colossalai.nn.layer.parallel_1d

class colossalai.nn.layer.parallel_1d.Linear1D(in_features, out_features, bias=True, dtype=None, gather_output=False, skip_bias_add=False, weight_initializer=<function kaiming_uniform_.<locals>.initializer>, bias_initializer=<function xavier_uniform_.<locals>.initializer>)[source]

Linear layer for 1D parallelism.

Parameters

  • in_features (int) – size of each input sample.

  • out_features (int) – size of each output sample.

  • bias (bool, optional) – If set to False, the layer will not learn an additive bias, defaults to True.

  • dtype (torch.dtype, optional) – The dtype of parameters, defaults to None.

  • gather_output (bool, optional) – Whether to call all-gather on output, defaults to False.

  • skip_bias_add (bool, optional) – If set to True, it will skip bias add for linear layer, which is preserved for kernel fusion, defaults to False

  • weight_initializer (typing.Callable, optional) – The initializer of weight, defaults to kaiming uniform initializer.

  • bias_initializer (typing.Callable, optional) – The initializer of bias, defaults to xavier uniform initializer.

More details about initializer please refer to init.

class colossalai.nn.layer.parallel_1d.Linear1D_Col(in_features, out_features, bias=True, dtype=None, gather_output=False, skip_bias_add=False, weight_initializer=<function kaiming_uniform_.<locals>.initializer>, bias_initializer=<function xavier_uniform_.<locals>.initializer>)[source]

Linear layer with column parallelism.

The linear layer is defined as \(Y = XA + b\). A is parallelized along its second dimension as \(A = [A_1, ..., A_p]\).

Parameters

  • in_features (int) – size of each input sample.

  • out_features (int) – size of each output sample.

  • bias (bool, optional) – If set to False, the layer will not learn an additive bias, defaults to True.

  • dtype (torch.dtype, optional) – The dtype of parameters, defaults to None.

  • gather_output (bool, optional) – If true, call all-gather on output and make Y available to all GPUs, otherwise, every GPU will have its output which is \(Y_i = XA_i\), defaults to False

  • skip_bias_add (bool, optional) – If set to True, it will skip bias add for linear layer, which is preserved for kernel fusion, defaults to Fals

  • weight_initializer (typing.Callable, optional) – The initializer of weight, defaults to kaiming uniform initializer.

  • bias_initializer (typing.Callable, optional) – The initializer of bias, defaults to xavier uniform initializer.

More details about initializer please refer to init.

class colossalai.nn.layer.parallel_1d.Linear1D_Row(in_features, out_features, bias=True, dtype=None, parallel_input=True, skip_bias_add=False, weight_initializer=<function kaiming_uniform_.<locals>.initializer>, bias_initializer=<function xavier_uniform_.<locals>.initializer>)[source]

Linear layer with row parallelism

Parameters

  • in_features (int) – size of each input sample.

  • out_features (int) – size of each output sample.

  • bias (bool, optional) – If set to False, the layer will not learn an additive bias, defaults to True.

  • dtype (torch.dtype, optional) – The dtype of parameters, defaults to None.

  • parallel_input (bool, optional) – If set to True, it’s assumed that the input is split, defaults to False.

  • skip_bias_add (bool, optional) – If set to True, it will skip bias add for linear layer, which is preserved for kernel fusion, defaults to Fals

  • weight_initializer (typing.Callable, optional) – The initializer of weight, defaults to kaiming uniform initializer.

  • bias_initializer (typing.Callable, optional) – The initializer of bias, defaults to xavier uniform initializer.

More details about initializer please refer to init.

class colossalai.nn.layer.parallel_1d.Embedding1D(num_embeddings, embedding_dim, padding_idx=None, dtype=None, weight_initializer=<function normal_.<locals>.initializer>, *args, **kwargs)[source]

Embedding for 1D parallelism.

Parameters

  • num_embeddings (int) – number of embeddings.

  • embedding_dim (int) – dimension of embedding.

  • padding_idx (int, optional) – If specified, the entries at padding_idx do not contribute to the gradient; therefore, the embedding vector at padding_idx is not updated during training, i.e. it remains as a fixed “pad”, defaults to None.

  • dtype (torch.dtype, optional) – The dtype of parameters, defaults to None.

  • weight_initializer (typing.Callable, optional) – he initializer of weight, defaults to normal initializer.

