Upload folder using huggingface_hub
Browse files- recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717/.hydra/config.yaml +233 -0
- recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717/.hydra/hydra.yaml +191 -0
- recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717/.hydra/overrides.yaml +17 -0
- recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717/.hydra/resolved_config.yaml +292 -0
- recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717/train.log +853 -0
recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717/.hydra/config.yaml
ADDED
|
@@ -0,0 +1,233 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
system:
|
| 2 |
+
seed: 0
|
| 3 |
+
distributed:
|
| 4 |
+
enable: null
|
| 5 |
+
nodes: null
|
| 6 |
+
nproc_per_node: null
|
| 7 |
+
backend: null
|
| 8 |
+
init_method: null
|
| 9 |
+
rdzv_id: null
|
| 10 |
+
rdzv_backend: null
|
| 11 |
+
rdzv_endpoint: null
|
| 12 |
+
max_restarts: null
|
| 13 |
+
cudnn:
|
| 14 |
+
benchmark: true
|
| 15 |
+
deterministic: false
|
| 16 |
+
amp:
|
| 17 |
+
enable: false
|
| 18 |
+
dtype: null
|
| 19 |
+
accelerator: cuda
|
| 20 |
+
compile:
|
| 21 |
+
enable: null
|
| 22 |
+
kwargs: null
|
| 23 |
+
preprocess:
|
| 24 |
+
dump_format: birdclef2025
|
| 25 |
+
list_path: null
|
| 26 |
+
wav_dir: null
|
| 27 |
+
feature_dir: null
|
| 28 |
+
max_workers: null
|
| 29 |
+
max_shard_size: 1000000000
|
| 30 |
+
vad:
|
| 31 |
+
raw_root: null
|
| 32 |
+
trimmed_root: null
|
| 33 |
+
threshold: null
|
| 34 |
+
min_duration: 15
|
| 35 |
+
csv_path: ???
|
| 36 |
+
submission_path: ???
|
| 37 |
+
audio_root: ???
|
| 38 |
+
subset: ???
|
| 39 |
+
train_ratio: 0.8
|
| 40 |
+
data:
|
| 41 |
+
audio:
|
| 42 |
+
sample_rate: 32000
|
| 43 |
+
duration: 15
|
| 44 |
+
melspectrogram:
|
| 45 |
+
_target_: birdclef2025.transforms.birdclef.BirdCLEF2024BaselineMelSpectrogram
|
| 46 |
+
sample_rate: ${..audio.sample_rate}
|
| 47 |
+
hop_length: 512
|
| 48 |
+
f_min: 20
|
| 49 |
+
f_max: 16000
|
| 50 |
+
pad: 0
|
| 51 |
+
n_mels: 128
|
| 52 |
+
window_fn:
|
| 53 |
+
_target_: torch.hann_window
|
| 54 |
+
_partial_: true
|
| 55 |
+
power: 1.0
|
| 56 |
+
normalized: false
|
| 57 |
+
wkwargs: null
|
| 58 |
+
center: true
|
| 59 |
+
pad_mode: constant
|
| 60 |
+
onesided: null
|
| 61 |
+
norm: slaney
|
| 62 |
+
mel_scale: slaney
|
| 63 |
+
take_log: true
|
| 64 |
+
freq_mask_param:
|
| 65 |
+
- 0.06
|
| 66 |
+
- 0.1
|
| 67 |
+
time_mask_param:
|
| 68 |
+
- 0.06
|
| 69 |
+
- 0.12
|
| 70 |
+
eps: null
|
| 71 |
+
train:
|
| 72 |
+
dataset:
|
| 73 |
+
train:
|
| 74 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025PrimaryLabelDataset
|
| 75 |
+
list_path: dump/birdclef2025_reshape_15s/list/train.txt
|
| 76 |
+
feature_dir: /kaggle/input/birdclef-2025
|
| 77 |
+
audio_key: audio
|
| 78 |
+
sample_rate_key: sample_rate
|
| 79 |
+
label_name_key: primary_label
|
| 80 |
+
filename_key: filename
|
| 81 |
+
validation:
|
| 82 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025PrimaryLabelDataset
|
| 83 |
+
list_path: dump/birdclef2025_reshape_15s/list/validation.txt
|
| 84 |
+
feature_dir: /kaggle/input/birdclef-2025
|
| 85 |
+
audio_key: ${..train.audio_key}
|
| 86 |
+
sample_rate_key: ${..train.sample_rate_key}
|
| 87 |
+
label_name_key: ${..train.label_name_key}
|
| 88 |
+
filename_key: ${..train.filename_key}
|
| 89 |
+
dataloader:
|
| 90 |
+
train:
|
| 91 |
+
_target_: torch.utils.data.DataLoader
|
| 92 |
+
batch_size: 64
|
| 93 |
+
shuffle: true
|
| 94 |
+
collate_fn:
|
| 95 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025BaselineCollator
|
| 96 |
+
composer:
|
| 97 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025ReshapePrimaryLabelComposer
|
| 98 |
+
melspectrogram_transform: ${data.melspectrogram}
|
| 99 |
+
audio_key: audio
|
| 100 |
+
sample_rate_key: sample_rate
|
| 101 |
+
label_name_key: primary_label
|
| 102 |
+
filename_key: filename
|
| 103 |
+
waveform_key: waveform
|
| 104 |
+
melspectrogram_key: log_melspectrogram
|
| 105 |
+
label_index_key: label_index
|
| 106 |
+
sample_rate: ${data.audio.sample_rate}
|
| 107 |
+
duration: ${data.audio.duration}
|
| 108 |
+
decode_audio_as_waveform: true
|
| 109 |
+
decode_audio_as_monoral: true
|
| 110 |
+
training: true
|
| 111 |
+
target_shape: 256
|
| 112 |
+
melspectrogram_key: ${.composer.melspectrogram_key}
|
| 113 |
+
label_index_key: ${.composer.label_index_key}
|
| 114 |
+
alpha: 0.4
|
| 115 |
+
num_workers: ${const:birdclef2025.utils.data.default_num_workers}
|
| 116 |
+
validation:
|
| 117 |
+
_target_: torch.utils.data.DataLoader
|
| 118 |
+
batch_size: 64
|
| 119 |
+
shuffle: false
|
| 120 |
+
collate_fn:
|
| 121 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025BaselineValidationCollator
|
| 122 |
+
composer:
|
| 123 |
+
_target_: ${....train.collate_fn.composer._target_}
|
| 124 |
+
melspectrogram_transform: ${....train.collate_fn.composer.melspectrogram_transform}
|
| 125 |
+
audio_key: ${....train.collate_fn.composer.audio_key}
|
| 126 |
+
sample_rate_key: ${....train.collate_fn.composer.sample_rate_key}
|
| 127 |
+
label_name_key: ${....train.collate_fn.composer.label_name_key}
|
| 128 |
+
filename_key: ${....train.collate_fn.composer.filename_key}
|
| 129 |
+
waveform_key: ${....train.collate_fn.composer.waveform_key}
|
| 130 |
+
melspectrogram_key: ${....train.collate_fn.composer.melspectrogram_key}
|
| 131 |
+
label_index_key: ${....train.collate_fn.composer.label_index_key}
|
| 132 |
+
sample_rate: ${....train.collate_fn.composer.sample_rate}
|
| 133 |
+
duration: ${....train.collate_fn.composer.duration}
|
| 134 |
+
decode_audio_as_waveform: ${....train.collate_fn.composer.decode_audio_as_waveform}
|
| 135 |
+
decode_audio_as_monoral: ${....train.collate_fn.composer.decode_audio_as_monoral}
|
| 136 |
+
training: false
|
| 137 |
+
target_shape: ${....train.collate_fn.composer.target_shape}
|
| 138 |
+
melspectrogram_key: ${...train.collate_fn.composer.melspectrogram_key}
|
| 139 |
+
label_index_key: ${...train.collate_fn.composer.label_index_key}
|
| 140 |
+
num_workers: ${const:birdclef2025.utils.data.default_num_workers}
|
| 141 |
+
clip_gradient: {}
|
| 142 |
+
record: {}
|
| 143 |
+
trainer:
|
| 144 |
+
_target_: birdclef2025.utils.driver.BaseTrainer
|
| 145 |
+
key_mapping:
|
| 146 |
+
train:
|
| 147 |
+
input:
|
| 148 |
+
input: ${....dataloader.train.collate_fn.composer.melspectrogram_key}
|
| 149 |
+
output: logit
|
| 150 |
+
validation: ${.train}
|
| 151 |
+
inference: ${.validation}
|
| 152 |
+
ddp_kwargs: null
|
| 153 |
+
resume:
|
| 154 |
+
continue_from: ''
|
| 155 |
+
output:
|
| 156 |
+
exp_dir: ./exp/20250504-153714
|
| 157 |
+
tensorboard_dir: ./tensorboard/20250504-153714
|
| 158 |
+
save_checkpoint:
|
| 159 |
+
iteration:
|
| 160 |
+
every: 10000
|
| 161 |
+
path: ${...exp_dir}/model/iteration{iteration}.pth
|
| 162 |
+
epoch:
|
| 163 |
+
every: 1
|
| 164 |
+
path: ${...exp_dir}/model/epoch{epoch}.pth
|
| 165 |
+
last:
|
| 166 |
+
path: ${...exp_dir}/model/last.pth
|
| 167 |
+
best_epoch:
|
| 168 |
+
path: ${...exp_dir}/model/best_epoch.pth
|
| 169 |
+
steps:
|
| 170 |
+
epochs: 10
|
| 171 |
+
iterations: null
|
| 172 |
+
lr_scheduler: epoch
|
| 173 |
+
test:
|
| 174 |
+
dataset:
|
| 175 |
+
test:
|
| 176 |
+
_target_: torch.utils.data.Dataset
|
| 177 |
+
dataloader:
|
| 178 |
+
test:
|
| 179 |
+
_target_: torch.utils.data.DataLoader
|
| 180 |
+
batch_size: 1
|
| 181 |
+
shuffle: false
|
| 182 |
+
key_mapping:
|
| 183 |
+
inference:
|
| 184 |
+
input: null
|
| 185 |
+
output: null
|
| 186 |
+
identifier: null
|
| 187 |
+
checkpoint: null
|
| 188 |
+
remove_weight_norm: null
|
| 189 |
+
output:
|
| 190 |
+
exp_dir: ./exp
|
| 191 |
+
inference_dir: ${.exp_dir}/inference
|
| 192 |
+
audio:
|
| 193 |
+
sample_rate: ${data.audio.sample_rate}
|
| 194 |
+
key_mapping:
|
| 195 |
+
inference:
|
| 196 |
+
output: null
|
| 197 |
+
reference: null
|
| 198 |
+
transforms:
|
| 199 |
+
inference:
|
| 200 |
+
output: null
|
| 201 |
+
reference: null
|
| 202 |
+
model:
|
| 203 |
+
_target_: birdclef2025.models.ConvNeXtTiny
|
| 204 |
+
weights: ${const:torchvision.models.ConvNeXt_Tiny_Weights.IMAGENET1K_V1}
|
| 205 |
+
num_classes: ${const:birdclef2025.utils.data.birdclef.num_birdclef2025_primary_labels}
|
| 206 |
+
optimizer:
|
| 207 |
+
_target_: torch.optim.AdamW
|
| 208 |
+
lr: 0.0001
|
| 209 |
+
weight_decay: 0.05
|
| 210 |
+
lr_scheduler: {}
|
| 211 |
+
criterion:
|
| 212 |
+
_target_: audyn.criterion.MultiCriteria
|
| 213 |
+
cross_entropy:
|
| 214 |
+
_target_: audyn.criterion.BaseCriterionWrapper
|
| 215 |
+
criterion:
|
| 216 |
+
_target_: torch.nn.CrossEntropyLoss
|
| 217 |
+
reduction: mean
|
| 218 |
+
weight: 1
|
| 219 |
+
key_mapping:
|
| 220 |
+
estimated:
|
| 221 |
+
input: logit
|
| 222 |
+
target:
|
| 223 |
+
target: ${train.dataloader.train.collate_fn.composer.label_index_key}
|
| 224 |
+
metrics:
|
| 225 |
+
roc_auc:
|
| 226 |
+
metric:
|
| 227 |
+
_target_: birdclef2025.metrics.ROCAUC
|
| 228 |
+
take_softmax: true
|
| 229 |
+
key_mapping:
|
| 230 |
+
estimated:
|
| 231 |
+
input: logit
|
| 232 |
+
target:
|
| 233 |
+
target: ${train.dataloader.train.collate_fn.composer.label_index_key}
|
recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717/.hydra/hydra.yaml
ADDED
|
@@ -0,0 +1,191 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
hydra:
|
| 2 |
+
run:
|
| 3 |
+
dir: ./exp/20250504-153714/log/20250504-153717
|
| 4 |
+
sweep:
|
| 5 |
+
dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
|
| 6 |
+
subdir: ${hydra.job.num}
|
| 7 |
+
launcher:
|
| 8 |
+
_target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
|
| 9 |
+
sweeper:
|
| 10 |
+
_target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
|
| 11 |
+
max_batch_size: null
|
| 12 |
+
params: null
|
| 13 |
+
help:
|
| 14 |
+
app_name: ${hydra.job.name}
|
| 15 |
+
header: '${hydra.help.app_name} is powered by Hydra.
