| ### model | |
| model_name_or_path: ckpts/Qwen2.5-Omni-3B | |
| image_max_pixels: 307200 | |
| video_max_pixels: 65536 | |
| trust_remote_code: true | |
| ### method | |
| stage: sft | |
| do_train: true | |
| finetuning_type: full | |
| deepspeed: examples/deepspeed/ds_z2_config.json # choices: [ds_z0_config.json, ds_z2_config.json, ds_z3_config.json] | |
| ### dataset | |
| # dataset: identity, mllm_audio_demo, mllm_demo, mllm_video_audio_demo | |
| # dataset: scannet_det_train_4frames, scan2cap_train_32frames, scanrefer_train_32frames | |
| dataset: hapticcap | |
| # streaming: true | |
| template: qwen2_omni | |
| cutoff_len: 8192 | |
| # max_samples: 1000 | |
| # max_steps: 2000 | |
| overwrite_cache: false | |
| # preprocessing_batch_size: 64 | |
| preprocessing_num_workers: 8 | |
| dataloader_num_workers: 16 | |
| # buffer_size: 8 | |
| ### output | |
| output_dir: saves/qwen2_omni-3b_haptic/full/sft_v2 | |
| logging_steps: 1 | |
| save_steps: 5000 | |
| plot_loss: true | |
| overwrite_output_dir: false | |
| save_only_model: false | |
| ### train | |
| per_device_train_batch_size: 1 | |
| gradient_accumulation_steps: 4 | |
| freeze_vision_tower: true | |
| learning_rate: 1.0e-5 | |
| num_train_epochs: 1.0 | |
| lr_scheduler_type: cosine | |
| warmup_ratio: 0.1 | |
| bf16: true | |
| ddp_timeout: 180000000 | |
| resume_from_checkpoint: null | |
| report_to: tensorboard |