diff --git a/transformers/configuration_albert.py b/transformers/configuration_albert.py
new file mode 100644
index 0000000000..c86c9565cb
--- /dev/null
+++ b/transformers/configuration_albert.py
@@ -0,0 +1,72 @@
+from .configuration_utils import PretrainedConfig
+
+class AlbertConfig(PretrainedConfig):
+    """Configuration for `AlbertModel`.
+
+    The default settings match the configuration of model `albert_xxlarge`.
+    """
+
+    def __init__(self,
+                 vocab_size_or_config_json_file,
+                 embedding_size=128,
+                 hidden_size=4096,
+                 num_hidden_layers=12,
+                 num_hidden_groups=1,
+                 num_attention_heads=64,
+                 intermediate_size=16384,
+                 inner_group_num=1,
+                 down_scale_factor=1,
+                 hidden_act="gelu",
+                 hidden_dropout_prob=0,
+                 attention_probs_dropout_prob=0,
+                 max_position_embeddings=512,
+                 type_vocab_size=2,
+                 initializer_range=0.02,
+                 layer_norm_eps=1e-12, **kwargs):
+        """Constructs AlbertConfig.
+
+        Args:
+            vocab_size: Vocabulary size of `inputs_ids` in `AlbertModel`.
+            embedding_size: size of voc embeddings.
+            hidden_size: Size of the encoder layers and the pooler layer.
+            num_hidden_layers: Number of hidden layers in the Transformer encoder.
+            num_hidden_groups: Number of group for the hidden layers, parameters in
+                the same group are shared.
+            num_attention_heads: Number of attention heads for each attention layer in
+                the Transformer encoder.
+            intermediate_size: The size of the "intermediate" (i.e., feed-forward)
+                layer in the Transformer encoder.
+            inner_group_num: int, number of inner repetition of attention and ffn.
+            down_scale_factor: float, the scale to apply
+            hidden_act: The non-linear activation function (function or string) in the
+                encoder and pooler.
+            hidden_dropout_prob: The dropout probability for all fully connected
+                layers in the embeddings, encoder, and pooler.
+            attention_probs_dropout_prob: The dropout ratio for the attention
+                probabilities.
+            max_position_embeddings: The maximum sequence length that this model might
+                ever be used with. Typically set this to something large just in case
+                (e.g., 512 or 1024 or 2048).
+            type_vocab_size: The vocabulary size of the `token_type_ids` passed into
+                `AlbertModel`.
+            initializer_range: The stdev of the truncated_normal_initializer for
+                initializing all weight matrices.
+        """
+        super(AlbertConfig, self).__init__(**kwargs)
+
+        self.vocab_size = vocab_size_or_config_json_file
+        self.embedding_size = embedding_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_hidden_groups = num_hidden_groups
+        self.num_attention_heads = num_attention_heads
+        self.inner_group_num = inner_group_num
+        self.down_scale_factor = down_scale_factor
+        self.hidden_act = hidden_act
+        self.intermediate_size = intermediate_size
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.type_vocab_size = type_vocab_size
+        self.initializer_range = initializer_range
+        self.layer_norm_eps = layer_norm_eps
\ No newline at end of file
diff --git a/transformers/convert_albert_original_tf_checkpoint_to_pytorch.py b/transformers/convert_albert_original_tf_checkpoint_to_pytorch.py
new file mode 100644
index 0000000000..04877d41b9
--- /dev/null
+++ b/transformers/convert_albert_original_tf_checkpoint_to_pytorch.py
@@ -0,0 +1,44 @@
+
+
+
+from transformers import AlbertConfig, BertForPreTraining, load_tf_weights_in_bert
+
+
+
+def convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, bert_config_file, pytorch_dump_path):
+    # Initialise PyTorch model
+    config = BertConfig.from_json_file(bert_config_file)
+    print("Building PyTorch model from configuration: {}".format(str(config)))
+    model = BertForPreTraining(config)
+
+    # Load weights from tf checkpoint
+    load_tf_weights_in_bert(model, config, tf_checkpoint_path)
+
+    # Save pytorch-model
+    print("Save PyTorch model to {}".format(pytorch_dump_path))
+    torch.save(model.state_dict(), pytorch_dump_path)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    ## Required parameters
+    parser.add_argument("--tf_checkpoint_path",
+                        default = None,
+                        type = str,
+                        required = True,
+                        help = "Path to the TensorFlow checkpoint path.")
+    parser.add_argument("--albert_config_file",
+                        default = None,
+                        type = str,
+                        required = True,
+                        help = "The config json file corresponding to the pre-trained BERT model. \n"
+                            "This specifies the model architecture.")
+    parser.add_argument("--pytorch_dump_path",
+                        default = None,
+                        type = str,
+                        required = True,
+                        help = "Path to the output PyTorch model.")