|
| 16 |
+
|
| 17 |
+
'
|
| 18 |
+
footer: 'Powered by Hydra (https://hydra.cc)
|
| 19 |
+
|
| 20 |
+
Use --hydra-help to view Hydra specific help
|
| 21 |
+
|
| 22 |
+
'
|
| 23 |
+
template: '${hydra.help.header}
|
| 24 |
+
|
| 25 |
+
== Configuration groups ==
|
| 26 |
+
|
| 27 |
+
Compose your configuration from those groups (group=option)
|
| 28 |
+
|
| 29 |
+
|
| 30 |
+
$APP_CONFIG_GROUPS
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
== Config ==
|
| 34 |
+
|
| 35 |
+
Override anything in the config (foo.bar=value)
|
| 36 |
+
|
| 37 |
+
|
| 38 |
+
$CONFIG
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
${hydra.help.footer}
|
| 42 |
+
|
| 43 |
+
'
|
| 44 |
+
hydra_help:
|
| 45 |
+
template: 'Hydra (${hydra.runtime.version})
|
| 46 |
+
|
| 47 |
+
See https://hydra.cc for more info.
|
| 48 |
+
|
| 49 |
+
|
| 50 |
+
== Flags ==
|
| 51 |
+
|
| 52 |
+
$FLAGS_HELP
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
== Configuration groups ==
|
| 56 |
+
|
| 57 |
+
Compose your configuration from those groups (For example, append hydra/job_logging=disabled
|
| 58 |
+
to command line)
|
| 59 |
+
|
| 60 |
+
|
| 61 |
+
$HYDRA_CONFIG_GROUPS
|
| 62 |
+
|
| 63 |
+
|
| 64 |
+
Use ''--cfg hydra'' to Show the Hydra config.
|
| 65 |
+
|
| 66 |
+
'
|
| 67 |
+
hydra_help: ???
|
| 68 |
+
hydra_logging:
|
| 69 |
+
version: 1
|
| 70 |
+
formatters:
|
| 71 |
+
simple:
|
| 72 |
+
format: '[%(asctime)s][HYDRA] %(message)s'
|
| 73 |
+
handlers:
|
| 74 |
+
console:
|
| 75 |
+
class: logging.StreamHandler
|
| 76 |
+
formatter: simple
|
| 77 |
+
stream: ext://sys.stdout
|
| 78 |
+
root:
|
| 79 |
+
level: INFO
|
| 80 |
+
handlers:
|
| 81 |
+
- console
|
| 82 |
+
loggers:
|
| 83 |
+
logging_example:
|
| 84 |
+
level: DEBUG
|
| 85 |
+
disable_existing_loggers: false
|
| 86 |
+
job_logging:
|
| 87 |
+
version: 1
|
| 88 |
+
formatters:
|
| 89 |
+
simple:
|
| 90 |
+
format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
|
| 91 |
+
handlers:
|
| 92 |
+
console:
|
| 93 |
+
class: logging.StreamHandler
|
| 94 |
+
formatter: simple
|
| 95 |
+
stream: ext://sys.stdout
|
| 96 |
+
file:
|
| 97 |
+
class: logging.FileHandler
|
| 98 |
+
formatter: simple
|
| 99 |
+
filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
|
| 100 |
+
root:
|
| 101 |
+
level: INFO
|
| 102 |
+
handlers:
|
| 103 |
+
- console
|
| 104 |
+
- file
|
| 105 |
+
disable_existing_loggers: false
|
| 106 |
+
env: {}
|
| 107 |
+
mode: RUN
|
| 108 |
+
searchpath: []
|
| 109 |
+
callbacks: {}
|
| 110 |
+
output_subdir: .hydra
|
| 111 |
+
overrides:
|
| 112 |
+
hydra:
|
| 113 |
+
- hydra.run.dir=./exp/20250504-153714/log/20250504-153717
|
| 114 |
+
- hydra.mode=RUN
|
| 115 |
+
task:
|
| 116 |
+
- system=cuda
|
| 117 |
+
- preprocess=birdclef2025
|
| 118 |
+
- data=birdclef2025_reshape_15s
|
| 119 |
+
- train=birdclef2025_reshape_convnext_tiny
|
| 120 |
+
- model=birdclef2025_convnext_tiny
|
| 121 |
+
- optimizer=adamw_1e-4_decay_5e-2
|
| 122 |
+
- lr_scheduler=none
|
| 123 |
+
- criterion=birdclef2025_categorical_cross_entropy
|
| 124 |
+
- +metrics=birdclef2025_categorical_cross_entropy
|
| 125 |
+
- preprocess.dump_format=birdclef2025
|
| 126 |
+
- train.dataset.train.list_path=dump/birdclef2025_reshape_15s/list/train.txt
|
| 127 |
+
- train.dataset.train.feature_dir=/kaggle/input/birdclef-2025
|
| 128 |
+
- train.dataset.validation.list_path=dump/birdclef2025_reshape_15s/list/validation.txt
|
| 129 |
+
- train.dataset.validation.feature_dir=/kaggle/input/birdclef-2025
|
| 130 |
+
- train.resume.continue_from=
|
| 131 |
+
- train.output.exp_dir=./exp/20250504-153714
|
| 132 |
+
- train.output.tensorboard_dir=./tensorboard/20250504-153714
|
| 133 |
+
job:
|
| 134 |
+
name: train
|
| 135 |
+
chdir: false
|
| 136 |
+
override_dirname: +metrics=birdclef2025_categorical_cross_entropy,criterion=birdclef2025_categorical_cross_entropy,data=birdclef2025_reshape_15s,lr_scheduler=none,model=birdclef2025_convnext_tiny,optimizer=adamw_1e-4_decay_5e-2,preprocess.dump_format=birdclef2025,preprocess=birdclef2025,system=cuda,train.dataset.train.feature_dir=/kaggle/input/birdclef-2025,train.dataset.train.list_path=dump/birdclef2025_reshape_15s/list/train.txt,train.dataset.validation.feature_dir=/kaggle/input/birdclef-2025,train.dataset.validation.list_path=dump/birdclef2025_reshape_15s/list/validation.txt,train.output.exp_dir=./exp/20250504-153714,train.output.tensorboard_dir=./tensorboard/20250504-153714,train.resume.continue_from=,train=birdclef2025_reshape_convnext_tiny
|
| 137 |
+
id: ???
|
| 138 |
+
num: ???
|
| 139 |
+
config_name: config
|
| 140 |
+
env_set: {}
|
| 141 |
+
env_copy: []
|
| 142 |
+
config:
|
| 143 |
+
override_dirname:
|
| 144 |
+
kv_sep: '='
|
| 145 |
+
item_sep: ','
|
| 146 |
+
exclude_keys: []
|
| 147 |
+
runtime:
|
| 148 |
+
version: 1.3.2
|
| 149 |
+
version_base: '1.2'
|
| 150 |
+
cwd: /kaggle/working/BirdCLEF2025/recipes/BirdCLEF2025/ConvNeXtTiny
|
| 151 |
+
config_sources:
|
| 152 |
+
- path: hydra.conf
|
| 153 |
+
schema: pkg
|
| 154 |
+
provider: hydra
|
| 155 |
+
- path: /usr/local/lib/python3.10/dist-packages/audyn/configs
|
| 156 |
+
schema: file
|
| 157 |
+
provider: main
|
| 158 |
+
- path: /kaggle/working/BirdCLEF2025/recipes/BirdCLEF2025/ConvNeXtTiny/conf
|
| 159 |
+
schema: file
|
| 160 |
+
provider: command-line
|
| 161 |
+
- path: ''
|
| 162 |
+
schema: structured
|
| 163 |
+
provider: schema
|
| 164 |
+
output_dir: /kaggle/working/BirdCLEF2025/recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717
|
| 165 |
+
choices:
|
| 166 |
+
metrics: birdclef2025_categorical_cross_entropy
|
| 167 |
+
criterion: birdclef2025_categorical_cross_entropy
|
| 168 |
+
lr_scheduler: none
|
| 169 |
+
optimizer: adamw_1e-4_decay_5e-2
|
| 170 |
+
model: birdclef2025_convnext_tiny
|
| 171 |
+
test: default
|
| 172 |
+
test/dataloader: default
|
| 173 |
+
test/dataset: default
|
| 174 |
+
train: birdclef2025_reshape_convnext_tiny
|
| 175 |
+
train/record: default
|
| 176 |
+
train/clip_gradient: default
|
| 177 |
+
train/dataloader: default
|
| 178 |
+
train/dataset: birdclef2025_primary-label
|
| 179 |
+
data: birdclef2025_reshape_15s
|
| 180 |
+
preprocess: birdclef2025
|
| 181 |
+
system: cuda
|
| 182 |
+
hydra/env: default
|
| 183 |
+
hydra/callbacks: null
|
| 184 |
+
hydra/job_logging: default
|
| 185 |
+
hydra/hydra_logging: default
|
| 186 |
+
hydra/hydra_help: default
|
| 187 |
+
hydra/help: default
|
| 188 |
+
hydra/sweeper: basic
|
| 189 |
+
hydra/launcher: basic
|
| 190 |
+
hydra/output: default
|
| 191 |
+
verbose: false
|
recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717/.hydra/overrides.yaml
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
- system=cuda
|
| 2 |
+
- preprocess=birdclef2025
|
| 3 |
+
- data=birdclef2025_reshape_15s
|
| 4 |
+
- train=birdclef2025_reshape_convnext_tiny
|
| 5 |
+
- model=birdclef2025_convnext_tiny
|
| 6 |
+
- optimizer=adamw_1e-4_decay_5e-2
|
| 7 |
+
- lr_scheduler=none
|
| 8 |
+
- criterion=birdclef2025_categorical_cross_entropy
|
| 9 |
+
- +metrics=birdclef2025_categorical_cross_entropy
|
| 10 |
+
- preprocess.dump_format=birdclef2025
|
| 11 |
+
- train.dataset.train.list_path=dump/birdclef2025_reshape_15s/list/train.txt
|
| 12 |
+
- train.dataset.train.feature_dir=/kaggle/input/birdclef-2025
|
| 13 |
+
- train.dataset.validation.list_path=dump/birdclef2025_reshape_15s/list/validation.txt
|
| 14 |
+
- train.dataset.validation.feature_dir=/kaggle/input/birdclef-2025
|
| 15 |
+
- train.resume.continue_from=
|
| 16 |
+
- train.output.exp_dir=./exp/20250504-153714
|
| 17 |
+
- train.output.tensorboard_dir=./tensorboard/20250504-153714
|
recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717/.hydra/resolved_config.yaml
ADDED
|
@@ -0,0 +1,292 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
system:
|
| 2 |
+
seed: 0
|
| 3 |
+
distributed:
|
| 4 |
+
enable: null
|
| 5 |
+
nodes: null
|
| 6 |
+
nproc_per_node: null
|
| 7 |
+
backend: null
|
| 8 |
+
init_method: null
|
| 9 |
+
rdzv_id: null
|
| 10 |
+
rdzv_backend: null
|
| 11 |
+
rdzv_endpoint: null
|
| 12 |
+
max_restarts: null
|
| 13 |
+
cudnn:
|
| 14 |
+
benchmark: true
|
| 15 |
+
deterministic: false
|
| 16 |
+
amp:
|
| 17 |
+
enable: false
|
| 18 |
+
dtype: null
|
| 19 |
+
accelerator: cuda
|
| 20 |
+
compile:
|
| 21 |
+
enable: false
|
| 22 |
+
kwargs: null
|
| 23 |
+
preprocess:
|
| 24 |
+
dump_format: birdclef2025
|
| 25 |
+
list_path: null
|
| 26 |
+
wav_dir: null
|
| 27 |
+
feature_dir: null
|
| 28 |
+
max_workers: 2
|
| 29 |
+
max_shard_size: 1000000000
|
| 30 |
+
vad:
|
| 31 |
+
raw_root: null
|
| 32 |
+
trimmed_root: null
|
| 33 |
+
threshold: null
|
| 34 |
+
min_duration: 15
|
| 35 |
+
csv_path: ???
|
| 36 |
+
submission_path: ???
|
| 37 |
+
audio_root: ???
|
| 38 |
+
subset: ???
