This function defines the grid and block sizes for the kernel launch and calls the kernel. chunk parallel size: siz_b_LQ chunk loop siz

Execute the parallel forward kernel for the H computation in the mLSTM chunkwise formulation. This function defines the grid and block sizes

Computes only the deltaC gradients for the backward pass. The other gradients are computed in the other (kernel) function. We do not need to c

Computes only the deltaC gradients for the backward pass. The other gradients are computed in the other (kernel) function. We do not need to c

Computes only the deltaC gradients for the backward pass. The other gradients are computed in the other (kernel) function. We do not need to c

Computes only the deltaC gradients for the backward pass. The other gradients are computed in the other (kernel) function. We do not nee

Computes only the deltaC gradients for the backward pass. The other gradients are computed in the other (kernel) function. We do not nee

This function computes the states of the mLSTM chunkwise formulation. It is the first part of the chunkwise forward pass and computes the inter ch

Computes the backward pass of the mLSTM chunkwise formulation. Args: matQ: Tensor containing the query vectors. Shape (B, NH, S, DHQ

Execute the forward pass of the mLSTM chunkwise formulation. Args: matQ: Tensor containing the queries. Shape (B, NH, S, DHQK).

Execute the forward pass of the mLSTM chunkwise formulation. Args: matQ: Tensor containing the queries. Shape (B, NH, S, DHQK).

This is a single step of the mLSTM operation in recurrent form. Args: matC_old: (B, NH, DHQK, DHV) vecN_old: (B, NH, DHQK)

This is a single step of the mLSTM operation in recurrent form. Args: matC_state (jax.Array): (B, NH, DHQK, DHV) vecN_state (jax.

Reduce tensor at a dimension by named reduction functions. Args: x (torch.Tensor): tensor to be reduced reduction (str): one

Plot benchmark results from a DataFrame as line plot. This function is used in the benchmarks for plotting the result raw data. It produces si

predicate: Return true if the module is to be replaced. func: Return new module to use.

predicate: Return true if the module is to be replaced. func: Return new module to use.

Reshape selected dimensions for all tensors in nested dictionary or list or tuple to a target dimension. Args: x (dict or list o

Rounds up x to the next multiple of multiple_of.

Main method to generate samples depending on the chosen sampler type. DDIM is the default as it works well in all settings.

Main method to generate samples depending on the chosen sampler type. DDIM is the default as it works well in all settings.

For each image, randomly sample @num_crops crops of size (@crop_height, @crop_width), from @images. Args: images (torch.Tensor):

Heuristic for setting the chunk size params for the XL chunk kernel. These params are used to determine the grid and block sizes for the kernel la

Sort a list by an other.

Store the weights of the entire agent

Sends all torch tensors in nested dictionary or list or tuple to device @device, and returns a new nested structure. Args: x (di

Converts all torch tensors and numpy arrays in nested dictionary or list or tuple to float type entries, and returns a new nested structure.

Convert tensor to one-hot representation, assuming a certain number of total class labels. Args: tensor (torch.Tensor): tensor conta

Converts all numpy arrays in nested dictionary or list or tuple to torch tensors (and leaves existing torch Tensors as-is), and returns a

Returns the tokenized representation of given input string(s) Parameters ---------- texts : Union[str, List[str]] An input s

Run a single training step.

Adds dimension of size 1 at dimension @dim in all torch tensors and numpy arrays in nested dictionary or list or tuple and returns a new nest

Unsqueeze and expand a tensor at a dimension @dim by @size. Args: x (dict or list or tuple): a possibly nested dictionary or list or

Update currently maintained parameters. Call this every time the parameters are updated, such as the result of the `optimizer