Implements the actions of the prompt process, which includes the actions pre-processing, execution, post-processing, and managing the state of related inferences. ``Prompt`` is responsible for pre-processing, executing, and post-processing of inferences and for managing end-to-end state
| 42 | |
| 43 | |
| 44 | class Prompt: |
| 45 | |
| 46 | """Implements the actions of the prompt process, which includes the actions pre-processing, execution, |
| 47 | post-processing, and managing the state of related inferences. |
| 48 | |
| 49 | ``Prompt`` is responsible for pre-processing, executing, and post-processing of inferences and for |
| 50 | managing end-to-end state of a series of related inferences. |
| 51 | |
| 52 | Parameters |
| 53 | ---------- |
| 54 | llm_name : str, default=None |
| 55 | The name of the llm to be used. |
| 56 | |
| 57 | tokenizer : object, default=None |
| 58 | The tokenizer to use. The default is to use the tokenizer specified by the ``Utilities`` class. |
| 59 | |
| 60 | model_card : dict, default=None |
| 61 | A dictionary describing the model to be used. If the dictionary contains the key ``model_name``, |
| 62 | then this model will be used instead of the one set by ``llm_name``. In other words, ``model_card`` |
| 63 | overrides ``llm_name``. |
| 64 | |
| 65 | library : object, default=None |
| 66 | A ``Library`` object. |
| 67 | |
| 68 | account_name : str, default="llmware" |
| 69 | The name of the account to be used. This is one of the attributes of the prompt. |
| 70 | |
| 71 | prompt_id : int, default=None |
| 72 | The ID of the prompt. If a prompt ID is given, then the state of this prompt is loaded. Otherwise, a |
| 73 | new prompt ID is generated. This is part of the state of a prompt. |
| 74 | |
| 75 | save_state : bool, default=True |
| 76 | Actually, this is a dead variable and should be removed in a future release. |
| 77 | |
| 78 | llm_api_key : str, default=None |
| 79 | The API key that is used to load the large language model. |
| 80 | |
| 81 | llm_model : str, default=None |
| 82 | The name of the model to load. |
| 83 | |
| 84 | from_hf : bool, default=False |
| 85 | Indicates whether the model should be loaded from hugging face. |
| 86 | |
| 87 | prompt_catalog : object, default=None |
| 88 | An object of type ``PromptCatalog``. |
| 89 | |
| 90 | temperature : float, default=0.5 |
| 91 | Sets the temperature of the large language model. |
| 92 | |
| 93 | prompt_wrapper : str, default="human_bot" |
| 94 | Sets the prompt wrapper. Possible values are "alpaca", "human_bot", "chatgpt", "<INST>", "open_chat", |
| 95 | "hf_chat", and "chat_ml". |
| 96 | |
| 97 | instruction_following : bool, default=False |
| 98 | Sets whether the large language model should follow instructions. Note that this has an effect |
| 99 | if and only if the model specified has a version that is trained to follow instructions. |
| 100 | |
| 101 | """ |
no outgoing calls