--- library_name: transformers license: apache-2.0 datasets: - isek-ai/danbooru-tags-2023 base_model: p1atdev/dart-v1-base tags: - trl - sft - optimum - danbooru inference: false --- # Dart (Danbooru Tags Transformer) v1 This model is a fine-tuned Dart (**Da**nboo**r**u **T**ags Transformer) model that generates danbooru tags. Demo: [🤗 Space](https://huggingface.co/spaces/p1atdev/danbooru-tags-transformer) If you are a developer and want to finetune, it's recommended using the base version, [p1atdev/dart-v1-base](https://huggingface.co/p1atdev/dart-v1-base), instead ## Usage ### Using AutoModel 🤗 Transformers library is required. ```bash pip install -U transformers ``` ```py import torch from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig MODEL_NAME = "p1atdev/dart-v1-sft" tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, trust_remote_code=True) # trust_remote_code is required for tokenizer model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype=torch.bfloat16) prompt = "<|bos|>rating:sfw, rating:generaloriginal<|long|>1girl<|input_end|>" inputs = tokenizer(prompt, return_tensors="pt").input_ids with torch.no_grad(): outputs = model.generate(inputs, generation_config=generation_config) print(tokenizer.decode(outputs[0], skip_special_tokens=True)) # rating:sfw, rating:general, 1girl, ahoge, braid, closed eyes, collared dress, dress, flower, full body, hair flower, hair ornament, long hair, night, night sky, outdoors, parted lips, pink flower, pink hair, short sleeves, sky, solo, straight hair, sunflower, very long hair, white flower ``` You can use `tokenizer.apply_chat_template` to simplify constructiing of prompts: ```py inputs = tokenizer.apply_chat_template({ "rating": "rating:sfw, rating:general", "copyright": "original", "character": "", "general": "1girl", "length": "<|long|>" }, return_tensors="pt", tokenize=True) # tokenize=False to preview prompt # same as input_ids of "<|bos|>rating:sfw, rating:generaloriginal<|long|>1girl<|input_end|>" with torch.no_grad(): outputs = model.generate(inputs, generation_config=generation_config) ``` See [chat_templating document](https://huggingface.co/docs/transformers/main/en/chat_templating) for more detail about `apply_chat_template`. #### Flash attention (optional) Using flash attention can optimize computations, but it is currently only compatible with Linux. ```bash pip install flash_attn ``` ### Accelerate with ORTModel 🤗 Optimum library is also compatible, for the high performance inference using ONNX. ```bash pip install "optimum[onnxruntime]" ``` Two ONNX models are provided: - [Normal](./model.onnx) - [Quantized](./model_quantized.onnx) Both can be utilized based on the following code: ```py import torch from transformers import AutoTokenizer, GenerationConfig from optimum.onnxruntime import ORTModelForCausalLM MODEL_NAME = "p1atdev/dart-v1-sft" tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, trust_remote_code=True) # normal version ort_model = ORTModelForCausalLM.from_pretrained(MODEL_NAME) # qunatized version # ort_model = ORTModelForCausalLM.from_pretrained(MODEL_NAME, file_name="model_quantized.onnx") inputs = tokenizer.apply_chat_template({ "rating": "rating:sfw, rating:general", "copyright": "original", "character": "", "general": "1girl", "length": "<|long|>" }, return_tensors="pt", tokenize=True) with torch.no_grad(): outputs = ort_model.generate(inputs, generation_config=generation_config) print(tokenizer.decode(outputs[0], skip_special_tokens=True)) ``` ### Prompt guide Due to training with a specialized prompt format, **natural language is not supported**. The trained sentences are essentially composed of the following elements, arranged in the strict order shown below: - `<|bos|>`: The bos (begin of sentence) token - `[RATING_PARENT], [RATING_CHILD]`: The block of rating tags - [RATING_PARENT]: `rating:sfw`, `rating:nsfw` - [RATING_CHILD]: - if `[RATING_PARENT]` is `rating:sfw`: `rating:general`, `rating:sensitive` - else: `rating:questionable`, `rating:explicit` - `[COPYRIGHT, ...]