| {"llm_checkpoint_path": "/gsdata/home/cyzhao/model_zoo/TinyLlama-1.1B-Chat-v1.0", "vae_checkpoint_path": "/gsdata/home/cyzhao/model_zoo/sd-vae-ft-mse", "use_iterable_dataset": false, "datasets": [{"type": "imagenet", "data_root": "/gsdata/home/wangwh/ILSVRC2012_RAW_PYTORCH/train/", "weight": 9}], "img_data_root": "/root/paddlejob/workspace/vae_features/256resolution/noflip", "load_vae_features": false, "patch_size": 2, "in_channels": 4, "learn_sigma": true, "use_flash_attn": false, "use_pos_embed": true, "decoder_t_embed": "add_before_image_tokens", "use_adaln_final_layer": true, "use_bi_attn_img_tokens": true, "use_chat_template": false, "chat_template": "{%- for message in messages %}{%- if message['role'] == 'user' %}{{- bos_token + '### Human: ' + message['content'].strip() }}{%- elif message['role'] == 'assistant' %}{{- '### Assistant: ' + message['content'] + eos_token }}{%- endif %}{%- if loop.last and add_generation_prompt %}{{- '### Assistant: '}}{%- endif %}{%- endfor %}", "tokenizer_max_length": 512, "source_max_length": 128, "llm_lr": null, "use_lr_decay": false, "warmup_iters": 500, "lr_decay_iters": 300000, "lr_decay_rate": 0.1, "loss_dict": {"loss_lm": 1.0, "loss_dit": 1.0}} |