End the current task with a failure

r"""Works like shutil.copy() for file, and shutil.copytree for dir, and supports hdfs. Copy data and mode bits ("cp src dst"). Return the file's

Reload.

Gets the state of the environment; i.e., screenshot & UI tree. In practice this will usually be called after executing an action. Logic shoul

Create a DataProto from a dict of tensors. This assumes that 1. All the tensor in tensors have the same dim0 2. Only dim0 is the batch

r"""Works like os.makedirs() but supports hdfs. Super-mkdir; create a leaf directory and all intermediate ones. Works like mkdir, except tha

Gets the state of the environment; i.e., screenshot & UI tree. In practice this will usually be called after executing an action. Logic shoul

In megatron, the model and optimizer storage are: - bf16 parameter data chunked in model parallel group - fp32 grad chunked in model para

Get ulysses sequence parallel world size.

Generates a random note.

Generates a random note.

Aggregate the loss matrix into a scalar. Args: loss_mat: `(torch.Tensor)`: shape: (bs, response_length) loss_mas

Create a huggingface processor to process multimodal data. Args: name_or_path (str): The name of the processor. Returns: tra

Calling multimodal LLM with a prompt and a list of images. Args: text_prompt: Text prompt. Returns: Text output, is_safe, and ra

Calling multimodal LLM with a prompt and a list of images. Args: text_prompt: Text prompt. Returns: Text output, is_safe, and ra

Split a batch into micro-batches by total token count, with optional DP sync and padding. Args: batch (TensorDict): must include "at