`: The block of copyright tags. - [COPYRIGHT, ...]: All supported copyright tags can be seen in [TODO]() - `[CHARACTER, ...]`: The block of character tags. - [CHARACTER, ...]: All supported character tags can be seen in [TODO]() - `[LENGTH_TOKEN][GENERAL, ...]<|input_end|>[COMPLETION]`: The block of general tags. - [LENGTH_TOKEN]: A token to specify **total** amount of general tags. - Avaiable: - `<|very_short|>`: less than 10 tags - `<|short|>`: less than 20 tags - `<|long|>`: less than 40 tags (recommended) - `<|very_long|>`: more than 40 tags - [GENERAL, ...]: All supported general tags can be seen in [TODO]() - `<|input_end|>`: A tag to show the end of input. Set this token at last of prompt. - [COMPLETION]: The model complete tags in alphabetical order. - `<|eos|>`: The eos (end of sentence) token - Tags other than special tokens are separated by commas. - You can place tags in any order you like in each block. Example sentence: ``` <|bos|>rating:sfw, rating:generalvocaloidhatsune miku<|long|>solo, 1girl, very long hair<|input_end|>blue hair, cowboy shot, ...<|eos|> ``` Therefore, to complete the tags, the input prompt should be as follows: 1. without any copyright and character tags ``` <|bos|>rating:sfw, rating:general<|very_long|>1girl, solo, cat ears<|input_end|> ``` 2. specifing copyright and character tags ``` <|bos|>rating:sfw, rating:generalsousou no frierenfrieren<|long|>1girl, solo, from side<|input_end|> ``` ## Model Details ### Model Description - **Developed by:** Plat - **Model type:** Causal language model - **Language(s) (NLP):** Danbooru tags - **License:** Apache-2.0 - **Demo:** Avaiable on [🤗Space](https://huggingface.co/spaces/p1atdev/danbooru-tags-transformer) ## Bias, Risks, and Limitations Since this model is a pre-trained model, it cannot accommodate flexible specifications. ## Training Details ### Training Data This model was trained with: - [isek-ai/danbooru-tags-2023](https://huggingface.co/datasets/isek-ai/danbooru-tags-2023): 6M size of danbooru tags dataset since 2005 to 2023 Only data from 2020 onwards was used for SFT. ### Training Procedure Trained using 🤗 transformers' trainer. #### Preprocessing Preprocessing was conducted through the following process: 1. Remove data where `general` tags is null. 2. Remove `general` tags that appear less than 100 times. 3. Remove undesirable tags such as `watermark` and `bad anatomy`. 4. Remove based on the number of tags attached to a single post (following rules): - Remove if more than 100 for `general` tags. - Remove if more than 5 for `copyright` tags. - Remove if more than 10 for `character` tags. 5. Remove posts created before 2020 6. Set length token according to each tags length 7. Shuffle some tags in the following rule: - Include people tags (e.g. `1girl`, `no humans`) tags in the shuffle-group with a 95% probability, and do not do so with a 5% probability. - Get tags at a random percentage between 0% and 75% to create a shuffle-group. - Shuffle tags in shuffle-group and concatnate with `<|input_end|>` token and remains in alphabetical order. 8. Concatnate all categories #### Training Hyperparameters The following hyperparameters were used during training: - learning_rate: 0.0001 - train_batch_size: 32 - eval_batch_size: 8 - seed: 42 - gradient_accumulation_steps: 2 - total_train_batch_size: 64 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 500 - num_epochs: 1 ## Evaluation Evaluation has not been done yet and it needs to evaluate. ## Technical Specifications ### Model Architecture and Objective The architecture of this model is [OPT (Open Pretrained Transformer)](https://huggingface.co/docs/transformers/model_doc/opt), but the position embeddings was not trained. ### Compute Infrastructure In house #### Hardware 1x RTX 3070 Ti #### Software - Dataset processing: [🤗 Datasets](https://github.com/huggingface/datasets) - Training: [🤗 Transformers](https://github.com/huggingface/transformers) - Optimizing: [🤗 Optimum](https://github.com/huggingface/optimum) - SFT: [🤗 TRL](https://github.com/huggingface/trl) ## More Information [optional] [More Information Needed]