This post is related to:
- BART configuration parameters overview
- BERT configuration parameters overview
- Exploring OpenAI models locally without APIs (DRAFT-GUIDE)
| Parameter | Description | Location in Documentation |
|---|---|---|
| pad_token_id | The token ID used for padding sequences. This value indicates to the model where padding occurs. | Check the tokenizer documentation: Hugging Face Tokenizers |
| eos_token_id | The token ID that signifies the end of a sequence (End Of Sequence). | Check model documentation for specifics: Hugging Face Models |
| attention_mask | A mask that indicates which tokens should be attended to by the model (1 for actual tokens, 0 for padding). | Refer to model-specific usage in the Hugging Face library: Transformers Usage |
| input_ids | The token IDs representing the input sequence of text. | General model input specifications: Hugging Face Model Input |
| output_attention | A boolean flag indicating if the model should return attention weights in addition to outputs. | Detailed in specific model API docs: Transformers API |
| max_length | The maximum length of sequences for generation tasks, limiting how long the output can be. | Check generation section in model docs: Text Generation |
| num_return_sequences | The number of sequences to return from the model output, useful for generating multiple responses. The number of independently computed returned sequences for each element in the batch. | Refer to the text generation configuration: Text Generation |
| temperature | Strictly positive float value used to modulate the logits distribution. A value smaller than 1 decreases randomness (and vice versa), with 0 being equivalent to shifting all probability mass to the most likely token. | Discussed in generation options: Generation Parameters |
| top_k | The number of highest probability vocabulary tokens to keep for sampling. | Detailed in sampling strategies: Sampling Methods |
| top_p | Controls the cumulative probability distribution for nucleus sampling. | Explained in sampling calculations: Nucleus Sampling |