2025-04-25 01:06:09,289 INFO [train.py:653] {
  "allowed_excess_duration_ratio": 0.1,
  "audio_key": "question_audio",
  "batch_idx_train": 0,
  "best_train_epoch": -1,
  "best_train_loss": Infinity,
  "best_valid_epoch": -1,
  "best_valid_loss": Infinity,
  "bucketing_sampler": true,
  "deepscale": false,
  "deepscale_config": null,
  "deepspeed": true,
  "deepspeed_config": "./slam_omni/ds_config_zero1.json",
  "drop_last": true,
  "enable_musan": false,
  "enable_spec_aug": true,
  "enable_speech_output": true,
  "encoder_projector_ds_rate": 8,
  "env_info": {
    "IP address": "0.114.183.253",
    "hostname": "7518205",
    "icefall-git-branch": null,
    "icefall-git-date": null,
    "icefall-git-sha1": null,
    "icefall-path": "/workspace/slam/icefall_omni",
    "k2-build-type": "Release",
    "k2-git-date": "Tue Oct 29 09:02:19 2024",
    "k2-git-sha1": "75e2ed6b2fd87c22b7f3f34bad48a69984bb8755",
    "k2-path": "/opt/conda/lib/python3.11/site-packages/k2/__init__.py",
    "k2-version": "1.24.4",
    "k2-with-cuda": true,
    "lhotse-path": "/workspace/slam/lhotse/lhotse/__init__.py",
    "lhotse-version": "1.30.0.dev+git.13c7616f.dirty",
    "python-version": "3.11",
    "torch-cuda-available": true,
    "torch-cuda-version": "12.4",
    "torch-version": "2.4.0"
  },
  "exp_dir": "slam_omni/exp_speech2speech_rerun",
  "frame_shift_ms": 10,
  "huggingface_dataset_path_or_name": "/workspace/Belle_1.4M-SLAM-Omni",
  "input_strategy": "PrecomputedFeatures",
  "llm_path_or_name": "models/Qwen2.5-0.5B-Instruct",
  "log_interval": 50,
  "manifest_dir": "data/fbank",
  "max_duration": 50,
  "num_buckets": 30,
  "num_epochs": 10,
  "num_workers": 2,
  "on_the_fly_feats": false,
  "pretrained_model_path": "./slam_omni/exp_speech2speech_rerun/epoch-1-checkpoint-15000.pt/pytorch_model.bin",
  "resample_to_16kHz": true,
  "reset_interval": 200,
  "return_cuts": true,
  "sampler_state_dict_path": "./slam_omni/exp_speech2speech_rerun/epoch-1-checkpoint-15000-sampler.pt",
  "seed": 42,
  "shuffle": true,
  "spec_aug_time_warp_factor": 80,
  "speech_encoder_path_or_name": "models/whisper/v1.1/whisper-large-v2-multi-hans-zh-epoch-3-avg-10.pt",
  "start_epoch": 1,
  "subsampling_factor": 2,
  "tensorboard": true,
  "text_key": "answer",
  "unfreeze_llm": true,
  "unfreeze_speech_projector": true,
  "use_flash_attn": true,
  "use_fp16": true,
  "use_lora": true,
  "valid_interval": 5000
}
2025-04-25 01:06:09,289 INFO [train.py:655] About to create model