The args and kwargs used in torch.nn.functional.embedding should contain:

max_norm (float, optional): If given, each embedding vector with norm larger than max_norm is
            renormalized to have norm max_norm. Note: this will modify weight in-place.
norm_type (float, optional): The p of the p-norm to compute for the max_norm option. Default 2.
scale_grad_by_freq (bool, optional): If given, this will scale gradients by the inverse
            of frequency of the words in the mini-batch. Default False.
sparse (bool, optional): If True, gradient w.r.t. weight will be a sparse tensor. Default False.

More details about args and kwargs could be found in Embedding.

More details about initializer please refer to init

class colossalai.nn.layer.parallel_1d.Dropout1D(p=0.5, inplace=False)[source]

Dropout layer of 1D parallelism.

Parameters

  • p (float, optional) – probability of an element to be zeroed, defaults 0.5.

  • inplace (bool, optional) – whether to do dropout in-place, default to be False.

class colossalai.nn.layer.parallel_1d.Classifier1D(in_features, num_classes, weight=None, bias=True, dtype=None, weight_initializer=<function kaiming_uniform_.<locals>.initializer>, bias_initializer=<function xavier_uniform_.<locals>.initializer>)[source]

RowLinear with given weight. Classifier of 1D parallelism.

Parameters

  • in_features (int) – size of each input sample.

  • num_classes (int) – number of classes.

  • weight (torch.nn.Parameter, optional) – weight of the classifier, defaults to None.

  • bias (bool, optional) – If set to False, the layer will not learn an additive bias, defaults to True.

  • dtype (torch.dtype, optional) – The dtype of parameters, defaults to None.

  • weight_initializer (typing.Callable, optional) – The initializer of weight, defaults to kaiming uniform initializer.

  • bias_initializer (typing.Callable, optional) – The initializer of bias, defaults to xavier uniform initializer.

More details about initializer please refer to init.

class colossalai.nn.layer.parallel_1d.VocabParallelClassifier1D(in_features, num_classes, weight=None, bias=True, dtype=None, weight_initializer=<function kaiming_uniform_.<locals>.initializer>, bias_initializer=<function xavier_uniform_.<locals>.initializer>)[source]

ColLinear with given weight. Classifier of 1D parallelism.

Parameters

  • in_features (int) – size of each input sample.

  • num_classes (int) – number of classes.

  • weight (torch.nn.Parameter, optional) – weight of the classifier, defaults to None.

  • bias (bool, optional) – If set to False, the layer will not learn an additive bias, defaults to True.

  • dtype (torch.dtype, optional) – The dtype of parameters, defaults to None.

  • weight_initializer (typing.Callable, optional) – The initializer of weight, defaults to kaiming uniform initializer.

  • bias_initializer (typing.Callable, optional) – The initializer of bias, defaults to xavier uniform initializer.

More details about initializer please refer to init.

class colossalai.nn.layer.parallel_1d.VocabParallelEmbedding1D(num_embeddings, embedding_dim, padding_idx=None, dtype=None, weight_initializer=<function normal_.<locals>.initializer>, *args, **kwargs)[source]

Embedding parallelized in the vocabulary dimension.

Parameters

  • num_embeddings (int) – number of embeddings.

  • embedding_dim (int) – dimension of embedding.

  • padding_idx (int, optional) – If specified, the entries at padding_idx do not contribute to the gradient; therefore, the embedding vector at padding_idx is not updated during training, i.e. it remains as a fixed “pad”, defaults to None.

  • dtype (torch.dtype, optional) – The dtype of parameters, defaults to None.

  • weight_initializer (typing.Callable, optional) – he initializer of weight, defaults to normal initializer.

The args and kwargs used in :class:torch.nn.functional.embedding should contain:

max_norm (float, optional): If given, each embedding vector with norm larger than max_norm is
            renormalized to have norm max_norm. Note: this will modify weight in-place.
norm_type (float, optional): The p of the p-norm to compute for the max_norm option. Default 2.
scale_grad_by_freq (bool, optional): If given, this will scale gradients by the inverse
            of frequency of the words in the mini-batch. Default False.
sparse (bool, optional): If True, gradient w.r.t. weight will be a sparse tensor. Default False.

More details about args and kwargs could be found in Embedding.

More details about initializer please refer to init.