|
| 39 |
+
train_ratio: 0.8
|
| 40 |
+
data:
|
| 41 |
+
audio:
|
| 42 |
+
sample_rate: 32000
|
| 43 |
+
duration: 15
|
| 44 |
+
melspectrogram:
|
| 45 |
+
_target_: birdclef2025.transforms.birdclef.BirdCLEF2024BaselineMelSpectrogram
|
| 46 |
+
sample_rate: 32000
|
| 47 |
+
hop_length: 512
|
| 48 |
+
f_min: 20
|
| 49 |
+
f_max: 16000
|
| 50 |
+
pad: 0
|
| 51 |
+
n_mels: 128
|
| 52 |
+
window_fn:
|
| 53 |
+
_target_: torch.hann_window
|
| 54 |
+
_partial_: true
|
| 55 |
+
power: 1.0
|
| 56 |
+
normalized: false
|
| 57 |
+
wkwargs: null
|
| 58 |
+
center: true
|
| 59 |
+
pad_mode: constant
|
| 60 |
+
onesided: null
|
| 61 |
+
norm: slaney
|
| 62 |
+
mel_scale: slaney
|
| 63 |
+
take_log: true
|
| 64 |
+
freq_mask_param:
|
| 65 |
+
- 0.06
|
| 66 |
+
- 0.1
|
| 67 |
+
time_mask_param:
|
| 68 |
+
- 0.06
|
| 69 |
+
- 0.12
|
| 70 |
+
eps: null
|
| 71 |
+
train:
|
| 72 |
+
dataset:
|
| 73 |
+
train:
|
| 74 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025PrimaryLabelDataset
|
| 75 |
+
list_path: dump/birdclef2025_reshape_15s/list/train.txt
|
| 76 |
+
feature_dir: /kaggle/input/birdclef-2025
|
| 77 |
+
audio_key: audio
|
| 78 |
+
sample_rate_key: sample_rate
|
| 79 |
+
label_name_key: primary_label
|
| 80 |
+
filename_key: filename
|
| 81 |
+
validation:
|
| 82 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025PrimaryLabelDataset
|
| 83 |
+
list_path: dump/birdclef2025_reshape_15s/list/validation.txt
|
| 84 |
+
feature_dir: /kaggle/input/birdclef-2025
|
| 85 |
+
audio_key: audio
|
| 86 |
+
sample_rate_key: sample_rate
|
| 87 |
+
label_name_key: primary_label
|
| 88 |
+
filename_key: filename
|
| 89 |
+
dataloader:
|
| 90 |
+
train:
|
| 91 |
+
_target_: torch.utils.data.DataLoader
|
| 92 |
+
batch_size: 64
|
| 93 |
+
shuffle: true
|
| 94 |
+
collate_fn:
|
| 95 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025BaselineCollator
|
| 96 |
+
composer:
|
| 97 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025ReshapePrimaryLabelComposer
|
| 98 |
+
melspectrogram_transform:
|
| 99 |
+
_target_: birdclef2025.transforms.birdclef.BirdCLEF2024BaselineMelSpectrogram
|
| 100 |
+
sample_rate: 32000
|
| 101 |
+
hop_length: 512
|
| 102 |
+
f_min: 20
|
| 103 |
+
f_max: 16000
|
| 104 |
+
pad: 0
|
| 105 |
+
n_mels: 128
|
| 106 |
+
window_fn:
|
| 107 |
+
_target_: torch.hann_window
|
| 108 |
+
_partial_: true
|
| 109 |
+
power: 1.0
|
| 110 |
+
normalized: false
|
| 111 |
+
wkwargs: null
|
| 112 |
+
center: true
|
| 113 |
+
pad_mode: constant
|
| 114 |
+
onesided: null
|
| 115 |
+
norm: slaney
|
| 116 |
+
mel_scale: slaney
|
| 117 |
+
take_log: true
|
| 118 |
+
freq_mask_param:
|
| 119 |
+
- 0.06
|
| 120 |
+
- 0.1
|
| 121 |
+
time_mask_param:
|
| 122 |
+
- 0.06
|
| 123 |
+
- 0.12
|
| 124 |
+
eps: null
|
| 125 |
+
audio_key: audio
|
| 126 |
+
sample_rate_key: sample_rate
|
| 127 |
+
label_name_key: primary_label
|
| 128 |
+
filename_key: filename
|
| 129 |
+
waveform_key: waveform
|
| 130 |
+
melspectrogram_key: log_melspectrogram
|
| 131 |
+
label_index_key: label_index
|
| 132 |
+
sample_rate: 32000
|
| 133 |
+
duration: 15
|
| 134 |
+
decode_audio_as_waveform: true
|
| 135 |
+
decode_audio_as_monoral: true
|
| 136 |
+
training: true
|
| 137 |
+
target_shape: 256
|
| 138 |
+
melspectrogram_key: log_melspectrogram
|
| 139 |
+
label_index_key: label_index
|
| 140 |
+
alpha: 0.4
|
| 141 |
+
num_workers: 2
|
| 142 |
+
validation:
|
| 143 |
+
_target_: torch.utils.data.DataLoader
|
| 144 |
+
batch_size: 64
|
| 145 |
+
shuffle: false
|
| 146 |
+
collate_fn:
|
| 147 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025BaselineValidationCollator
|
| 148 |
+
composer:
|
| 149 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025ReshapePrimaryLabelComposer
|
| 150 |
+
melspectrogram_transform:
|
| 151 |
+
_target_: birdclef2025.transforms.birdclef.BirdCLEF2024BaselineMelSpectrogram
|
| 152 |
+
sample_rate: 32000
|
| 153 |
+
hop_length: 512
|
| 154 |
+
f_min: 20
|
| 155 |
+
f_max: 16000
|
| 156 |
+
pad: 0
|
| 157 |
+
n_mels: 128
|
| 158 |
+
window_fn:
|
| 159 |
+
_target_: torch.hann_window
|
| 160 |
+
_partial_: true
|
| 161 |
+
power: 1.0
|
| 162 |
+
normalized: false
|
| 163 |
+
wkwargs: null
|
| 164 |
+
center: true
|
| 165 |
+
pad_mode: constant
|
| 166 |
+
onesided: null
|
| 167 |
+
norm: slaney
|
| 168 |
+
mel_scale: slaney
|
| 169 |
+
take_log: true
|
| 170 |
+
freq_mask_param:
|
| 171 |
+
- 0.06
|
| 172 |
+
- 0.1
|
| 173 |
+
time_mask_param:
|
| 174 |
+
- 0.06
|
| 175 |
+
- 0.12
|
| 176 |
+
eps: null
|
| 177 |
+
audio_key: audio
|
| 178 |
+
sample_rate_key: sample_rate
|
| 179 |
+
label_name_key: primary_label
|
| 180 |
+
filename_key: filename
|
| 181 |
+
waveform_key: waveform
|
| 182 |
+
melspectrogram_key: log_melspectrogram
|
| 183 |
+
label_index_key: label_index
|
| 184 |
+
sample_rate: 32000
|
| 185 |
+
duration: 15
|
| 186 |
+
decode_audio_as_waveform: true
|
| 187 |
+
decode_audio_as_monoral: true
|
| 188 |
+
training: false
|
| 189 |
+
target_shape: 256
|
| 190 |
+
melspectrogram_key: log_melspectrogram
|
| 191 |
+
label_index_key: label_index
|
| 192 |
+
num_workers: 2
|
| 193 |
+
clip_gradient: {}
|
| 194 |
+
record: {}
|
| 195 |
+
trainer:
|
| 196 |
+
_target_: birdclef2025.utils.driver.BaseTrainer
|
| 197 |
+
key_mapping:
|
| 198 |
+
train:
|
| 199 |
+
input:
|
| 200 |
+
input: log_melspectrogram
|
| 201 |
+
output: logit
|
| 202 |
+
validation:
|
| 203 |
+
input:
|
| 204 |
+
input: log_melspectrogram
|
| 205 |
+
output: logit
|
| 206 |
+
inference:
|
| 207 |
+
input:
|
| 208 |
+
input: log_melspectrogram
|
| 209 |
+
output: logit
|
| 210 |
+
ddp_kwargs: null
|
| 211 |
+
resume:
|
| 212 |
+
continue_from: ''
|
| 213 |
+
output:
|
| 214 |
+
exp_dir: ./exp/20250504-153714
|
| 215 |
+
tensorboard_dir: ./tensorboard/20250504-153714
|
| 216 |
+
save_checkpoint:
|
| 217 |
+
iteration:
|
| 218 |
+
every: 10000
|
| 219 |
+
path: ./exp/20250504-153714/model/iteration{iteration}.pth
|
| 220 |
+
epoch:
|
| 221 |
+
every: 1
|
| 222 |
+
path: ./exp/20250504-153714/model/epoch{epoch}.pth
|
| 223 |
+
last:
|
| 224 |
+
path: ./exp/20250504-153714/model/last.pth
|
| 225 |
+
best_epoch:
|
| 226 |
+
path: ./exp/20250504-153714/model/best_epoch.pth
|
| 227 |
+
steps:
|
| 228 |
+
epochs: 10
|
| 229 |
+
iterations: null
|
| 230 |
+
lr_scheduler: epoch
|
| 231 |
+
test:
|
| 232 |
+
dataset:
|
| 233 |
+
test:
|
| 234 |
+
_target_: torch.utils.data.Dataset
|
| 235 |
+
dataloader:
|
| 236 |
+
test:
|
| 237 |
+
_target_: torch.utils.data.DataLoader
|
| 238 |
+
batch_size: 1
|
| 239 |
+
shuffle: false
|
| 240 |
+
key_mapping:
|
| 241 |
+
inference:
|
| 242 |
+
input: null
|
| 243 |
+
output: null
|
| 244 |
+
identifier: null
|
| 245 |
+
checkpoint: null
|
| 246 |
+
remove_weight_norm: null
|
| 247 |
+
output:
|
| 248 |
+
exp_dir: ./exp
|
| 249 |
+
inference_dir: ./exp/inference
|
| 250 |
+
audio:
|
| 251 |
+
sample_rate: 32000
|
| 252 |
+
key_mapping:
|
| 253 |
+
inference:
|
| 254 |
+
output: null
|
| 255 |
+
reference: null
|
| 256 |
+
transforms:
|
| 257 |
+
inference:
|
| 258 |
+
output: null
|
| 259 |
+
reference: null
|
| 260 |
+
ddp_kwargs: null
|
| 261 |
+
model:
|
| 262 |
+
_target_: birdclef2025.models.ConvNeXtTiny
|
| 263 |
+
weights: IMAGENET1K_V1
|
| 264 |
+
num_classes: 206
|
| 265 |
+
optimizer:
|
| 266 |
+
_target_: torch.optim.AdamW
|
| 267 |
+
lr: 0.0001
|
| 268 |
+
weight_decay: 0.05
|
| 269 |
+
lr_scheduler: {}
|
| 270 |
+
criterion:
|
| 271 |
+
_target_: audyn.criterion.MultiCriteria
|
| 272 |
+
cross_entropy:
|
| 273 |
+
_target_: audyn.criterion.BaseCriterionWrapper
|
| 274 |
+
criterion:
|
| 275 |
+
_target_: torch.nn.CrossEntropyLoss
|
| 276 |
+
reduction: mean
|
| 277 |
+
weight: 1
|
| 278 |
+
key_mapping:
|
| 279 |
+
estimated:
|
| 280 |
+
input: logit
|
| 281 |
+
target:
|
| 282 |
+
target: label_index
|
| 283 |
+
metrics:
|
| 284 |
+
roc_auc:
|
| 285 |
+
metric:
|
| 286 |
+
_target_: birdclef2025.metrics.ROCAUC
|
| 287 |
+
take_softmax: true
|
| 288 |
+
key_mapping:
|
| 289 |
+
estimated:
|
| 290 |
+
input: logit
|
| 291 |
+
target:
|
| 292 |
+
target: label_index
|
recipes/BirdCLEF2025/ConvNeXtTiny/exp/20250504-153714/log/20250504-153717/train.log
ADDED
|
@@ -0,0 +1,853 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[2025-05-04 15:37:45,581][BaseTrainer][INFO] - system:
|
| 2 |
+
seed: 0
|
| 3 |
+
distributed:
|
| 4 |
+
enable: null
|
| 5 |
+
nodes: null
|
| 6 |
+
nproc_per_node: null
|
| 7 |
+
backend: null
|
| 8 |
+
init_method: null
|
| 9 |
+
rdzv_id: null
|
| 10 |
+
rdzv_backend: null
|
| 11 |
+
rdzv_endpoint: null
|
| 12 |
+
max_restarts: null
|
| 13 |
+
cudnn:
|
| 14 |
+
benchmark: true
|
| 15 |
+
deterministic: false
|
| 16 |
+
amp:
|
| 17 |
+
enable: false
|
| 18 |
+
dtype: null
|
| 19 |
+
accelerator: cuda
|
| 20 |
+
compile:
|
| 21 |
+
enable: false
|
| 22 |
+
kwargs: null
|
| 23 |
+
preprocess:
|
| 24 |
+
dump_format: birdclef2025
|
| 25 |
+
list_path: null
|
| 26 |
+
wav_dir: null
|
| 27 |
+
feature_dir: null
|
| 28 |
+
max_workers: 2
|
| 29 |
+
max_shard_size: 1000000000
|
| 30 |
+
vad:
|
| 31 |
+
raw_root: null
|
| 32 |
+
trimmed_root: null
|
| 33 |
+
threshold: null
|
| 34 |
+
min_duration: 15
|
| 35 |
+
csv_path: ???
|
| 36 |
+
submission_path: ???
|
| 37 |
+
audio_root: ???
|
| 38 |
+
subset: ???
