accum_freq: 1
audio_ext: flac
aug_cfg: {}
batch_size: 64
beta1: 0.9
beta2: 0.95
cache_dir: None
caption_key_map: None
caption_weight: 0.0
checkpoint_path: vctrans_1n/checkpoints
classification_label_key: labels
classification_num_classes: 527
classification_weight: 0.0
coca_caption_loss_weight: 2.0
coca_contrastive_loss_weight: 1.0
compile_mode: default
contrastive_weight: 1.0
copy_codebase: False
csv_caption_key: title
csv_img_key: filepath
csv_separator: 	
dac_model_type: 44khz
dac_num_codebooks: 1
dac_tokenize: False
data_filling: pad
data_truncating: rand_trunc
dataset_resampled: True
dataset_type: webdataset-audio
ddp_static_graph: True
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
emotion_weight: 0.0
enable_fusion: False
epochs: 8
epochs_cooldown: None
eps: 1e-08
exclude_nc_emilia: False
force_context_length: 128
force_custom_text: False
force_image_size: None
force_patch_dropout: None
force_quick_gelu: False
fsdp2: False
gather_with_grad: True
grad_checkpointing: False
grad_clip_norm: 1.0
hard_neg_k: 3
hard_neg_max_frac: 0.5
hard_neg_min_frac: 0.1
hard_neg_weight: 1.0
holdout_eval_mod: 1000
horovod: False
image_interpolation: None
image_mean: None
image_resize_mode: None
image_std: None
imagenet_v2: None
imagenet_val: None
int16_normalize: False
local_loss: True
local_rank: 0
lock_audio: False
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: vctrans_1n/out.log
log_step_timings: False
logs: voiceclap/checkpoints
loss_dist_impl: None
lr: 0.0001
lr_cooldown_end: 0.0
lr_cooldown_power: 1.0
lr_scheduler: cosine
model: whisper-small-mMiniLM
momentum: None
moss_join_sep:  
moss_sample_k: 3
naflex: False
name: vctrans_1n
no_set_device_rank: False
opt: adamw
pack_max_clips: 4
pcl_embed_dim: 768
pcl_min_z: 1.5
pcl_proto_lr: 0.001
pcl_temperature: 0.1
pcl_warmup_batches: 50
pcl_weight: 0.0
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: tensorboard
resume: None
save_frequency: 1
save_most_recent: True
seed: 0
sequence_packing: False
siglip: True
skip_scheduler: False
synthetic_seq_len: 0
tensorboard: True
tensorboard_path: vctrans_1n/tensorboard
torchcompile: False
torchscript: False
trace: False
train_data: emolia-balanced-5M-subset-flac48k-wds-paired/emolia-{000000..001051}.tar
train_data_file: None
train_data_upsampling_factors: None
train_num_samples: 125000
use_bn_sync: False
use_bnb_linear: None
val_data: None
val_frequency: 1
val_num_samples: None
wandb: False
wandb_notes: 
wandb_project_name: open-clip
warmup: 500
wd: 0.2
workers: 12
world_size: 4
zeroshot_frequency: 2