Converts macs in numeric form to string form. Args: macs (int): Calculate the results of the model macs in numerical form. units

Converts flops in numeric form to string form. Args: flops (int): Calculate the results of the model flops in numerical form. uni

Converts params in numeric form to string form. Args: params_num (int): Calculate the results of the model param in numerical form.

Returns the total floating-point operations, MACs, and parameters of a model. Args: model ([torch.nn.Module]): The model of input must be

Automatically generates data in the form of transformes model input format. Args: input_shape (tuple):transformers model input shape:

Returns the total flops of the model. Args: as_string (bool, optional): whether to output the flops as string. Defaults to False.

Returns the total MACs of the model. Args: as_string (bool, optional): whether to output the flops as string. Defaults to False.

Returns the total number of parameters stored per rank. Args: as_string (bool, optional): whether to output the parameters as str

Recursively compute the FLOP s of the model Args: module (pytorch module): model format must be pytorch is_sparse (bool, Optional

Recursively compute the macs s of the model Args: module (pytorch module): model format must be pytorch is_sparse (bool, Optional

Ends the pipline of calculating FLOPs. The added attributes and handles are removed recursively on all the modules.

Starts the pipline of calculating FLOPs. Extra attributes are added recursively to all the modules and the calculate torch.nn.functionals are

Returns the total floating-point operations, MACs, and parameters of a model. Args: model_name (str): The model name in huggingface platf

Creates an empty model from its parent library on the `Hub` to calculate the overall memory consumption. Args: model_name (`str`):

Prints the model graph with the calculateing pipline attached to each module. Args: module_depth (int, optional): The depth of th

Prints the model graph with the calculateing pipline attached to each module. Args: module_depth (int, optional): The depth of th

Resets the pipline of calculating FLOPs. Adds or resets the extra attributes, include flops、macs、params.

Stop the pipline of calculating FLOPs. All torch.nn.functionals are restored to their originals.

Verifies that the model is on the hub and returns the model info.

Init Pipline of Calculating the FLOPs about model. Args: model (pytorch model): The model must be a pytorh model now.

Count flops for the addmm operation.

Count flops for the einsum operation.

Count flops for the matmul operation.

Count flops for the tensor addmm operation.

Converts bytes in numeric form to string form. Args: b (int): Calculate the results of the bytes in numerical form. units (str, o

Checks what library spawned `error` when a model is not found

Converts `size` from bytes to the largest possible unit

Creates a pretty table from a list of rows, minimal version of `tabulate`.