versatile_diffusion/configs/model/optimus.yaml

optimus:
  symbol: optimus
  find_unused_parameters: false
  args: {}

optimus_bert_encoder:
  super_cfg: optimus
  type: optimus_bert_connector
  # pth: pretrained/optimus_bert_encoder.pth
  args:
    config:
      architectures: 
        - BertForMaskedLM
      attention_probs_dropout_prob: 0.1
      finetuning_task: null
      hidden_act: gelu
      hidden_dropout_prob: 0.1
      hidden_size: 768
      initializer_range: 0.02
      intermediate_size: 3072
      layer_norm_eps: 1.e-12
      max_position_embeddings: 512
      num_attention_heads: 12
      num_hidden_layers: 12
      num_labels: 2
      output_attentions: false
      output_hidden_states: false
      pruned_heads: {}
      torchscript: false
      type_vocab_size: 2
      vocab_size: 28996
    latent_size: 768

optimus_bert_tokenizer:
  super_cfg: optimus
  type: optimus_bert_tokenizer
  args:
    do_lower_case: false 
    max_len: 512
    vocab_file: versatile_diffusion/lib/model_zoo/optimus_models/vocab/bert-base-cased-vocab.txt

optimus_gpt2_decoder:
  super_cfg: optimus
  type: optimus_gpt2_connector
  # pth: pretrained/optimus_gpt2_decoder.pth
  args:
    config:
      architectures:
        - GPT2LMHeadModel
      attn_pdrop: 0.1
      embd_pdrop: 0.1
      finetuning_task: null
      hidden_size: 768
      initializer_range: 0.02
      latent_size: 768
      layer_norm_epsilon: 1.e-05
      max_position_embeddings: 1024
      n_ctx: 1024
      n_embd: 768
      n_head: 12
      n_layer: 12
      n_positions: 1024
      num_attention_heads: 12
      num_hidden_layers: 12
      num_labels: 1
      output_attentions: false
      output_hidden_states: false
      pretrained_config_archive_map:
        gpt2        : https://s3.amazonaws.com/models.huggingface.co/bert/gpt2-config.json
        gpt2-medium : https://s3.amazonaws.com/models.huggingface.co/bert/gpt2-medium-config.json
        gpt2-large  : https://s3.amazonaws.com/models.huggingface.co/bert/gpt2-large-config.json
      pruned_heads: {}
      resid_pdrop: 0.1
      summary_activation: null
      summary_first_dropout: 0.1
      summary_proj_to_labels: true
      summary_type: cls_index
      summary_use_proj: true
      torchscript: false
      vocab_size: 50260

optimus_gpt2_tokenizer:
  super_cfg: optimus
  type: optimus_gpt2_tokenizer
  args:
    do_lower_case: false 
    max_len: 1024
    vocab_file: versatile_diffusion/lib/model_zoo/optimus_models/vocab/gpt2-vocab.json
    merges_file: versatile_diffusion/lib/model_zoo/optimus_models/vocab/gpt2-merges.txt

optimus_vae:
  super_cfg: optimus
  type: optimus_vae
  pth: versatile_diffusion/pretrained/optimus-vae.pth
  args:
    encoder: MODEL(optimus_bert_encoder)
    decoder: MODEL(optimus_gpt2_decoder)
    tokenizer_encoder: MODEL(optimus_bert_tokenizer)
    tokenizer_decoder: MODEL(optimus_gpt2_tokenizer)
    args:
      latent_size: 768