r""" Gets the merged datasets in the standard format.

r""" Preprocesses the dataset, including format checking and tokenization.

r""" Computes video sample frames according to fps.

r""" input: a dict with keys {"question", "A", "B", "C", "D", "answer"} output: a tuple of (prompt, response)

r""" Pre-processes a single image.

r""" Regularizes images to avoid error. Including reading and pre-processing.

r""" Asynchronously gets a list of responses of the chat model.

r""" Asynchronously gets a list of scores of the reward model.

r""" Asynchronously gets the response token-by-token of the chat model.

r""" Gets a list of responses of the chat model.

r""" Removes args with NoneType or False or empty string value.

r""" Computes the sum log probabilities of the labels under given logits if loss_type is not IPO, ORPO or SimPO. Otherwise the averag

r""" Returns the number of trainable parameters and number of all parameters in the model.

r""" Returns multiple pairs of token ids representing prompts and responses respectively.

r""" Extracts tool message.

r""" Finds all available modules to apply lora or galore.

r""" Runs forward pass and computes the log probabilities.

r""" Computes the log probabilities of the given labels under the given logits. Returns: logps: A tensor of shape (batch_size,) conta

r""" Gets the path to user config.

r""" Freezes vision tower and language model for VLM full/freeze tuning.

r""" Gets a list of scores of the reward model.

r""" Gets the sequnce lengths in the current batch. e.g. ```python # input [ [1, 1, 2, 2, 2, 0], [1, 2, 2, 3, 3,

r""" Loads saved arguments.

r""" Loads value head parameters from Hugging Face Hub or local disk. Returns: dict with keys `v_head.summary.weight` and `v_head.summary.bia

r""" Expands the attention mask with indices from (batch_size, seq_len) to (batch_size, 1, seq_len, seq_len), while handles packed sequences a

Removes punctuation from the text and replaces it with a space.

Removes punctuation from the text and replaces it with a space.

r""" Saves arguments.

r""" EMA implementation according to TensorBoard.

Fixes extra whitespace in the text by collapsing multiple spaces into one.

Fixes extra whitespace in the text by collapsing multiple spaces into one.

r""" Configures root logger using a stdout stream handler with an explicit format.

r""" Returns the default logging level.

r""" Gets paligemma token type ids for computing loss. Returns: batch_token_type_ids: shape (batch_size, sequence_length)

r""" Prepares the tokenized dataset to perform AutoGPTQ. Do not use tensor output for JSON serialization.

r""" Loads a single dataset and aligns it to the standard format.

r""" Pads the tensor to the same length as the target tensor.

r""" Regularizes videos to avoid error. Including reading, resizing and converting.