|
| 39 |
+
train_ratio: 0.8
|
| 40 |
+
data:
|
| 41 |
+
audio:
|
| 42 |
+
sample_rate: 32000
|
| 43 |
+
duration: 15
|
| 44 |
+
melspectrogram:
|
| 45 |
+
_target_: birdclef2025.transforms.birdclef.BirdCLEF2024BaselineMelSpectrogram
|
| 46 |
+
sample_rate: 32000
|
| 47 |
+
hop_length: 512
|
| 48 |
+
f_min: 20
|
| 49 |
+
f_max: 16000
|
| 50 |
+
pad: 0
|
| 51 |
+
n_mels: 128
|
| 52 |
+
window_fn:
|
| 53 |
+
_target_: torch.hann_window
|
| 54 |
+
_partial_: true
|
| 55 |
+
power: 1.0
|
| 56 |
+
normalized: false
|
| 57 |
+
wkwargs: null
|
| 58 |
+
center: true
|
| 59 |
+
pad_mode: constant
|
| 60 |
+
onesided: null
|
| 61 |
+
norm: slaney
|
| 62 |
+
mel_scale: slaney
|
| 63 |
+
take_log: true
|
| 64 |
+
freq_mask_param:
|
| 65 |
+
- 0.06
|
| 66 |
+
- 0.1
|
| 67 |
+
time_mask_param:
|
| 68 |
+
- 0.06
|
| 69 |
+
- 0.12
|
| 70 |
+
eps: null
|
| 71 |
+
train:
|
| 72 |
+
dataset:
|
| 73 |
+
train:
|
| 74 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025PrimaryLabelDataset
|
| 75 |
+
list_path: dump/birdclef2025_reshape_15s/list/train.txt
|
| 76 |
+
feature_dir: /kaggle/input/birdclef-2025
|
| 77 |
+
audio_key: audio
|
| 78 |
+
sample_rate_key: sample_rate
|
| 79 |
+
label_name_key: primary_label
|
| 80 |
+
filename_key: filename
|
| 81 |
+
validation:
|
| 82 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025PrimaryLabelDataset
|
| 83 |
+
list_path: dump/birdclef2025_reshape_15s/list/validation.txt
|
| 84 |
+
feature_dir: /kaggle/input/birdclef-2025
|
| 85 |
+
audio_key: ${..train.audio_key}
|
| 86 |
+
sample_rate_key: ${..train.sample_rate_key}
|
| 87 |
+
label_name_key: ${..train.label_name_key}
|
| 88 |
+
filename_key: ${..train.filename_key}
|
| 89 |
+
dataloader:
|
| 90 |
+
train:
|
| 91 |
+
_target_: torch.utils.data.DataLoader
|
| 92 |
+
batch_size: 64
|
| 93 |
+
shuffle: true
|
| 94 |
+
collate_fn:
|
| 95 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025BaselineCollator
|
| 96 |
+
composer:
|
| 97 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025ReshapePrimaryLabelComposer
|
| 98 |
+
melspectrogram_transform: ${data.melspectrogram}
|
| 99 |
+
audio_key: audio
|
| 100 |
+
sample_rate_key: sample_rate
|
| 101 |
+
label_name_key: primary_label
|
| 102 |
+
filename_key: filename
|
| 103 |
+
waveform_key: waveform
|
| 104 |
+
melspectrogram_key: log_melspectrogram
|
| 105 |
+
label_index_key: label_index
|
| 106 |
+
sample_rate: ${data.audio.sample_rate}
|
| 107 |
+
duration: ${data.audio.duration}
|
| 108 |
+
decode_audio_as_waveform: true
|
| 109 |
+
decode_audio_as_monoral: true
|
| 110 |
+
training: true
|
| 111 |
+
target_shape: 256
|
| 112 |
+
melspectrogram_key: ${.composer.melspectrogram_key}
|
| 113 |
+
label_index_key: ${.composer.label_index_key}
|
| 114 |
+
alpha: 0.4
|
| 115 |
+
num_workers: ${const:birdclef2025.utils.data.default_num_workers}
|
| 116 |
+
validation:
|
| 117 |
+
_target_: torch.utils.data.DataLoader
|
| 118 |
+
batch_size: 64
|
| 119 |
+
shuffle: false
|
| 120 |
+
collate_fn:
|
| 121 |
+
_target_: birdclef2025.utils.data.birdclef.BirdCLEF2025BaselineValidationCollator
|
| 122 |
+
composer:
|
| 123 |
+
_target_: ${....train.collate_fn.composer._target_}
|
| 124 |
+
melspectrogram_transform: ${....train.collate_fn.composer.melspectrogram_transform}
|
| 125 |
+
audio_key: ${....train.collate_fn.composer.audio_key}
|
| 126 |
+
sample_rate_key: ${....train.collate_fn.composer.sample_rate_key}
|
| 127 |
+
label_name_key: ${....train.collate_fn.composer.label_name_key}
|
| 128 |
+
filename_key: ${....train.collate_fn.composer.filename_key}
|
| 129 |
+
waveform_key: ${....train.collate_fn.composer.waveform_key}
|
| 130 |
+
melspectrogram_key: ${....train.collate_fn.composer.melspectrogram_key}
|
| 131 |
+
label_index_key: ${....train.collate_fn.composer.label_index_key}
|
| 132 |
+
sample_rate: ${....train.collate_fn.composer.sample_rate}
|
| 133 |
+
duration: ${....train.collate_fn.composer.duration}
|
| 134 |
+
decode_audio_as_waveform: ${....train.collate_fn.composer.decode_audio_as_waveform}
|
| 135 |
+
decode_audio_as_monoral: ${....train.collate_fn.composer.decode_audio_as_monoral}
|
| 136 |
+
training: false
|
| 137 |
+
target_shape: ${....train.collate_fn.composer.target_shape}
|
| 138 |
+
melspectrogram_key: ${...train.collate_fn.composer.melspectrogram_key}
|
| 139 |
+
label_index_key: ${...train.collate_fn.composer.label_index_key}
|
| 140 |
+
num_workers: ${const:birdclef2025.utils.data.default_num_workers}
|
| 141 |
+
clip_gradient: {}
|
| 142 |
+
record: {}
|
| 143 |
+
trainer:
|
| 144 |
+
_target_: birdclef2025.utils.driver.BaseTrainer
|
| 145 |
+
_partial_: true
|
| 146 |
+
key_mapping:
|
| 147 |
+
train:
|
| 148 |
+
input:
|
| 149 |
+
input: ${....dataloader.train.collate_fn.composer.melspectrogram_key}
|
| 150 |
+
output: logit
|
| 151 |
+
validation: ${.train}
|
| 152 |
+
inference: ${.validation}
|
| 153 |
+
ddp_kwargs: null
|
| 154 |
+
resume:
|
| 155 |
+
continue_from: ''
|
| 156 |
+
output:
|
| 157 |
+
exp_dir: ./exp/20250504-153714
|
| 158 |
+
tensorboard_dir: ./tensorboard/20250504-153714
|
| 159 |
+
save_checkpoint:
|
| 160 |
+
iteration:
|
| 161 |
+
every: 10000
|
| 162 |
+
path: ${...exp_dir}/model/iteration{iteration}.pth
|
| 163 |
+
epoch:
|
| 164 |
+
every: 1
|
| 165 |
+
path: ${...exp_dir}/model/epoch{epoch}.pth
|
| 166 |
+
last:
|
| 167 |
+
path: ${...exp_dir}/model/last.pth
|
| 168 |
+
best_epoch:
|
| 169 |
+
path: ${...exp_dir}/model/best_epoch.pth
|
| 170 |
+
steps:
|
| 171 |
+
epochs: 10
|
| 172 |
+
iterations: null
|
| 173 |
+
lr_scheduler: epoch
|
| 174 |
+
test:
|
| 175 |
+
dataset:
|
| 176 |
+
test:
|
| 177 |
+
_target_: torch.utils.data.Dataset
|
| 178 |
+
dataloader:
|
| 179 |
+
test:
|
| 180 |
+
_target_: torch.utils.data.DataLoader
|
| 181 |
+
batch_size: 1
|
| 182 |
+
shuffle: false
|
| 183 |
+
key_mapping:
|
| 184 |
+
inference:
|
| 185 |
+
input: null
|
| 186 |
+
output: null
|
| 187 |
+
identifier: null
|
| 188 |
+
checkpoint: null
|
| 189 |
+
remove_weight_norm: null
|
| 190 |
+
output:
|
| 191 |
+
exp_dir: ./exp
|
| 192 |
+
inference_dir: ${.exp_dir}/inference
|
| 193 |
+
audio:
|
| 194 |
+
sample_rate: ${data.audio.sample_rate}
|
| 195 |
+
key_mapping:
|
| 196 |
+
inference:
|
| 197 |
+
output: null
|
| 198 |
+
reference: null
|
| 199 |
+
transforms:
|
| 200 |
+
inference:
|
| 201 |
+
output: null
|
| 202 |
+
reference: null
|
| 203 |
+
ddp_kwargs: null
|
| 204 |
+
model:
|
| 205 |
+
_target_: birdclef2025.models.ConvNeXtTiny
|
| 206 |
+
weights: ${const:torchvision.models.ConvNeXt_Tiny_Weights.IMAGENET1K_V1}
|
| 207 |
+
num_classes: ${const:birdclef2025.utils.data.birdclef.num_birdclef2025_primary_labels}
|
| 208 |
+
optimizer:
|
| 209 |
+
_target_: torch.optim.AdamW
|
| 210 |
+
lr: 0.0001
|
| 211 |
+
weight_decay: 0.05
|
| 212 |
+
lr_scheduler: {}
|
| 213 |
+
criterion:
|
| 214 |
+
_target_: audyn.criterion.MultiCriteria
|
| 215 |
+
cross_entropy:
|
| 216 |
+
_target_: audyn.criterion.BaseCriterionWrapper
|
| 217 |
+
criterion:
|
| 218 |
+
_target_: torch.nn.CrossEntropyLoss
|
| 219 |
+
reduction: mean
|
| 220 |
+
weight: 1
|
| 221 |
+
key_mapping:
|
| 222 |
+
estimated:
|
| 223 |
+
input: logit
|
| 224 |
+
target:
|
| 225 |
+
target: ${train.dataloader.train.collate_fn.composer.label_index_key}
|
| 226 |
+
metrics:
|
| 227 |
+
roc_auc:
|
| 228 |
+
metric:
|
| 229 |
+
_target_: birdclef2025.metrics.ROCAUC
|
| 230 |
+
take_softmax: true
|
| 231 |
+
key_mapping:
|
| 232 |
+
estimated:
|
| 233 |
+
input: logit
|
| 234 |
+
target:
|
| 235 |
+
target: ${train.dataloader.train.collate_fn.composer.label_index_key}
|
| 236 |
+
|
| 237 |
+
[2025-05-04 15:37:45,581][BaseTrainer][INFO] - ConvNeXtTiny(
|
| 238 |
+
(backbone): Sequential(
|
| 239 |
+
(0): Conv2dNormActivation(
|
| 240 |
+
(0): Conv2d(3, 96, kernel_size=(4, 4), stride=(4, 4))
|
| 241 |
+
(1): LayerNorm2d((96,), eps=1e-06, elementwise_affine=True)
|
| 242 |
+
)
|
| 243 |
+
(1): Sequential(
|
| 244 |
+
(0): CNBlock(
|
| 245 |
+
(block): Sequential(
|
| 246 |
+
(0): Conv2d(96, 96, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=96)
|
| 247 |
+
(1): Permute()
|
| 248 |
+
(2): LayerNorm((96,), eps=1e-06, elementwise_affine=True)
|
| 249 |
+
(3): Linear(in_features=96, out_features=384, bias=True)
|
| 250 |
+
(4): GELU(approximate='none')
|
| 251 |
+
(5): Linear(in_features=384, out_features=96, bias=True)
|
| 252 |
+
(6): Permute()
|
| 253 |
+
)
|
| 254 |
+
(stochastic_depth): StochasticDepth(p=0.0, mode=row)
|
| 255 |
+
)
|
| 256 |
+
(1): CNBlock(
|
| 257 |
+
(block): Sequential(
|
| 258 |
+
(0): Conv2d(96, 96, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=96)
|
| 259 |
+
(1): Permute()
|
| 260 |
+
(2): LayerNorm((96,), eps=1e-06, elementwise_affine=True)
|
| 261 |
+
(3): Linear(in_features=96, out_features=384, bias=True)
|
| 262 |
+
(4): GELU(approximate='none')
|
| 263 |
+
(5): Linear(in_features=384, out_features=96, bias=True)
|
| 264 |
+
(6): Permute()
|
| 265 |
+
)
|
| 266 |
+
(stochastic_depth): StochasticDepth(p=0.0058823529411764705, mode=row)
|
| 267 |
+
)
|
| 268 |
+
(2): CNBlock(
|
| 269 |
+
(block): Sequential(
|
| 270 |
+
(0): Conv2d(96, 96, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=96)
|
| 271 |
+
(1): Permute()
|
| 272 |
+
(2): LayerNorm((96,), eps=1e-06, elementwise_affine=True)
|
| 273 |
+
(3): Linear(in_features=96, out_features=384, bias=True)
|
| 274 |
+
(4): GELU(approximate='none')
|
| 275 |
+
(5): Linear(in_features=384, out_features=96, bias=True)
|
| 276 |
+
(6): Permute()
|
| 277 |
+
)
|
| 278 |
+
(stochastic_depth): StochasticDepth(p=0.011764705882352941, mode=row)
|
| 279 |
+
)
|
| 280 |
+
)
|
| 281 |
