Performs a graph convolution operation. Args: input_tensor (torch.Tensor): Input tensor representing node features.

Performs forward pass of the Graph Convolutional Network (GCN). Args: input_tensor (torch.Tensor): Input node feature ma

Take a conditioning sequence of indices idx (LongTensor of shape (b,t)) and complete the sequence max_new_tokens times, feeding the p

create date and time for file save uniqueness example: 2022-05-07-08:31:12_PM'

we register our main layer class and use the fsdp transformer wrapping policy ensures embedding layers are in the root fsdp unit for shared access

load local checkpoint to rank0 cpu must be called * before * passing to FSDP

load an fdsp optimizer full_state checkpoint using scatter method this ensures only rank 0 loads the optimizer state dict and scatters to other ra

r""" Observers call this function to report rewards.

r""" Run one episode of n_steps. Args: agent_rref (RRef): an RRef referencing the agent object. n_steps (int)

r""" Run one episode of n_steps. Args: agent_rref (RRef): an RRef referencing the agent object. n_steps (int)

r""" A wrapper function that initializes RPC, calls the function, and shuts down RPC.

r""" This is the entry point for all processes. The rank 0 is the agent. All other ranks are observers.

r""" A wrapper function that initializes RPC, calls the function, and shuts down RPC.

r""" This is the entry point for all processes. The rank 0 is the agent. All other ranks are observers.

save model and optimizer via sharded_state_dict to save_dir

saving model via rank0 cpu streaming and full_state_dict

save optimizer state via full state dict

r""" This function is mostly borrowed from the Reinforcement Learning example. See https://github.com/pytorch/examples/tree/main/reinf

r""" Non-batching select_action, return the action right away.

r""" Batching select_action: In each step, the agent waits for states from all observers, and process them together. This helps to red