config.json

{
  "architectures": [
    "E2TextToSemanticWLenSMPL"
  ],
  "bos_token_id": 1,
  "eos_token_id": 2,
  "hidden_size": 512,
  "main_encoder_args": {
    "attn_dropout": 0.0,
    "attn_flash": true,
    "conv_dropout": 0.0,
    "conv_kernel_size": 5,
    "depth": 8,
    "dim": 512,
    "dim_head": 32,
    "ff_dropout": 0.0,
    "ff_mult": 4,
    "heads": 16
  },
  "model_type": "e2_text_to_semantic_w_length_smpl",
  "pad_token_id": 0,
  "pre_text_encoder_args": {
    "attn_dropout": 0.0,
    "attn_flash": true,
    "conv_dropout": 0.0,
    "conv_kernel_size": 5,
    "depth": 4,
    "dim": 512,
    "dim_head": 32,
    "ff_dropout": 0.0,
    "ff_mult": 4,
    "heads": 16
  },
  "semantic_vocab_size": 1024,
  "special_tokens": {
    "mask": 6,
    "pad": 0,
    "prompt_end": 3,
    "prompt_start": 2,
    "sep": 5,
    "speech": 4,
    "text": 1
  },
  "text_vocab_size": 256,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.42.3"
}