+
(2): Sequential(
|
| 282 |
+
(0): LayerNorm2d((96,), eps=1e-06, elementwise_affine=True)
|
| 283 |
+
(1): Conv2d(96, 192, kernel_size=(2, 2), stride=(2, 2))
|
| 284 |
+
)
|
| 285 |
+
(3): Sequential(
|
| 286 |
+
(0): CNBlock(
|
| 287 |
+
(block): Sequential(
|
| 288 |
+
(0): Conv2d(192, 192, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=192)
|
| 289 |
+
(1): Permute()
|
| 290 |
+
(2): LayerNorm((192,), eps=1e-06, elementwise_affine=True)
|
| 291 |
+
(3): Linear(in_features=192, out_features=768, bias=True)
|
| 292 |
+
(4): GELU(approximate='none')
|
| 293 |
+
(5): Linear(in_features=768, out_features=192, bias=True)
|
| 294 |
+
(6): Permute()
|
| 295 |
+
)
|
| 296 |
+
(stochastic_depth): StochasticDepth(p=0.017647058823529415, mode=row)
|
| 297 |
+
)
|
| 298 |
+
(1): CNBlock(
|
| 299 |
+
(block): Sequential(
|
| 300 |
+
(0): Conv2d(192, 192, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=192)
|
| 301 |
+
(1): Permute()
|
| 302 |
+
(2): LayerNorm((192,), eps=1e-06, elementwise_affine=True)
|
| 303 |
+
(3): Linear(in_features=192, out_features=768, bias=True)
|
| 304 |
+
(4): GELU(approximate='none')
|
| 305 |
+
(5): Linear(in_features=768, out_features=192, bias=True)
|
| 306 |
+
(6): Permute()
|
| 307 |
+
)
|
| 308 |
+
(stochastic_depth): StochasticDepth(p=0.023529411764705882, mode=row)
|
| 309 |
+
)
|
| 310 |
+
(2): CNBlock(
|
| 311 |
+
(block): Sequential(
|
| 312 |
+
(0): Conv2d(192, 192, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=192)
|
| 313 |
+
(1): Permute()
|
| 314 |
+
(2): LayerNorm((192,), eps=1e-06, elementwise_affine=True)
|
| 315 |
+
(3): Linear(in_features=192, out_features=768, bias=True)
|
| 316 |
+
(4): GELU(approximate='none')
|
| 317 |
+
(5): Linear(in_features=768, out_features=192, bias=True)
|
| 318 |
+
(6): Permute()
|
| 319 |
+
)
|
| 320 |
+
(stochastic_depth): StochasticDepth(p=0.029411764705882353, mode=row)
|
| 321 |
+
)
|
| 322 |
+
)
|
| 323 |
+
(4): Sequential(
|
| 324 |
+
(0): LayerNorm2d((192,), eps=1e-06, elementwise_affine=True)
|
| 325 |
+
(1): Conv2d(192, 384, kernel_size=(2, 2), stride=(2, 2))
|
| 326 |
+
)
|
| 327 |
+
(5): Sequential(
|
| 328 |
+
(0): CNBlock(
|
| 329 |
+
(block): Sequential(
|
| 330 |
+
(0): Conv2d(384, 384, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=384)
|
| 331 |
+
(1): Permute()
|
| 332 |
+
(2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)
|
| 333 |
+
(3): Linear(in_features=384, out_features=1536, bias=True)
|
| 334 |
+
(4): GELU(approximate='none')
|
| 335 |
+
(5): Linear(in_features=1536, out_features=384, bias=True)
|
| 336 |
+
(6): Permute()
|
| 337 |
+
)
|
| 338 |
+
(stochastic_depth): StochasticDepth(p=0.03529411764705883, mode=row)
|
| 339 |
+
)
|
| 340 |
+
(1): CNBlock(
|
| 341 |
+
(block): Sequential(
|
| 342 |
+
(0): Conv2d(384, 384, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=384)
|
| 343 |
+
(1): Permute()
|
| 344 |
+
(2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)
|
| 345 |
+
(3): Linear(in_features=384, out_features=1536, bias=True)
|
| 346 |
+
(4): GELU(approximate='none')
|
| 347 |
+
(5): Linear(in_features=1536, out_features=384, bias=True)
|
| 348 |
+
(6): Permute()
|
| 349 |
+
)
|
| 350 |
+
(stochastic_depth): StochasticDepth(p=0.0411764705882353, mode=row)
|
| 351 |
+
)
|
| 352 |
+
(2): CNBlock(
|
| 353 |
+
(block): Sequential(
|
| 354 |
+
(0): Conv2d(384, 384, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=384)
|
| 355 |
+
(1): Permute()
|
| 356 |
+
(2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)
|
| 357 |
+
(3): Linear(in_features=384, out_features=1536, bias=True)
|
| 358 |
+
(4): GELU(approximate='none')
|
| 359 |
+
(5): Linear(in_features=1536, out_features=384, bias=True)
|
| 360 |
+
(6): Permute()
|
| 361 |
+
)
|
| 362 |
+
(stochastic_depth): StochasticDepth(p=0.047058823529411764, mode=row)
|
| 363 |
+
)
|
| 364 |
+
(3): CNBlock(
|
| 365 |
+
(block): Sequential(
|
| 366 |
+
(0): Conv2d(384, 384, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=384)
|
| 367 |
+
(1): Permute()
|
| 368 |
+
(2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)
|
| 369 |
+
(3): Linear(in_features=384, out_features=1536, bias=True)
|
| 370 |
+
(4): GELU(approximate='none')
|
| 371 |
+
(5): Linear(in_features=1536, out_features=384, bias=True)
|
| 372 |
+
(6): Permute()
|
| 373 |
+
)
|
| 374 |
+
(stochastic_depth): StochasticDepth(p=0.052941176470588235, mode=row)
|
| 375 |
+
)
|
| 376 |
+
(4): CNBlock(
|
| 377 |
+
(block): Sequential(
|
| 378 |
+
(0): Conv2d(384, 384, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=384)
|
| 379 |
+
(1): Permute()
|
| 380 |
+
(2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)
|
| 381 |
+
(3): Linear(in_features=384, out_features=1536, bias=True)
|
| 382 |
+
(4): GELU(approximate='none')
|
| 383 |
+
(5): Linear(in_features=1536, out_features=384, bias=True)
|
| 384 |
+
(6): Permute()
|
| 385 |
+
)
|
| 386 |
+
(stochastic_depth): StochasticDepth(p=0.058823529411764705, mode=row)
|
| 387 |
+
)
|
| 388 |
+
(5): CNBlock(
|
| 389 |
+
(block): Sequential(
|
| 390 |
+
(0): Conv2d(384, 384, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=384)
|
| 391 |
+
(1): Permute()
|
| 392 |
+
(2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)
|
| 393 |
+
(3): Linear(in_features=384, out_features=1536, bias=True)
|
| 394 |
+
(4): GELU(approximate='none')
|
| 395 |
+
(5): Linear(in_features=1536, out_features=384, bias=True)
|
| 396 |
+
(6): Permute()
|
| 397 |
+
)
|
| 398 |
+
(stochastic_depth): StochasticDepth(p=0.06470588235294118, mode=row)
|
| 399 |
+
)
|
| 400 |
+
(6): CNBlock(
|
| 401 |
+
(block): Sequential(
|
| 402 |
+
(0): Conv2d(384, 384, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=384)
|
| 403 |
+
(1): Permute()
|
| 404 |
+
(2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)
|
| 405 |
+
(3): Linear(in_features=384, out_features=1536, bias=True)
|
| 406 |
+
(4): GELU(approximate='none')
|
| 407 |
+
(5): Linear(in_features=1536, out_features=384, bias=True)
|
| 408 |
+
(6): Permute()
|
| 409 |
+
)
|
| 410 |
+
(stochastic_depth): StochasticDepth(p=0.07058823529411766, mode=row)
|
| 411 |
+
)
|
| 412 |
+
(7): CNBlock(
|
| 413 |
+
(block): Sequential(
|
| 414 |
+
(0): Conv2d(384, 384, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=384)
|
| 415 |
+
(1): Permute()
|
| 416 |
+
(2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)
|
| 417 |
+
(3): Linear(in_features=384, out_features=1536, bias=True)
|
| 418 |
+
(4): GELU(approximate='none')
|
| 419 |
+
(5): Linear(in_features=1536, out_features=384, bias=True)
|
| 420 |
+
(6): Permute()
|
| 421 |
+
)
|
| 422 |
+
(stochastic_depth): StochasticDepth(p=0.07647058823529412, mode=row)
|
| 423 |
+
)
|
| 424 |
+
(8): CNBlock(
|
| 425 |
+
(block): Sequential(
|
| 426 |
+
(0): Conv2d(384, 384, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=384)
|
| 427 |
+
(1): Permute()
|
| 428 |
+
(2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)
|
| 429 |
+
(3): Linear(in_features=384, out_features=1536, bias=True)
|
| 430 |
+
(4): GELU(approximate='none')
|
| 431 |
+
(5): Linear(in_features=1536, out_features=384, bias=True)
|
| 432 |
+
(6): Permute()
|
| 433 |
+
)
|
| 434 |
+
(stochastic_depth): StochasticDepth(p=0.0823529411764706, mode=row)
|
| 435 |
+
)
|
| 436 |
+
)
|
| 437 |
+
(6): Sequential(
|
| 438 |
+
(0): LayerNorm2d((384,), eps=1e-06, elementwise_affine=True)
|
| 439 |
+
(1): Conv2d(384, 768, kernel_size=(2, 2), stride=(2, 2))
|
| 440 |
+
)
|
| 441 |
+
(7): Sequential(
|
| 442 |
+
(0): CNBlock(
|
| 443 |
+
(block): Sequential(
|
| 444 |
+
(0): Conv2d(768, 768, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=768)
|
| 445 |
+
(1): Permute()
|
| 446 |
+
(2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
|
| 447 |
+
(3): Linear(in_features=768, out_features=3072, bias=True)
|
| 448 |
+
(4): GELU(approximate='none')
|
| 449 |
+
(5): Linear(in_features=3072, out_features=768, bias=True)
|
| 450 |
+
(6): Permute()
|
| 451 |
+
)
|
| 452 |
+
(stochastic_depth): StochasticDepth(p=0.08823529411764706, mode=row)
|
| 453 |
+
)
|
| 454 |
+
(1): CNBlock(
|
| 455 |
+
(block): Sequential(
|
| 456 |
+
(0): Conv2d(768, 768, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=768)
|
| 457 |
+
(1): Permute()
|
| 458 |
+
(2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
|
| 459 |
+
(3): Linear(in_features=768, out_features=3072, bias=True)
|
| 460 |
+
(4): GELU(approximate='none')
|
| 461 |
+
(5): Linear(in_features=3072, out_features=768, bias=True)
|
| 462 |
+
(6): Permute()
|
| 463 |
+
)
|
| 464 |
+
(stochastic_depth): StochasticDepth(p=0.09411764705882353, mode=row)
|
| 465 |
+
)
|
| 466 |
+
(2): CNBlock(
|
| 467 |
+
(block): Sequential(
|
| 468 |
+
(0): Conv2d(768, 768, kernel_size=(7, 7), stride=(1, 1), padding=(3, 3), groups=768)
|
| 469 |
+
(1): Permute()
|
| 470 |
+
(2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
|
| 471 |
+
(3): Linear(in_features=768, out_features=3072, bias=True)
|
| 472 |
+
(4): GELU(approximate='none')
|
| 473 |
+
(5): Linear(in_features=3072, out_features=768, bias=True)
|
| 474 |
+
(6): Permute()
|
| 475 |
+
)
|
| 476 |
+
(stochastic_depth): StochasticDepth(p=0.1, mode=row)
|
| 477 |
+
)
|
| 478 |
+
)
|
| 479 |
+
)
|
| 480 |
+
(avgpool): AdaptiveAvgPool2d(output_size=1)
|
| 481 |
+
(classifier): Sequential(
|
| 482 |
+
(0): LayerNorm2d((768,), eps=1e-06, elementwise_affine=True)
|
| 483 |
+
(1): Flatten(start_dim=1, end_dim=-1)
|
| 484 |
+
(2): Linear(in_features=768, out_features=206, bias=True)
|
| 485 |
+
)
|
| 486 |
+
)
|
| 487 |
+
[2025-05-04 15:37:45,584][BaseTrainer][INFO] - # of parameters: 27978542.
