Update model

meta.yaml

@@ -2,7 +2,7 @@ espnet: '202503'
 files:
   classification_model_file: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/audioset2m/cls_earlarge3/valid.epoch_mAP.ave_1best.pth
 python: "3.9.18 | packaged by conda-forge | (main, Dec 23 2023, 17:20:25) \n[GCC 12.3.0]"
-timestamp: 1763330218.090096
+timestamp: 1763330253.214344
 torch: 2.1.2
 yaml_files:
   classification_train_config: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/audioset2m/cls_earlarge3/config.yaml

work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/audioset2m/cls_earlarge3/lightning_logs/version_0/events.out.tfevents.1746445643.gh132.hsn.cm.delta.internal.ncsa.edu.3804512.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6fe910543e2907bf59e7d16e8af07e973e3b0a28dd7416442b21156ad86128c
+size 282738

work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/audioset2m/cls_earlarge3/lightning_logs/version_0/hparams.yaml

@@ -0,0 +1,208 @@
+args: !!python/object:argparse.Namespace
+  accum_grad: 2
+  adapter: lora
+  adapter_conf: {}
+  allow_multi_rates: false
+  allow_variable_data_keys: false
+  batch_bins: 20000000
+  batch_size: 20
+  batch_type: length_weighted
+  best_model_criterion:
+  - - valid
+    - epoch_mAP
+    - max
+  category_sample_size: 10
+  chunk_default_fs: null
+  chunk_discard_short_samples: true
+  chunk_excluded_key_prefixes: []
+  chunk_length: 500
+  chunk_max_abs_length: null
+  chunk_shift_ratio: 0.5
+  collect_stats: false
+  config: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/earlarge3.final/conf/ear_large/audioset2m.yaml
+  create_graph_in_tensorboard: false
+  cudnn_benchmark: false
+  cudnn_deterministic: true
+  cudnn_enabled: true
+  ddp_comm_hook: null
+  decoder: linear
+  decoder_conf: {}
+  deepspeed_config: null
+  detect_anomaly: false
+  dist_backend: nccl
+  dist_init_method: env://
+  dist_launcher: null
+  dist_master_addr: null
+  dist_master_port: null
+  dist_rank: null
+  dist_world_size: null
+  drop_last_iter: false
+  dry_run: false
+  early_stopping_criterion: !!python/tuple
+  - valid
+  - loss
+  - min
+  embedding_fusion: null
+  embedding_fusion_conf: {}
+  encoder: beats
+  encoder_conf:
+    beats_ckpt_path: /work/nvme/bbjs/sbharadwaj/7Msounds/exp/beats_iter2_large2.tune_lr1.0e-4_warmup40000_bins1600000_totalsteps400000/epoch_latest.pt
+    beats_config:
+      dropout: 0.0
+      encoder_layerdrop: 0.1
+      layer_wise_gradient_decay_ratio: 0.6
+    roll_augment: true
+    roll_interval: 1
+    specaug_config:
+      apply_freq_mask: false
+      apply_time_mask: true
+      apply_time_warp: true
+      num_time_mask: 1
+      time_mask_width_ratio_range:
+      - 0
+      - 0.06
+    use_weighted_representation: false
+  exclude_weight_decay: false
+  exclude_weight_decay_conf: {}
+  fold_length:
+  - 160000
+  - 600
+  freeze_param: []
+  frontend: null
+  frontend_conf:
+    fs: 16k
+  grad_clip: 1
+  grad_clip_type: 2.0
+  grad_noise: false
+  gradient_as_bucket_view: true
+  ignore_init_mismatch: false
+  init: xavier_normal
+  init_param: []
+  input_size: 1
+  iterator_type: sequence
+  keep_nbest_models: 1
+  lightning_conf:
+    best_model_criterion:
+    - - valid/epoch_mAP
+      - max
+      - 1
+    default_root_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/audioset2m/cls_earlarge3.final
+    devices: 4
+    log_every_n_steps: 250
+    max_epochs: 25
+    num_nodes: 1
+    strategy: ddp
+    strategy_conf:
+      find_unused_parameters: true
+  local_rank: null
+  log_interval: null
+  log_level: INFO
+  max_cache_fd: 32
+  max_cache_size: 0.0
+  max_epoch: 25
+  model: espnet
+  model_conf:
+    classification_type: multi-label
+    log_epoch_metrics: true
+    lsm_weight: 0.1
+    mixup_probability: 0.8
+  multi_task_dataset: false
+  multiple_iterator: false
+  multiprocessing_distributed: false
+  nbest_averaging_interval: 0
+  ngpu: 0
+  no_forward_run: false
+  normalize: null
+  normalize_conf: {}
+  num_att_plot: 0
+  num_cache_chunks: 1024
+  num_iters_per_epoch: null
+  num_workers: 2
+  optim: adamw
+  optim_conf:
+    betas:
+    - 0.9
+    - 0.98
+    lr: 0.0001
+    weight_decay: 0.01
+  output_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/audioset2m/cls_earlarge3.final
+  patience: null
+  preencoder: null
+  preencoder_conf: {}
+  pretrain_path: null
+  print_config: false
+  required:
+  - output_dir
+  - token_list
+  resume: true
+  save_strategy: all
+  scheduler: cosineannealingwarmuprestarts
+  scheduler_conf:
+    first_cycle_steps: 400000
+    max_lr: 0.0001
+    min_lr: 5.0e-06
+    warmup_steps: 10000
+  seed: 0
+  sharded_ddp: false
+  shuffle_within_batch: false
+  sort_batch: descending
+  sort_in_batch: descending
+  specaug: null
+  specaug_conf: {}
+  task: cls
+  text_bpemodel: null
+  text_encoder: null
+  text_encoder_conf: {}
+  text_token_list: null
+  token_list: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/data/audioset2m/token_list
+  train_data_path_and_name_and_type:
+  - !!python/tuple
+    - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/audioset2m/train/wav.scp
+    - speech
+    - sound
+  - !!python/tuple
+    - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/audioset2m/train/text
+    - label
+    - text
+  train_dtype: float32
+  train_shape_file:
+  - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/audioset2m/cls_stats_16k/train/speech_shape
+  - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/audioset2m/cls_stats_16k/train/label_shape
+  unused_parameters: true
+  use_adapter: false
+  use_amp: false
+  use_deepspeed: false
+  use_matplotlib: true
+  use_preprocessor: true
+  use_tensorboard: true
+  use_tf32: false
+  use_wandb: true
+  user_callbacks:
+  - mAP_logging
+  utt2weight_file: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/data/audioset2m/utt2weight
+  val_scheduler_criterion: !!python/tuple
+  - valid
+  - loss
+  valid_batch_bins: null
+  valid_batch_size: null
+  valid_batch_type: length
+  valid_data_path_and_name_and_type:
+  - !!python/tuple
+    - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/audioset2m/eval/wav.scp
+    - speech
+    - sound
+  - !!python/tuple
+    - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/audioset2m/eval/text
+    - label
+    - text
+  valid_iterator_type: null
+  valid_max_cache_size: null
+  valid_shape_file:
+  - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/audioset2m/cls_stats_16k/valid/speech_shape
+  - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/audioset2m/cls_stats_16k/valid/label_shape
+  wandb_entity: shikhar
+  wandb_id: null
+  wandb_model_log_interval: -1
+  wandb_name: audioset2m.earlarge3.final
+  wandb_project: audioverse
+  write_collected_feats: false