| accum_freq: 1 |
| audio_ext: flac |
| audio_fill: repeatpad |
| audio_fusion: False |
| audio_int16_normalize: False |
| audio_trunc: rand_trunc |
| audio_zeroshot_audio_key: audio |
| audio_zeroshot_class_key: category |
| audio_zeroshot_dataset: None |
| audio_zeroshot_multiprocessing_context: forkserver |
| audio_zeroshot_split: train |
| audio_zeroshot_target_key: target |
| audio_zeroshot_templates: None |
| audio_zeroshot_workers: 0 |
| aug_cfg: {} |
| batch_size: 4096 |
| beta1: 0.9 |
| beta2: 0.98 |
| cache_dir: None |
| checkpoint_path: ./logs/ViT-B-32-Vanilla-resume/checkpoints |
| coca_caption_loss_weight: 2.0 |
| coca_contrastive_loss_weight: 1.0 |
| copy_codebase: False |
| csv_caption_key: title |
| csv_img_key: filepath |
| csv_separator: |
| dataset_resampled: False |
| dataset_type: webdataset |
| ddp_static_graph: False |
| debug: False |
| delete_previous_checkpoint: False |
| device: cuda:0 |
| dist_backend: None |
| dist_url: None |
| distill: False |
| distill_model: None |
| distill_pretrained: None |
| distributed: True |
| epochs: 32 |
| epochs_cooldown: None |
| eps: 1e-06 |
| force_context_length: None |
| force_custom_text: False |
| force_image_size: None |
| force_naflex_vision: False |
| force_patch_dropout: None |
| force_quick_gelu: False |
| fsdp: False |
| fsdp_checkpoint: full |
| fsdp_no_reshard_after_forward: False |
| fsdp_offload_cpu: False |
| gather_with_grad: True |
| grad_checkpointing: False |
| grad_clip_norm: None |
| image_interpolation: None |
| image_mean: None |
| image_resize_mode: None |
| image_std: None |
| imagenet_v2: None |
| imagenet_val: None |
| local_loss: True |
| local_rank: 0 |
| lock_image: False |
| lock_image_freeze_bn_stats: False |
| lock_image_unlocked_groups: 0 |
| lock_text: False |
| lock_text_freeze_layer_norm: False |
| lock_text_unlocked_layers: 0 |
| log_every_n_steps: 100 |
| log_level: 20 |
| log_local: False |
| log_path: ./logs/ViT-B-32-Vanilla-resume/out.log |
| logs: ./logs/ |
| loss_dist_impl: None |
| lr: 0.0005 |
| lr_cooldown_end: 0.0 |
| lr_cooldown_power: 1.0 |
| lr_scheduler: cosine |
| model: ViT-B-32 |
| momentum: None |
| naflex_batch_divisor: 8 |
| naflex_loss_scale: none |
| naflex_max_image_tokens_per_batch: 16384 |
| naflex_num_train_image_tokens: None |
| naflex_patch_size_probs: None |
| naflex_patch_sizes: None |
| naflex_seq_lens: None |
| name: ViT-B-32-Vanilla-resume |
| no_set_device_rank: False |
| opt: adamw |
| precision: amp_bfloat16 |
| pretrained: |
| pretrained_audio: None |
| pretrained_image: False |
| rank: 0 |
| remote_sync: None |
| remote_sync_frequency: 300 |
| remote_sync_protocol: s3 |
| report_to: wandb |
| resume: /scratch/work/zhul2/code/open_clip/logs/ViT-B-32-Vanilla/checkpoints/epoch_24.pt |
| save_frequency: 1 |
| save_most_recent: False |
| seed: 0 |
| siglip: False |
| skip_scheduler: False |
| tensorboard: False |
| tensorboard_path: |
| torchcompile: True |
| torchcompile_backend: None |
| torchcompile_mode: None |
| torchcompile_strategy: task |
| train_data: /scratch/shareddata/dldata/laion400M/img2dataset/laion400m-data/{00000..41407}.tar |
| train_data_upsampling_factors: None |
| train_num_samples: 268836185 |
| use_bn_sync: False |
| use_bnb_linear: None |
| use_naflex: False |
| val_data: None |
| val_frequency: 1 |
| val_num_samples: None |
| wandb: True |
| wandb_notes: |
| wandb_project_name: open-clip |
| warmup: 2000 |
| wd: 0.2 |
| workers: 12 |
| world_size: 8 |
| zeroshot_frequency: 2 |
|
|