|
| 488 |
+
[2025-05-04 15:38:05,641][BaseTrainer][INFO] - [Epoch 1/10, Iter 1/3560] 5.376641273498535, cross_entropy: 5.376641273498535
|
| 489 |
+
[2025-05-04 15:38:06,677][BaseTrainer][INFO] - [Epoch 1/10, Iter 2/3560] 5.312026023864746, cross_entropy: 5.312026023864746
|
| 490 |
+
[2025-05-04 15:38:07,712][BaseTrainer][INFO] - [Epoch 1/10, Iter 3/3560] 5.27780818939209, cross_entropy: 5.27780818939209
|
| 491 |
+
[2025-05-04 15:38:08,783][BaseTrainer][INFO] - [Epoch 1/10, Iter 4/3560] 5.321934700012207, cross_entropy: 5.321934700012207
|
| 492 |
+
[2025-05-04 15:38:09,838][BaseTrainer][INFO] - [Epoch 1/10, Iter 5/3560] 5.203423500061035, cross_entropy: 5.203423500061035
|
| 493 |
+
[2025-05-04 15:38:12,173][BaseTrainer][INFO] - [Epoch 1/10, Iter 6/3560] 5.188291549682617, cross_entropy: 5.188291549682617
|
| 494 |
+
[2025-05-04 15:38:13,279][BaseTrainer][INFO] - [Epoch 1/10, Iter 7/3560] 5.233580589294434, cross_entropy: 5.233580589294434
|
| 495 |
+
[2025-05-04 15:38:16,962][BaseTrainer][INFO] - [Epoch 1/10, Iter 8/3560] 5.191223621368408, cross_entropy: 5.191223621368408
|
| 496 |
+
[2025-05-04 15:38:18,014][BaseTrainer][INFO] - [Epoch 1/10, Iter 9/3560] 5.152655601501465, cross_entropy: 5.152655601501465
|
| 497 |
+
[2025-05-04 15:38:21,246][BaseTrainer][INFO] - [Epoch 1/10, Iter 10/3560] 5.084836959838867, cross_entropy: 5.084836959838867
|
| 498 |
+
[2025-05-04 15:38:22,332][BaseTrainer][INFO] - [Epoch 1/10, Iter 11/3560] 5.160213470458984, cross_entropy: 5.160213470458984
|
| 499 |
+
[2025-05-04 15:38:25,576][BaseTrainer][INFO] - [Epoch 1/10, Iter 12/3560] 5.043108940124512, cross_entropy: 5.043108940124512
|
| 500 |
+
[2025-05-04 15:38:26,627][BaseTrainer][INFO] - [Epoch 1/10, Iter 13/3560] 4.957451820373535, cross_entropy: 4.957451820373535
|
| 501 |
+
[2025-05-04 15:38:30,608][BaseTrainer][INFO] - [Epoch 1/10, Iter 14/3560] 4.992649078369141, cross_entropy: 4.992649078369141
|
| 502 |
+
[2025-05-04 15:38:31,670][BaseTrainer][INFO] - [Epoch 1/10, Iter 15/3560] 4.82185173034668, cross_entropy: 4.82185173034668
|
| 503 |
+
[2025-05-04 15:38:34,801][BaseTrainer][INFO] - [Epoch 1/10, Iter 16/3560] 5.031904220581055, cross_entropy: 5.031904220581055
|
| 504 |
+
[2025-05-04 15:38:35,859][BaseTrainer][INFO] - [Epoch 1/10, Iter 17/3560] 5.100973129272461, cross_entropy: 5.100973129272461
|
| 505 |
+
[2025-05-04 15:38:38,908][BaseTrainer][INFO] - [Epoch 1/10, Iter 18/3560] 5.039768218994141, cross_entropy: 5.039768218994141
|
| 506 |
+
[2025-05-04 15:38:39,968][BaseTrainer][INFO] - [Epoch 1/10, Iter 19/3560] 4.945712089538574, cross_entropy: 4.945712089538574
|
| 507 |
+
[2025-05-04 15:38:42,835][BaseTrainer][INFO] - [Epoch 1/10, Iter 20/3560] 4.910451412200928, cross_entropy: 4.910451412200928
|
| 508 |
+
[2025-05-04 15:38:43,887][BaseTrainer][INFO] - [Epoch 1/10, Iter 21/3560] 4.891071319580078, cross_entropy: 4.891071319580078
|
| 509 |
+
[2025-05-04 15:38:47,421][BaseTrainer][INFO] - [Epoch 1/10, Iter 22/3560] 4.818423271179199, cross_entropy: 4.818423271179199
|
| 510 |
+
[2025-05-04 15:38:48,469][BaseTrainer][INFO] - [Epoch 1/10, Iter 23/3560] 4.651951789855957, cross_entropy: 4.651951789855957
|
| 511 |
+
[2025-05-04 15:38:51,379][BaseTrainer][INFO] - [Epoch 1/10, Iter 24/3560] 5.034425735473633, cross_entropy: 5.034425735473633
|
| 512 |
+
[2025-05-04 15:38:52,424][BaseTrainer][INFO] - [Epoch 1/10, Iter 25/3560] 4.565324783325195, cross_entropy: 4.565324783325195
|
| 513 |
+
[2025-05-04 15:38:55,513][BaseTrainer][INFO] - [Epoch 1/10, Iter 26/3560] 4.9761152267456055, cross_entropy: 4.9761152267456055
|
| 514 |
+
[2025-05-04 15:38:56,547][BaseTrainer][INFO] - [Epoch 1/10, Iter 27/3560] 4.841283798217773, cross_entropy: 4.841283798217773
|
| 515 |
+
[2025-05-04 15:38:59,982][BaseTrainer][INFO] - [Epoch 1/10, Iter 28/3560] 4.716811180114746, cross_entropy: 4.716811180114746
|
| 516 |
+
[2025-05-04 15:39:01,035][BaseTrainer][INFO] - [Epoch 1/10, Iter 29/3560] 5.003815650939941, cross_entropy: 5.003815650939941
|
| 517 |
+
[2025-05-04 15:39:04,335][BaseTrainer][INFO] - [Epoch 1/10, Iter 30/3560] 4.7728986740112305, cross_entropy: 4.7728986740112305
|
| 518 |
+
[2025-05-04 15:39:05,380][BaseTrainer][INFO] - [Epoch 1/10, Iter 31/3560] 4.807315826416016, cross_entropy: 4.807315826416016
|
| 519 |
+
[2025-05-04 15:39:07,805][BaseTrainer][INFO] - [Epoch 1/10, Iter 32/3560] 4.718504905700684, cross_entropy: 4.718504905700684
|
| 520 |
+
[2025-05-04 15:39:08,836][BaseTrainer][INFO] - [Epoch 1/10, Iter 33/3560] 4.847777843475342, cross_entropy: 4.847777843475342
|
| 521 |
+
[2025-05-04 15:39:12,467][BaseTrainer][INFO] - [Epoch 1/10, Iter 34/3560] 4.851305961608887, cross_entropy: 4.851305961608887
|
| 522 |
+
[2025-05-04 15:39:13,516][BaseTrainer][INFO] - [Epoch 1/10, Iter 35/3560] 4.6127824783325195, cross_entropy: 4.6127824783325195
|
| 523 |
+
[2025-05-04 15:39:16,284][BaseTrainer][INFO] - [Epoch 1/10, Iter 36/3560] 4.892227649688721, cross_entropy: 4.892227649688721
|
| 524 |
+
[2025-05-04 15:39:17,320][BaseTrainer][INFO] - [Epoch 1/10, Iter 37/3560] 4.760605812072754, cross_entropy: 4.760605812072754
|
| 525 |
+
[2025-05-04 15:39:20,215][BaseTrainer][INFO] - [Epoch 1/10, Iter 38/3560] 4.747416973114014, cross_entropy: 4.747416973114014
|
| 526 |
+
[2025-05-04 15:39:21,269][BaseTrainer][INFO] - [Epoch 1/10, Iter 39/3560] 4.610195159912109, cross_entropy: 4.610195159912109
|
| 527 |
+
[2025-05-04 15:39:24,466][BaseTrainer][INFO] - [Epoch 1/10, Iter 40/3560] 4.788339614868164, cross_entropy: 4.788339614868164
|
| 528 |
+
[2025-05-04 15:39:25,501][BaseTrainer][INFO] - [Epoch 1/10, Iter 41/3560] 4.6864423751831055, cross_entropy: 4.6864423751831055
|
| 529 |
+
[2025-05-04 15:39:28,466][BaseTrainer][INFO] - [Epoch 1/10, Iter 42/3560] 4.7742438316345215, cross_entropy: 4.7742438316345215
|
| 530 |
+
[2025-05-04 15:39:29,499][BaseTrainer][INFO] - [Epoch 1/10, Iter 43/3560] 4.873464584350586, cross_entropy: 4.873464584350586
|
| 531 |
+
[2025-05-04 15:39:32,250][BaseTrainer][INFO] - [Epoch 1/10, Iter 44/3560] 4.77091646194458, cross_entropy: 4.77091646194458
|
| 532 |
+
[2025-05-04 15:39:33,285][BaseTrainer][INFO] - [Epoch 1/10, Iter 45/3560] 4.725327014923096, cross_entropy: 4.725327014923096
|
| 533 |
+
[2025-05-04 15:39:36,199][BaseTrainer][INFO] - [Epoch 1/10, Iter 46/3560] 4.5556559562683105, cross_entropy: 4.5556559562683105
|
| 534 |
+
[2025-05-04 15:39:37,253][BaseTrainer][INFO] - [Epoch 1/10, Iter 47/3560] 4.744180202484131, cross_entropy: 4.744180202484131
|
| 535 |
+
[2025-05-04 15:39:40,399][BaseTrainer][INFO] - [Epoch 1/10, Iter 48/3560] 4.5664777755737305, cross_entropy: 4.5664777755737305
|
| 536 |
+
[2025-05-04 15:39:41,456][BaseTrainer][INFO] - [Epoch 1/10, Iter 49/3560] 4.5559892654418945, cross_entropy: 4.5559892654418945
|
| 537 |
+
[2025-05-04 15:39:46,237][BaseTrainer][INFO] - [Epoch 1/10, Iter 50/3560] 4.736114501953125, cross_entropy: 4.736114501953125
|
| 538 |
+
[2025-05-04 15:39:47,272][BaseTrainer][INFO] - [Epoch 1/10, Iter 51/3560] 4.541403770446777, cross_entropy: 4.541403770446777
|
| 539 |
+
[2025-05-04 15:39:49,705][BaseTrainer][INFO] - [Epoch 1/10, Iter 52/3560] 4.843523979187012, cross_entropy: 4.843523979187012
|
| 540 |
+
[2025-05-04 15:39:50,759][BaseTrainer][INFO] - [Epoch 1/10, Iter 53/3560] 4.459508895874023, cross_entropy: 4.459508895874023
|
| 541 |
+
[2025-05-04 15:39:53,462][BaseTrainer][INFO] - [Epoch 1/10, Iter 54/3560] 4.814382553100586, cross_entropy: 4.814382553100586
|
| 542 |
+
[2025-05-04 15:39:54,506][BaseTrainer][INFO] - [Epoch 1/10, Iter 55/3560] 4.580887794494629, cross_entropy: 4.580887794494629
|
| 543 |
+
[2025-05-04 15:39:57,395][BaseTrainer][INFO] - [Epoch 1/10, Iter 56/3560] 4.71505069732666, cross_entropy: 4.71505069732666
|
| 544 |
+
[2025-05-04 15:39:58,429][BaseTrainer][INFO] - [Epoch 1/10, Iter 57/3560] 4.5077738761901855, cross_entropy: 4.5077738761901855
|
| 545 |
+
[2025-05-04 15:40:01,088][BaseTrainer][INFO] - [Epoch 1/10, Iter 58/3560] 4.75019645690918, cross_entropy: 4.75019645690918
|
| 546 |
+
[2025-05-04 15:40:02,121][BaseTrainer][INFO] - [Epoch 1/10, Iter 59/3560] 4.728370666503906, cross_entropy: 4.728370666503906
|
| 547 |
+
[2025-05-04 15:40:04,650][BaseTrainer][INFO] - [Epoch 1/10, Iter 60/3560] 4.706501483917236, cross_entropy: 4.706501483917236
|
| 548 |
+
[2025-05-04 15:40:05,703][BaseTrainer][INFO] - [Epoch 1/10, Iter 61/3560] 4.695921897888184, cross_entropy: 4.695921897888184
|
| 549 |
+
[2025-05-04 15:40:09,063][BaseTrainer][INFO] - [Epoch 1/10, Iter 62/3560] 4.792250156402588, cross_entropy: 4.792250156402588
|
| 550 |
+
[2025-05-04 15:40:10,106][BaseTrainer][INFO] - [Epoch 1/10, Iter 63/3560] 4.864690780639648, cross_entropy: 4.864690780639648
|
| 551 |
+
[2025-05-04 15:40:14,049][BaseTrainer][INFO] - [Epoch 1/10, Iter 64/3560] 4.703803062438965, cross_entropy: 4.703803062438965
|
| 552 |
+
[2025-05-04 15:40:15,112][BaseTrainer][INFO] - [Epoch 1/10, Iter 65/3560] 4.695968151092529, cross_entropy: 4.695968151092529
|
| 553 |
+
[2025-05-04 15:40:18,446][BaseTrainer][INFO] - [Epoch 1/10, Iter 66/3560] 4.7378034591674805, cross_entropy: 4.7378034591674805
|
| 554 |
+
[2025-05-04 15:40:19,485][BaseTrainer][INFO] - [Epoch 1/10, Iter 67/3560] 4.630002975463867, cross_entropy: 4.630002975463867
|
| 555 |
+
[2025-05-04 15:40:22,457][BaseTrainer][INFO] - [Epoch 1/10, Iter 68/3560] 4.627466201782227, cross_entropy: 4.627466201782227
|
| 556 |
+
[2025-05-04 15:40:23,508][BaseTrainer][INFO] - [Epoch 1/10, Iter 69/3560] 4.664760112762451, cross_entropy: 4.664760112762451
|
| 557 |
+
[2025-05-04 15:40:26,421][BaseTrainer][INFO] - [Epoch 1/10, Iter 70/3560] 4.678044319152832, cross_entropy: 4.678044319152832
|
| 558 |
+
