Make batch API calls with caching. Args: system_contents: List of system prompts message_contents: List of processed mes

Make batch API calls with caching. Args: system_contents: List of system prompts message_contents: List of processed mes

Make batch API calls with caching. Args: system_contents: List of system prompts message_contents: List of processed mes

Clear all cache files

This function is used to convert the custom configurations to dict.

Decode base64 string to PIL Image object and ensure consistent format

returns either a context list that includes one that will disable zero.Init or an empty context list

Convert image to base64 string with consistent format

Encode video to base64 string by extracting and processing frames. This function takes a video input (URL, file path, list of images

Evaluate the model on the evaluation dataset.

Returns the largest integer less than or equal to 'number' that is divisible by 'factor'.

Formatting input sample, and change the related keys according to the training dataset.

Formatting input sample, and change the related keys according to the training dataset.

Load pre-trained model and tokenizer from a given path.

Load pre-trained model and tokenizer from a given path.

Load pre-trained model and tokenizer from a given path.

Print a message only on the main process.

Print a table in the main process.

Perform a single update step with PTX loss.

Run benchmark Args: task_list (list[str]): task list Returns: evaluation_details (dict[str, list[EvaluationRe

Save model and tokenizer in Hugging Face format.

Train the model.

Training main loop

Train the model.

Evaluate the model on the validation dataset.

Inputs: subset_dir: path to the directory containing subdirectories with hdf5_sensors.hdf5 files

Return a string representation of the dataset.

Get the modality type of current task. Returns: str: Modality type (multimodal, text-to-text, etc.)

Compute the log-probabilities of the model on the specified tokens.

Load and cache HuggingFace model. Args: kwargs: Additional keyword arguments for initializing HF models. Returns:

Load and cache HuggingFace tokenizer.

Visualize bounding boxes on an image and save the image to disk. Parameters: - frame: numpy array of shape (height, width, 3) representi

Visualize bounding boxes on an image and save the image to disk. Parameters: - frame: numpy array of shape (height, width, 3) representi

Analyze preference distribution

Match the specified pattern in the text. Args: data (list/iterable): the text to be matched Returns:

Make API requests with caching to avoid duplicate calls. Args: messages: List of message dictionaries to send to the API

Make API requests with caching to avoid duplicate calls. Args: messages: List of message dictionaries to send to the API

Calculate additional evaluation metrics

Calculate Pearson correlation coefficient with confidence interval

检查模型文件是否已存在