configuration_amdetr.py

"""
Copyright $today.year LY Corporation
LY Corporation licenses this file to you under the Apache License,
version 2.0 (the "License"); you may not use this file except in compliance
with the License. You may obtain a copy of the License at:
  https://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
License for the specific language governing permissions and limitations
under the License.
Moment-DETR (https://github.com/jayleicn/moment_detr)
Copyright (c) 2021 Jie Lei
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
"""
from transformers import PretrainedConfig


class AMDETRConfig(PretrainedConfig):
    model_type = "amdetr"

    def __init__(
        self,
        seed: int = 2023,
        device: str = "cuda",
        num_workers: int = 4,
        lr: float = 0.0001,
        lr_drop: int = 400,
        wd: float = 0.0001,
        n_epoch: int = 100,
        max_es_cnt: int = 200,
        bsz: int = 32,
        eval_bsz: int = 100,
        grad_clip: float = 0.1,
        max_q_l: int = 32,
        max_v_l: int = 75,
        max_windows: int = 5,
        clip_length: int = 1,
        eval_epoch_interval: int = 1,
        position_embedding: str = "sine",
        enc_layers: int = 2,
        dec_layers: int = 2,
        dim_feedforward: int = 1024,
        hidden_dim: int = 256,
        input_dropout: float = 0.5,
        dropout: float = 0.1,
        nheads: int = 8,
        num_queries: int = 10,
        n_input_proj: int = 2,
        saliency_margin: float = 0.2,
        span_loss_type: str = "l1",
        set_cost_span: int = 10,
        set_cost_giou: int = 1,
        set_cost_class: int = 4,
        span_loss_coef: int = 10,
        giou_loss_coef: int = 1,
        label_loss_coef: int = 4,
        eos_coef: float = 0.1,
        lw_saliency: int = 1,
        ckpt_filename: str = "best.ckpt",
        train_log_filename: str = "train.log",
        eval_log_filename: str = "val.log",
        eval_split_name: str = "val",
        aux_loss: bool = True,
        model_ema: bool = False,
        ema_decay: float = 0.9,
        results_dir: str = "results/qd_detr/clotho-moment/clap",
        ctx_mode: str = "audio_tef",
        v_feat_types: None = None,
        a_feat_types: str = "clap",
        t_feat_type: str = "clap",
        v_feat_dim: int = 2,
        a_feat_dim: int = 768,
        t_feat_dim: int = 768,
        model_name: str = "qd_detr",
        dset_name: str = "clotho-moment",
        train_path: str = "data/clotho_moment/clotho_moment_train_release.jsonl",
        eval_path: str = "data/clotho_moment/clotho_moment_val_release.jsonl",
        ckpt_filepath: str = "results/qd_detr/clotho-moment/clap/best.ckpt",
        train_log_filepath: str = "results/qd_detr/clotho-moment/clap/train.log",
        eval_log_filepath: str = "results/qd_detr/clotho-moment/clap/val.log",
        v_feat_dirs: None = None,
        t_feat_dir: str = "features/clotho-moment/clap_text",
        a_feat_dirs: list = ['features/clotho-moment/clap'],
        t_feat_dir_pretrain_eval: None = None,
        train_log_txt_formatter: str = "{time_str} [Epoch] {epoch:03d} [Loss] {loss_str}\n",
        eval_log_txt_formatter: str = "{time_str} [Epoch] {epoch:03d} [Loss] {loss_str} [Metrics] {eval_metrics_str}\n", 
        **kwargs
    ) -> None:
        args_and_values = locals()
        for arg_name, arg_value in args_and_values.items():
            if arg_name != 'self':
                setattr(self, arg_name, arg_value)
        super().__init__(**kwargs)


if __name__ == "__main__":
    cfg = AMDETRConfig()
    print(cfg)