[2025-05-04 15:40:27,454][BaseTrainer][INFO] - [Epoch 1/10, Iter 71/3560] 4.816387176513672, cross_entropy: 4.816387176513672
|
| 559 |
+
[2025-05-04 15:40:30,671][BaseTrainer][INFO] - [Epoch 1/10, Iter 72/3560] 4.488626480102539, cross_entropy: 4.488626480102539
|
| 560 |
+
[2025-05-04 15:40:31,717][BaseTrainer][INFO] - [Epoch 1/10, Iter 73/3560] 4.663271903991699, cross_entropy: 4.663271903991699
|
| 561 |
+
[2025-05-04 15:40:34,328][BaseTrainer][INFO] - [Epoch 1/10, Iter 74/3560] 4.632810592651367, cross_entropy: 4.632810592651367
|
| 562 |
+
[2025-05-04 15:40:35,392][BaseTrainer][INFO] - [Epoch 1/10, Iter 75/3560] 4.598424911499023, cross_entropy: 4.598424911499023
|
| 563 |
+
[2025-05-04 15:40:38,288][BaseTrainer][INFO] - [Epoch 1/10, Iter 76/3560] 4.5883259773254395, cross_entropy: 4.5883259773254395
|
| 564 |
+
[2025-05-04 15:40:39,340][BaseTrainer][INFO] - [Epoch 1/10, Iter 77/3560] 4.6476922035217285, cross_entropy: 4.6476922035217285
|
| 565 |
+
[2025-05-04 15:40:42,592][BaseTrainer][INFO] - [Epoch 1/10, Iter 78/3560] 4.831206798553467, cross_entropy: 4.831206798553467
|
| 566 |
+
[2025-05-04 15:40:43,642][BaseTrainer][INFO] - [Epoch 1/10, Iter 79/3560] 4.717135906219482, cross_entropy: 4.717135906219482
|
| 567 |
+
[2025-05-04 15:40:47,329][BaseTrainer][INFO] - [Epoch 1/10, Iter 80/3560] 4.5759453773498535, cross_entropy: 4.5759453773498535
|
| 568 |
+
[2025-05-04 15:40:48,376][BaseTrainer][INFO] - [Epoch 1/10, Iter 81/3560] 4.878664970397949, cross_entropy: 4.878664970397949
|
| 569 |
+
[2025-05-04 15:40:51,664][BaseTrainer][INFO] - [Epoch 1/10, Iter 82/3560] 4.611279487609863, cross_entropy: 4.611279487609863
|
| 570 |
+
[2025-05-04 15:40:52,698][BaseTrainer][INFO] - [Epoch 1/10, Iter 83/3560] 4.751533508300781, cross_entropy: 4.751533508300781
|
| 571 |
+
[2025-05-04 15:40:55,983][BaseTrainer][INFO] - [Epoch 1/10, Iter 84/3560] 4.457396984100342, cross_entropy: 4.457396984100342
|
| 572 |
+
[2025-05-04 15:40:57,029][BaseTrainer][INFO] - [Epoch 1/10, Iter 85/3560] 4.3965020179748535, cross_entropy: 4.3965020179748535
|
| 573 |
+
[2025-05-04 15:41:00,237][BaseTrainer][INFO] - [Epoch 1/10, Iter 86/3560] 4.610512733459473, cross_entropy: 4.610512733459473
|
| 574 |
+
[2025-05-04 15:41:01,276][BaseTrainer][INFO] - [Epoch 1/10, Iter 87/3560] 4.730538845062256, cross_entropy: 4.730538845062256
|
| 575 |
+
[2025-05-04 15:41:04,260][BaseTrainer][INFO] - [Epoch 1/10, Iter 88/3560] 4.436934471130371, cross_entropy: 4.436934471130371
|
| 576 |
+
[2025-05-04 15:41:05,302][BaseTrainer][INFO] - [Epoch 1/10, Iter 89/3560] 4.580585956573486, cross_entropy: 4.580585956573486
|
| 577 |
+
[2025-05-04 15:41:07,921][BaseTrainer][INFO] - [Epoch 1/10, Iter 90/3560] 4.503360748291016, cross_entropy: 4.503360748291016
|
| 578 |
+
[2025-05-04 15:41:08,952][BaseTrainer][INFO] - [Epoch 1/10, Iter 91/3560] 4.2591400146484375, cross_entropy: 4.2591400146484375
|
| 579 |
+
[2025-05-04 15:41:12,215][BaseTrainer][INFO] - [Epoch 1/10, Iter 92/3560] 4.643581390380859, cross_entropy: 4.643581390380859
|
| 580 |
+
[2025-05-04 15:41:13,274][BaseTrainer][INFO] - [Epoch 1/10, Iter 93/3560] 4.443016052246094, cross_entropy: 4.443016052246094
|
| 581 |
+
[2025-05-04 15:41:16,352][BaseTrainer][INFO] - [Epoch 1/10, Iter 94/3560] 4.586359977722168, cross_entropy: 4.586359977722168
|
| 582 |
+
[2025-05-04 15:41:17,403][BaseTrainer][INFO] - [Epoch 1/10, Iter 95/3560] 4.471511363983154, cross_entropy: 4.471511363983154
|
| 583 |
+
[2025-05-04 15:41:20,271][BaseTrainer][INFO] - [Epoch 1/10, Iter 96/3560] 4.5110931396484375, cross_entropy: 4.5110931396484375
|
| 584 |
+
[2025-05-04 15:41:21,302][BaseTrainer][INFO] - [Epoch 1/10, Iter 97/3560] 4.563319206237793, cross_entropy: 4.563319206237793
|
| 585 |
+
[2025-05-04 15:41:24,537][BaseTrainer][INFO] - [Epoch 1/10, Iter 98/3560] 4.672548294067383, cross_entropy: 4.672548294067383
|
| 586 |
+
[2025-05-04 15:41:25,569][BaseTrainer][INFO] - [Epoch 1/10, Iter 99/3560] 4.615906715393066, cross_entropy: 4.615906715393066
|
| 587 |
+
[2025-05-04 15:41:29,032][BaseTrainer][INFO] - [Epoch 1/10, Iter 100/3560] 4.54000997543335, cross_entropy: 4.54000997543335
|
| 588 |
+
[2025-05-04 15:41:30,065][BaseTrainer][INFO] - [Epoch 1/10, Iter 101/3560] 4.559946060180664, cross_entropy: 4.559946060180664
|
| 589 |
+
[2025-05-04 15:41:32,638][BaseTrainer][INFO] - [Epoch 1/10, Iter 102/3560] 4.563340187072754, cross_entropy: 4.563340187072754
|
| 590 |
+
[2025-05-04 15:41:33,691][BaseTrainer][INFO] - [Epoch 1/10, Iter 103/3560] 4.632449150085449, cross_entropy: 4.632449150085449
|
| 591 |
+
[2025-05-04 15:41:36,213][BaseTrainer][INFO] - [Epoch 1/10, Iter 104/3560] 4.516786575317383, cross_entropy: 4.516786575317383
|
| 592 |
+
[2025-05-04 15:41:37,271][BaseTrainer][INFO] - [Epoch 1/10, Iter 105/3560] 4.154585838317871, cross_entropy: 4.154585838317871
|
| 593 |
+
[2025-05-04 15:41:40,329][BaseTrainer][INFO] - [Epoch 1/10, Iter 106/3560] 4.5116777420043945, cross_entropy: 4.5116777420043945
|
| 594 |
+
[2025-05-04 15:41:41,384][BaseTrainer][INFO] - [Epoch 1/10, Iter 107/3560] 4.780315399169922, cross_entropy: 4.780315399169922
|
| 595 |
+
[2025-05-04 15:41:44,375][BaseTrainer][INFO] - [Epoch 1/10, Iter 108/3560] 4.336567401885986, cross_entropy: 4.336567401885986
|
| 596 |
+
[2025-05-04 15:41:45,410][BaseTrainer][INFO] - [Epoch 1/10, Iter 109/3560] 4.303857326507568, cross_entropy: 4.303857326507568
|
| 597 |
+
[2025-05-04 15:41:48,564][BaseTrainer][INFO] - [Epoch 1/10, Iter 110/3560] 4.627621173858643, cross_entropy: 4.627621173858643
|
| 598 |
+
[2025-05-04 15:41:49,628][BaseTrainer][INFO] - [Epoch 1/10, Iter 111/3560] 4.187675476074219, cross_entropy: 4.187675476074219
|
| 599 |
+
[2025-05-04 15:41:53,056][BaseTrainer][INFO] - [Epoch 1/10, Iter 112/3560] 4.382655143737793, cross_entropy: 4.382655143737793
|
| 600 |
+
[2025-05-04 15:41:54,103][BaseTrainer][INFO] - [Epoch 1/10, Iter 113/3560] 4.612682819366455, cross_entropy: 4.612682819366455
|
| 601 |
+
[2025-05-04 15:41:57,509][BaseTrainer][INFO] - [Epoch 1/10, Iter 114/3560] 4.566553115844727, cross_entropy: 4.566553115844727
|
| 602 |
+
[2025-05-04 15:41:58,566][BaseTrainer][INFO] - [Epoch 1/10, Iter 115/3560] 4.172827243804932, cross_entropy: 4.172827243804932
|
| 603 |
+
[2025-05-04 15:42:02,412][BaseTrainer][INFO] - [Epoch 1/10, Iter 116/3560] 4.707499027252197, cross_entropy: 4.707499027252197
|
| 604 |
+
[2025-05-04 15:42:03,462][BaseTrainer][INFO] - [Epoch 1/10, Iter 117/3560] 4.260464668273926, cross_entropy: 4.260464668273926
|
| 605 |
+
[2025-05-04 15:42:06,088][BaseTrainer][INFO] - [Epoch 1/10, Iter 118/3560] 4.38938045501709, cross_entropy: 4.38938045501709
|
| 606 |
+
[2025-05-04 15:42:07,123][BaseTrainer][INFO] - [Epoch 1/10, Iter 119/3560] 4.397576332092285, cross_entropy: 4.397576332092285
|
| 607 |
+
[2025-05-04 15:42:09,779][BaseTrainer][INFO] - [Epoch 1/10, Iter 120/3560] 4.336797714233398, cross_entropy: 4.336797714233398
|
| 608 |
+
[2025-05-04 15:42:10,810][BaseTrainer][INFO] - [Epoch 1/10, Iter 121/3560] 4.763151168823242, cross_entropy: 4.763151168823242
|
| 609 |
+
[2025-05-04 15:42:13,750][BaseTrainer][INFO] - [Epoch 1/10, Iter 122/3560] 4.443341255187988, cross_entropy: 4.443341255187988
|
| 610 |
+
[2025-05-04 15:42:14,798][BaseTrainer][INFO] - [Epoch 1/10, Iter 123/3560] 4.7432756423950195, cross_entropy: 4.7432756423950195
|
| 611 |
+
[2025-05-04 15:42:17,478][BaseTrainer][INFO] - [Epoch 1/10, Iter 124/3560] 4.287132263183594, cross_entropy: 4.287132263183594
|
| 612 |
+
[2025-05-04 15:42:18,533][BaseTrainer][INFO] - [Epoch 1/10, Iter 125/3560] 4.172094345092773, cross_entropy: 4.172094345092773
|
| 613 |
+
[2025-05-04 15:42:21,268][BaseTrainer][INFO] - [Epoch 1/10, Iter 126/3560] 4.541016101837158, cross_entropy: 4.541016101837158
|
| 614 |
+
[2025-05-04 15:42:22,325][BaseTrainer][INFO] - [Epoch 1/10, Iter 127/3560] 4.759217262268066, cross_entropy: 4.759217262268066
|
| 615 |
+
[2025-05-04 15:42:25,181][BaseTrainer][INFO] - [Epoch 1/10, Iter 128/3560] 4.484977722167969, cross_entropy: 4.484977722167969
|
| 616 |
+
[2025-05-04 15:42:26,231][BaseTrainer][INFO] - [Epoch 1/10, Iter 129/3560] 4.228219985961914, cross_entropy: 4.228219985961914
|
| 617 |
+
[2025-05-04 15:42:28,768][BaseTrainer][INFO] - [Epoch 1/10, Iter 130/3560] 4.172601699829102, cross_entropy: 4.172601699829102
|
| 618 |
+
[2025-05-04 15:42:29,830][BaseTrainer][INFO] - [Epoch 1/10, Iter 131/3560] 4.271279811859131, cross_entropy: 4.271279811859131
|
| 619 |
+
[2025-05-04 15:42:32,993][BaseTrainer][INFO] - [Epoch 1/10, Iter 132/3560] 4.379694938659668, cross_entropy: 4.379694938659668
|
| 620 |
+
[2025-05-04 15:42:34,026][BaseTrainer][INFO] - [Epoch 1/10, Iter 133/3560] 3.976646900177002, cross_entropy: 3.976646900177002
|
| 621 |
+
[2025-05-04 15:42:36,696][BaseTrainer][INFO] - [Epoch 1/10, Iter 134/3560] 4.2220282554626465, cross_entropy: 4.2220282554626465
|
| 622 |
+
[2025-05-04 15:42:37,735][BaseTrainer][INFO] - [Epoch 1/10, Iter 135/3560] 4.535080909729004, cross_entropy: 4.535080909729004
|
| 623 |
+
[2025-05-04 15:42:40,535][BaseTrainer][INFO] - [Epoch 1/10, Iter 136/3560] 4.378235816955566, cross_entropy: 4.378235816955566
|
| 624 |
+
[2025-05-04 15:42:41,574][BaseTrainer][INFO] - [Epoch 1/10, Iter 137/3560] 4.188755035400391, cross_entropy: 4.188755035400391
|
| 625 |
+
[2025-05-04 15:42:44,207][BaseTrainer][INFO] - [Epoch 1/10, Iter 138/3560] 4.1770477294921875, cross_entropy: 4.1770477294921875
|
| 626 |
+
[2025-05-04 15:42:45,241][BaseTrainer][INFO] - [Epoch 1/10, Iter 139/3560] 4.205631256103516, cross_entropy: 4.205631256103516
|
| 627 |
+
[2025-05-04 15:42:48,137][BaseTrainer][INFO] - [Epoch 1/10, Iter 140/3560] 4.314923286437988, cross_entropy: 4.314923286437988
