avae.v0.1 / config.json
Hieuman's picture
Push model using huggingface_hub.
32cf53e verified
raw
history blame contribute delete
949 Bytes
{
"adapter_name": null,
"attn_implementation": "flash_attention_2",
"constraint_loss_weight": 0.0,
"content_discriminator_loss_weight": 1.0,
"content_encoder_model_name_or_path": "Alibaba-NLP/gte-Qwen2-1.5B-instruct",
"content_encoder_model_type": "qwen2",
"content_encoder_use_lora": true,
"content_loss_weight": 0.0,
"dropout_prob": 0.1,
"embedding_dim": 1536,
"generator_model_name_or_path": "Qwen/Qwen2.5-1.5B-Instruct",
"generator_use_lora": true,
"lora_alpha": 32,
"lora_dropout": 0.1,
"lora_r": 16,
"mi_reg_weight": 0.0,
"pooling_method": "mean",
"reconstruction_loss_weight": 1.0,
"style_discriminator_loss_weight": 1.0,
"style_encoder_model_name_or_path": "Hieuman/qwen2-1.5b-hard-author-reps",
"style_encoder_model_type": "qwen2",
"style_encoder_use_lora": true,
"style_loss_weight": 0.0,
"target_modules": "all",
"token_mi_reg_weight": 0.0,
"use_vae": true,
"vae_loss_weight": 1e-05
}