+    args = parser.parse_args()
+    convert_tf_checkpoint_to_pytorch(args.tf_checkpoint_path,
+                                     args.bert_config_file,
+                                     args.pytorch_dump_path)
diff --git a/transformers/modeling_albert.py b/transformers/modeling_albert.py
new file mode 100644
index 0000000000..b006cbe8fd
--- /dev/null
+++ b/transformers/modeling_albert.py
@@ -0,0 +1,331 @@
+
+import os
+import math
+import logging
+import torch
+import torch.nn as nn
+from transformers.configuration_albert import AlbertConfig
+logger = logging.getLogger(__name__)
+
+def load_tf_weights_in_albert(model, config, tf_checkpoint_path):
+    """ Load tf checkpoints in a pytorch model."""
+    try:
+        import re
+        import numpy as np
+        import tensorflow as tf
+    except ImportError:
+        logger.error("Loading a TensorFlow model in PyTorch, requires TensorFlow to be installed. Please see "
+            "https://www.tensorflow.org/install/ for installation instructions.")
+        raise
+    tf_path = os.path.abspath(tf_checkpoint_path)
+    logger.info("Converting TensorFlow checkpoint from {}".format(tf_path))
+    # Load weights from TF model
+    init_vars = tf.train.list_variables(tf_path)
+    names = []
+    arrays = []
+    for name, shape in init_vars:
+        logger.info("Loading TF weight {} with shape {}".format(name, shape))
+        array = tf.train.load_variable(tf_path, name)
+        names.append(name)
+        arrays.append(array)
+
+    print(model)
+    
+    for name, array in zip(names, arrays):
+        og = name
+        name = name.replace("transformer/group_0/inner_group_0", "transformer")
+        name = name.replace("LayerNorm", "layer_norm")
+        name = name.replace("ffn_1", "ffn")
+        name = name.replace("ffn/intermediate/output", "ffn_output")
+        name = name.replace("attention_1", "attention")   
+        name = name.replace("cls/predictions/transform", "predictions")
+        name = name.replace("transformer/layer_norm_1", "transformer/attention/output/LayerNorm")        
+        name = name.split('/')
+
+        print(name)
+        pointer = model
+        for m_name in name:
+            if re.fullmatch(r'[A-Za-z]+_\d+', m_name):
+                l = re.split(r'_(\d+)', m_name)
+            else:
+                l = [m_name]
+
+            if l[0] == 'kernel' or l[0] == 'gamma':
+                pointer = getattr(pointer, 'weight')
+            elif l[0] == 'output_bias' or l[0] == 'beta':
+                pointer = getattr(pointer, 'bias')
+            elif l[0] == 'output_weights':
+                pointer = getattr(pointer, 'weight')
+            elif l[0] == 'squad':
+                pointer = getattr(pointer, 'classifier')
+            else:
+                try:
+                    pointer = getattr(pointer, l[0])
+                except AttributeError:
+                    logger.info("Skipping {}".format("/".join(name)))
+                    continue
+            if len(l) >= 2:
+                num = int(l[1])
+                pointer = pointer[num]
+
+        if m_name[-11:] == '_embeddings':
+            pointer = getattr(pointer, 'weight')
+        elif m_name == 'kernel':
+            array = np.transpose(array)
+            print("transposed")
+        try:
+            assert pointer.shape == array.shape
+        except AssertionError as e:
+            e.args += (pointer.shape, array.shape)
+            raise
+        print("Initialize PyTorch weight {} from {}".format(name, og))
+        pointer.data = torch.from_numpy(array)
+
+    return model
+
+
+class AlbertEmbeddings(nn.Module):
+    """
+    Construct the embeddings from word, position and token_type embeddings.
