bnb_4bit_compute_dtype: bfloat16 ### Framework versions - PEFT 0.4.0.dev0