codesage/codesage-large-v2

1#!/usr/bin/env python2# coding=utf-83# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.4 5from transformers.configuration_utils import PretrainedConfig6 7CODESAGE_PRETRAINED_CONFIG_ARCHIVE_MAP = {8    "codesage/codesage-small-v2": "https://huggingface.co/codesage/codesage-small-v2/resolve/main/config.json",9    "codesage/codesage-base-v2": "https://huggingface.co/codesage/codesage-base-v2/resolve/main/config.json",10    "codesage/codesage-large-v2": "https://huggingface.co/codesage/codesage-large-v2/resolve/main/config.json",11}12 13 14class CodeSageConfig(PretrainedConfig):15    model_type = "codesage"16 17    def __init__(18            self,19            vocab_size=50257,20            max_position_embeddings=1024,21            hidden_size=768,22            num_hidden_layers=12,23            num_attention_heads=12,24            intermediate_size=3072,25            activation_function="gelu_new",26            residual_dropout_prob=0.1,27            embedding_dropout_prob=0.1,28            attention_dropout_prob=0.1,29            layer_norm_epsilon=1e-5,30            initializer_range=0.02,31            position_embedding_type='absolute',32            bos_token_id=0,33            eos_token_id=0,34            pad_token_id=49153,35            **kwargs36    ):37        self.vocab_size = vocab_size38        self.max_position_embeddings = max_position_embeddings39        self.hidden_size = hidden_size40        self.num_hidden_layers = num_hidden_layers41        self.num_attention_heads = num_attention_heads42        self.intermediate_size = intermediate_size43        assert 'gelu' in activation_function44        self.activation_function = activation_function45        self.residual_dropout_prob = residual_dropout_prob46        self.embedding_dropout_prob = embedding_dropout_prob47        self.attention_dropout_prob = attention_dropout_prob48        self.layer_norm_epsilon = layer_norm_epsilon49        self.initializer_range = initializer_range50        self.position_embedding_type = position_embedding_type51 52        super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)53