+    """
+    def __init__(self, config):
+        super(AlbertEmbeddings, self).__init__()
+
+        self.word_embeddings = nn.Embedding(config.vocab_size, config.embedding_size, padding_idx=0)
+        self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.embedding_size)
+        self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.embedding_size)
+        self.layer_norm = torch.nn.LayerNorm(config.embedding_size, eps=config.layer_norm_eps)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+    def forward(self, input_ids, token_type_ids=None, position_ids=None):
+        seq_length = input_ids.size(1)
+        if position_ids is None:
+            position_ids = torch.arange(seq_length, dtype=torch.long, device=input_ids.device)
+            position_ids = position_ids.unsqueeze(0).expand_as(input_ids)
+        if token_type_ids is None:
+            token_type_ids = torch.zeros_like(input_ids)
+
+        word_embeddings = self.word_embeddings(input_ids)
+        position_embeddings = self.position_embeddings(position_ids)
+        token_type_embeddings = self.token_type_embeddings(token_type_ids)
+
+        embeddings = word_embeddings + position_embeddings + token_type_embeddings
+        embeddings = self.layer_norm(embeddings)
+        embeddings = self.dropout(embeddings)
+        return embeddings
+
+
+    def get_word_embeddings_table(self):
+        return self.word_embeddings
+
+
+class AlbertModel(nn.Module):
+    def __init__(self, config):
+        super(AlbertModel, self).__init__()
+
+        self.config = config
+        self.embeddings = AlbertEmbeddings(config)
+        self.encoder = AlbertEncoder(config)
+        self.pooler = nn.Linear(config.hidden_size, config.hidden_size)
+        self.pooler_activation = nn.Tanh()
+
+    def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None):
+        if attention_mask is None:
+            attention_mask = torch.ones_like(input_ids)
+        if token_type_ids is None:
+            token_type_ids = torch.zeros_like(input_ids)
+
+        extended_attention_mask = attention_mask.unsqueeze(1).unsqueeze(2)
+        extended_attention_mask = extended_attention_mask.to(dtype=next(self.parameters()).dtype) # fp16 compatibility
+        extended_attention_mask = (1.0 - extended_attention_mask) * -10000.0
+        if head_mask is not None:
+            if head_mask.dim() == 1:
+                head_mask = head_mask.unsqueeze(0).unsqueeze(0).unsqueeze(-1).unsqueeze(-1)
+                head_mask = head_mask.expand(self.config.num_hidden_layers, -1, -1, -1, -1)
+            elif head_mask.dim() == 2:
+                head_mask = head_mask.unsqueeze(1).unsqueeze(-1).unsqueeze(-1)  # We can specify head_mask for each layer
+            head_mask = head_mask.to(dtype=next(self.parameters()).dtype) # switch to fload if need + fp16 compatibility
+        else:
+            head_mask = [None] * self.config.num_hidden_layers
+
+        embedding_output = self.embeddings(input_ids, position_ids=position_ids, token_type_ids=token_type_ids)
+        encoder_outputs = self.encoder(embedding_output,
+                                       extended_attention_mask,
+                                       head_mask=head_mask)
+        sequence_output = encoder_outputs[0]
+        print(sequence_output.shape, sequence_output[:, 0].shape, self.pooler(sequence_output[:, 0]).shape)
+        pooled_output = self.pooler_activation(self.pooler(sequence_output[:, 0]))
+
+        outputs = (sequence_output, pooled_output,) + encoder_outputs[1:]  # add hidden_states and attentions if they are here
+        return outputs
+
+
+class AlbertForMaskedLM(nn.Module):
+    def __init__(self, config):
+        super(AlbertForMaskedLM, self).__init__()
+
+        self.config = config
+        self.bert = AlbertModel(config)
+        self.layer_norm = nn.LayerNorm(config.embedding_size)
+        self.bias = nn.Parameter(torch.zeros(config.vocab_size))
+        self.dense = nn.Linear(config.hidden_size, config.embedding_size)
+        self.word_embeddings = nn.Linear(config.embedding_size, config.vocab_size)
+
+    def tie_weights(self):
+        """ Make sure we are sharing the input and output embeddings.
+            Export to TorchScript can't handle parameter sharing so we are cloning them instead.
+        """
+        self._tie_or_clone_weights(self.classifier.word_embeddings,
+                                   self.transformer.embeddings.word_embeddings)
+
+    def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None):
+        hidden_states = self.bert(input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None)[0]
+        hidden_states = self.dense(hidden_states)
+        hidden_states = gelu_new(hidden_states)
+        hidden_states = self.layer_norm(hidden_states)
+
+        logits = self.word_embeddings(hidden_states)
+
+        return logits
+
+
+class AlbertAttention(nn.Module):
+    def __init__(self, config):
+        super(AlbertAttention, self).__init__()
+
+        if config.hidden_size % config.num_attention_heads != 0:
+            raise ValueError(
+                "The hidden size (%d) is not a multiple of the number of attention "
+                "heads (%d)" % (config.hidden_size, config.num_attention_heads))
+        self.output_attentions = config.output_attentions
+
+        self.num_attention_heads = config.num_attention_heads
+        self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
+        self.all_head_size = self.num_attention_heads * self.attention_head_size
+
+        self.query = nn.Linear(config.hidden_size, self.all_head_size)
+        self.key = nn.Linear(config.hidden_size, self.all_head_size)
+        self.value = nn.Linear(config.hidden_size, self.all_head_size)
+
+        self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+
+    def transpose_for_scores(self, x):
+        new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size)
+        x = x.view(*new_x_shape)
+        return x.permute(0, 2, 1, 3)
+
+    def forward(self, input_ids, attention_mask=None, head_mask=None):
+        mixed_query_layer = self.query(input_ids)
+        mixed_key_layer = self.key(input_ids)
+        mixed_value_layer = self.value(input_ids)
+
+        query_layer = self.transpose_for_scores(mixed_query_layer)
+        key_layer = self.transpose_for_scores(mixed_key_layer)
+        value_layer = self.transpose_for_scores(mixed_value_layer)
+
+        # Take the dot product between "query" and "key" to get the raw attention scores.
+        attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2))
+        attention_scores = attention_scores / math.sqrt(self.attention_head_size)
+        if attention_mask is not None:
+            # Apply the attention mask is (precomputed for all layers in BertModel forward() function)
+            attention_scores = attention_scores + attention_mask
+
+        # Normalize the attention scores to probabilities.
+        attention_probs = nn.Softmax(dim=-1)(attention_scores)
+
+        # This is actually dropping out entire tokens to attend to, which might
+        # seem a bit unusual, but is taken from the original Transformer paper.
+        attention_probs = self.dropout(attention_probs)
+
+        # Mask heads if we want to
+        if head_mask is not None:
+            attention_probs = attention_probs * head_mask
+
+        context_layer = torch.matmul(attention_probs, value_layer)
+
+        context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
+        new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
+        reshaped_context_layer = context_layer.view(*new_context_layer_shape)
+        w = self.dense.weight.T.view(16, 64, 1024)
+        b = self.dense.bias
+
+        projected_context_layer = torch.einsum("bfnd,ndh->bfh", context_layer, w) + b
+        projected_context_layer = self.dropout(projected_context_layer)
+        layernormed_context_layer = self.LayerNorm(input_ids + projected_context_layer)
+        return layernormed_context_layer, projected_context_layer, reshaped_context_layer, context_layer, attention_scores, attention_probs, attention_mask
+
+
+class AlbertTransformer(nn.Module):
+    def __init__(self, config):
+        super(AlbertTransformer, self).__init__()
+        
+        self.config =config
+        self.layer_norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.attention = AlbertAttention(config)
+        self.ffn = nn.Linear(config.hidden_size, config.intermediate_size) 
+        self.ffn_output = nn.Linear(config.intermediate_size, config.hidden_size)
+
+    def forward(self, hidden_states, attention_mask=None, head_mask=None):
+        for i in range(self.config.num_hidden_layers):
+            attention_output = self.attention(hidden_states, attention_mask)[0]
+            ffn_output = self.ffn(attention_output)
+            ffn_output = gelu_new(ffn_output)
+            ffn_output = self.ffn_output(ffn_output)
+            hidden_states = self.layer_norm(ffn_output + attention_output)
+
+        return hidden_states
+
+
+def gelu_new(x):
+    """ Implementation of the gelu activation function currently in Google Bert repo (identical to OpenAI GPT).
+        Also see https://arxiv.org/abs/1606.08415
+    """
+    return 0.5 * x * (1 + torch.tanh(math.sqrt(2 / math.pi) * (x + 0.044715 * torch.pow(x, 3))))
+
+
+class AlbertEncoder(nn.Module):
+    def __init__(self, config):
+        super(AlbertEncoder, self).__init__()
+
+        self.output_attentions = config.output_attentions
+        self.output_hidden_states = config.output_hidden_states
+        self.embedding_hidden_mapping_in = nn.Linear(config.embedding_size, config.hidden_size)
+        self.transformer = AlbertTransformer(config)
+
+    def forward(self, hidden_states, attention_mask=None, head_mask=None):
+        hidden_states = self.embedding_hidden_mapping_in(hidden_states)
+        hidden_states = self.transformer(hidden_states, attention_mask, head_mask)
+
+        outputs = (hidden_states,)
+        if self.output_hidden_states:
+            outputs = outputs + (all_hidden_states,)
+        if self.output_attentions:
+            outputs = outputs + (all_attentions,)
+        return outputs  # last-layer hidden state, (all hidden states), (all attentions)
+
+# config = AlbertConfig.from_json_file("config.json")
+# # model = AlbertForMaskedLM(config)
+# model = AlbertModel(config)
+
+# model = load_tf_weights_in_albert(model, config, "albert/albert")
+
+# print(model)
+
+# input_ids = torch.tensor([[31, 51, 99], [15, 5, 0]])
+# input_mask = torch.tensor([[1, 1, 1], [1, 1, 0]])
+# segment_ids = torch.tensor([[0, 0, 1], [0, 0, 0]])
+
+# # sequence_output, pooled_outputs = model()
+
+# logits = model(input_ids, attention_mask=input_mask, token_type_ids=segment_ids)[1]
+
+
+# embeddings_output = 
+# print("pooled output", logits)
+# # print("Pooled output", pooled_outputs)
+
+config = AlbertConfig.from_json_file("/home/hf/google-research/albert/config.json")
+model = AlbertModel(config)
+model = load_tf_weights_in_albert(model, config, "/home/hf/transformers/albert/albert")
\ No newline at end of file