|
| 628 |
+
[2025-05-04 15:42:49,193][BaseTrainer][INFO] - [Epoch 1/10, Iter 141/3560] 4.068292617797852, cross_entropy: 4.068292617797852
|
| 629 |
+
[2025-05-04 15:42:53,460][BaseTrainer][INFO] - [Epoch 1/10, Iter 142/3560] 4.10396671295166, cross_entropy: 4.10396671295166
|
| 630 |
+
[2025-05-04 15:42:54,513][BaseTrainer][INFO] - [Epoch 1/10, Iter 143/3560] 4.234318733215332, cross_entropy: 4.234318733215332
|
| 631 |
+
[2025-05-04 15:42:57,367][BaseTrainer][INFO] - [Epoch 1/10, Iter 144/3560] 4.579281806945801, cross_entropy: 4.579281806945801
|
| 632 |
+
[2025-05-04 15:42:58,400][BaseTrainer][INFO] - [Epoch 1/10, Iter 145/3560] 4.095945358276367, cross_entropy: 4.095945358276367
|
| 633 |
+
[2025-05-04 15:43:01,556][BaseTrainer][INFO] - [Epoch 1/10, Iter 146/3560] 4.273804664611816, cross_entropy: 4.273804664611816
|
| 634 |
+
[2025-05-04 15:43:02,612][BaseTrainer][INFO] - [Epoch 1/10, Iter 147/3560] 4.2708587646484375, cross_entropy: 4.2708587646484375
|
| 635 |
+
[2025-05-04 15:43:05,603][BaseTrainer][INFO] - [Epoch 1/10, Iter 148/3560] 3.9428675174713135, cross_entropy: 3.9428675174713135
|
| 636 |
+
[2025-05-04 15:43:06,658][BaseTrainer][INFO] - [Epoch 1/10, Iter 149/3560] 4.465112686157227, cross_entropy: 4.465112686157227
|
| 637 |
+
[2025-05-04 15:43:09,488][BaseTrainer][INFO] - [Epoch 1/10, Iter 150/3560] 4.029943943023682, cross_entropy: 4.029943943023682
|
| 638 |
+
[2025-05-04 15:43:10,519][BaseTrainer][INFO] - [Epoch 1/10, Iter 151/3560] 3.805082321166992, cross_entropy: 3.805082321166992
|
| 639 |
+
[2025-05-04 15:43:13,695][BaseTrainer][INFO] - [Epoch 1/10, Iter 152/3560] 3.835386276245117, cross_entropy: 3.835386276245117
|
| 640 |
+
[2025-05-04 15:43:14,727][BaseTrainer][INFO] - [Epoch 1/10, Iter 153/3560] 4.022421836853027, cross_entropy: 4.022421836853027
|
| 641 |
+
[2025-05-04 15:43:17,758][BaseTrainer][INFO] - [Epoch 1/10, Iter 154/3560] 4.256207466125488, cross_entropy: 4.256207466125488
|
| 642 |
+
[2025-05-04 15:43:18,811][BaseTrainer][INFO] - [Epoch 1/10, Iter 155/3560] 3.787353038787842, cross_entropy: 3.787353038787842
|
| 643 |
+
[2025-05-04 15:43:22,536][BaseTrainer][INFO] - [Epoch 1/10, Iter 156/3560] 3.814817428588867, cross_entropy: 3.814817428588867
|
| 644 |
+
[2025-05-04 15:43:23,588][BaseTrainer][INFO] - [Epoch 1/10, Iter 157/3560] 4.2981061935424805, cross_entropy: 4.2981061935424805
|
| 645 |
+
[2025-05-04 15:43:26,502][BaseTrainer][INFO] - [Epoch 1/10, Iter 158/3560] 4.3363037109375, cross_entropy: 4.3363037109375
|
| 646 |
+
[2025-05-04 15:43:27,561][BaseTrainer][INFO] - [Epoch 1/10, Iter 159/3560] 3.993943214416504, cross_entropy: 3.993943214416504
|
| 647 |
+
[2025-05-04 15:43:31,023][BaseTrainer][INFO] - [Epoch 1/10, Iter 160/3560] 3.8628408908843994, cross_entropy: 3.8628408908843994
|
| 648 |
+
[2025-05-04 15:43:32,057][BaseTrainer][INFO] - [Epoch 1/10, Iter 161/3560] 3.870957851409912, cross_entropy: 3.870957851409912
|
| 649 |
+
[2025-05-04 15:43:35,409][BaseTrainer][INFO] - [Epoch 1/10, Iter 162/3560] 3.827423572540283, cross_entropy: 3.827423572540283
|
| 650 |
+
[2025-05-04 15:43:36,447][BaseTrainer][INFO] - [Epoch 1/10, Iter 163/3560] 3.8112123012542725, cross_entropy: 3.8112123012542725
|
| 651 |
+
[2025-05-04 15:43:39,359][BaseTrainer][INFO] - [Epoch 1/10, Iter 164/3560] 3.975653886795044, cross_entropy: 3.975653886795044
|
| 652 |
+
[2025-05-04 15:43:40,393][BaseTrainer][INFO] - [Epoch 1/10, Iter 165/3560] 4.278404712677002, cross_entropy: 4.278404712677002
|
| 653 |
+
[2025-05-04 15:43:43,182][BaseTrainer][INFO] - [Epoch 1/10, Iter 166/3560] 4.291914939880371, cross_entropy: 4.291914939880371
|
| 654 |
+
[2025-05-04 15:43:44,215][BaseTrainer][INFO] - [Epoch 1/10, Iter 167/3560] 3.8860697746276855, cross_entropy: 3.8860697746276855
|
| 655 |
+
[2025-05-04 15:43:47,243][BaseTrainer][INFO] - [Epoch 1/10, Iter 168/3560] 3.97953200340271, cross_entropy: 3.97953200340271
|
| 656 |
+
[2025-05-04 15:43:48,273][BaseTrainer][INFO] - [Epoch 1/10, Iter 169/3560] 4.418849945068359, cross_entropy: 4.418849945068359
|
| 657 |
+
[2025-05-04 15:43:52,155][BaseTrainer][INFO] - [Epoch 1/10, Iter 170/3560] 3.99082612991333, cross_entropy: 3.99082612991333
|
| 658 |
+
[2025-05-04 15:43:53,212][BaseTrainer][INFO] - [Epoch 1/10, Iter 171/3560] 3.6433072090148926, cross_entropy: 3.6433072090148926
|
| 659 |
+
[2025-05-04 15:43:56,167][BaseTrainer][INFO] - [Epoch 1/10, Iter 172/3560] 3.890350818634033, cross_entropy: 3.890350818634033
|
| 660 |
+
[2025-05-04 15:43:57,217][BaseTrainer][INFO] - [Epoch 1/10, Iter 173/3560] 3.7742199897766113, cross_entropy: 3.7742199897766113
|
| 661 |
+
[2025-05-04 15:44:00,341][BaseTrainer][INFO] - [Epoch 1/10, Iter 174/3560] 4.139997959136963, cross_entropy: 4.139997959136963
|
| 662 |
+
[2025-05-04 15:44:01,397][BaseTrainer][INFO] - [Epoch 1/10, Iter 175/3560] 4.463659286499023, cross_entropy: 4.463659286499023
|
| 663 |
+
[2025-05-04 15:44:04,941][BaseTrainer][INFO] - [Epoch 1/10, Iter 176/3560] 3.935756206512451, cross_entropy: 3.935756206512451
|
| 664 |
+
[2025-05-04 15:44:05,996][BaseTrainer][INFO] - [Epoch 1/10, Iter 177/3560] 3.7262871265411377, cross_entropy: 3.7262871265411377
|
| 665 |
+
[2025-05-04 15:44:09,246][BaseTrainer][INFO] - [Epoch 1/10, Iter 178/3560] 3.7987825870513916, cross_entropy: 3.7987825870513916
|
| 666 |
+
[2025-05-04 15:44:10,308][BaseTrainer][INFO] - [Epoch 1/10, Iter 179/3560] 3.6240315437316895, cross_entropy: 3.6240315437316895
|
| 667 |
+
[2025-05-04 15:44:13,550][BaseTrainer][INFO] - [Epoch 1/10, Iter 180/3560] 3.700782060623169, cross_entropy: 3.700782060623169
|
| 668 |
+
[2025-05-04 15:44:14,581][BaseTrainer][INFO] - [Epoch 1/10, Iter 181/3560] 3.962472438812256, cross_entropy: 3.962472438812256
|
| 669 |
+
[2025-05-04 15:44:18,026][BaseTrainer][INFO] - [Epoch 1/10, Iter 182/3560] 3.7632155418395996, cross_entropy: 3.7632155418395996
|
| 670 |
+
[2025-05-04 15:44:19,058][BaseTrainer][INFO] - [Epoch 1/10, Iter 183/3560] 3.956826686859131, cross_entropy: 3.956826686859131
|
| 671 |
+
[2025-05-04 15:44:23,384][BaseTrainer][INFO] - [Epoch 1/10, Iter 184/3560] 4.246981620788574, cross_entropy: 4.246981620788574
|
| 672 |
+
[2025-05-04 15:44:24,441][BaseTrainer][INFO] - [Epoch 1/10, Iter 185/3560] 3.8731627464294434, cross_entropy: 3.8731627464294434
|
| 673 |
+
[2025-05-04 15:44:27,344][BaseTrainer][INFO] - [Epoch 1/10, Iter 186/3560] 4.063902378082275, cross_entropy: 4.063902378082275
|
| 674 |
+
[2025-05-04 15:44:28,377][BaseTrainer][INFO] - [Epoch 1/10, Iter 187/3560] 3.6894967555999756, cross_entropy: 3.6894967555999756
|
| 675 |
+
[2025-05-04 15:44:32,188][BaseTrainer][INFO] - [Epoch 1/10, Iter 188/3560] 4.099838733673096, cross_entropy: 4.099838733673096
|
| 676 |
+
[2025-05-04 15:44:33,243][BaseTrainer][INFO] - [Epoch 1/10, Iter 189/3560] 4.2275166511535645, cross_entropy: 4.2275166511535645
|
| 677 |
+
[2025-05-04 15:44:35,945][BaseTrainer][INFO] - [Epoch 1/10, Iter 190/3560] 3.468061923980713, cross_entropy: 3.468061923980713
|
| 678 |
+
[2025-05-04 15:44:36,977][BaseTrainer][INFO] - [Epoch 1/10, Iter 191/3560] 4.105533599853516, cross_entropy: 4.105533599853516
|
| 679 |
+
[2025-05-04 15:44:39,745][BaseTrainer][INFO] - [Epoch 1/10, Iter 192/3560] 3.783463716506958, cross_entropy: 3.783463716506958
|
| 680 |
+
[2025-05-04 15:44:40,797][BaseTrainer][INFO] - [Epoch 1/10, Iter 193/3560] 4.128844738006592, cross_entropy: 4.128844738006592
|
| 681 |
+
[2025-05-04 15:44:43,553][BaseTrainer][INFO] - [Epoch 1/10, Iter 194/3560] 3.705315351486206, cross_entropy: 3.705315351486206
|
| 682 |
+
[2025-05-04 15:44:44,584][BaseTrainer][INFO] - [Epoch 1/10, Iter 195/3560] 4.0855913162231445, cross_entropy: 4.0855913162231445
|
| 683 |
+
[2025-05-04 15:44:47,041][BaseTrainer][INFO] - [Epoch 1/10, Iter 196/3560] 3.921950578689575, cross_entropy: 3.921950578689575
|
| 684 |
+
[2025-05-04 15:44:48,223][BaseTrainer][INFO] - [Epoch 1/10, Iter 197/3560] 4.081811428070068, cross_entropy: 4.081811428070068
|
| 685 |
+
[2025-05-04 15:44:51,484][BaseTrainer][INFO] - [Epoch 1/10, Iter 198/3560] 3.353686571121216, cross_entropy: 3.353686571121216
|
| 686 |
+
[2025-05-04 15:44:52,563][BaseTrainer][INFO] - [Epoch 1/10, Iter 199/3560] 3.644573450088501, cross_entropy: 3.644573450088501
|
| 687 |
+
[2025-05-04 15:44:56,034][BaseTrainer][INFO] - [Epoch 1/10, Iter 200/3560] 3.7141542434692383, cross_entropy: 3.7141542434692383
|
| 688 |
+
[2025-05-04 15:44:57,084][BaseTrainer][INFO] - [Epoch 1/10, Iter 201/3560] 3.6063575744628906, cross_entropy: 3.6063575744628906
|
| 689 |
+
[2025-05-04 15:45:00,073][BaseTrainer][INFO] - [Epoch 1/10, Iter 202/3560] 3.51749849319458, cross_entropy: 3.51749849319458
|
| 690 |
+
[2025-05-04 15:45:01,284][BaseTrainer][INFO] - [Epoch 1/10, Iter 203/3560] 4.176861763000488, cross_entropy: 4.176861763000488
|
| 691 |
+
[2025-05-04 15:45:04,357][BaseTrainer][INFO] - [Epoch 1/10, Iter 204/3560] 3.717007875442505, cross_entropy: 3.717007875442505
|
| 692 |
+
[2025-05-04 15:45:05,403][BaseTrainer][INFO] - [Epoch 1/10, Iter 205/3560] 4.427414894104004, cross_entropy: 4.427414894104004
|
| 693 |
+
[2025-05-04 15:45:08,656][BaseTrainer][INFO] - [Epoch 1/10, Iter 206/3560] 3.8307533264160156, cross_entropy: 3.8307533264160156
|
| 694 |
+
[2025-05-04 15:45:09,853][BaseTrainer][INFO] - [Epoch 1/10, Iter 207/3560] 3.3228912353515625, cross_entropy: 3.3228912353515625
|
| 695 |
+
[2025-05-04 15:45:12,957][BaseTrainer][INFO] - [Epoch 1/10, Iter 208/3560] 3.6359074115753174, cross_entropy: 3.6359074115753174
|
| 696 |
+
[2025-05-04 15:45:14,007][BaseTrainer][INFO] - [Epoch 1/10, Iter 209/3560] 3.6332781314849854, cross_entropy: 3.6332781314849854
|
| 697 |
+
[2025-05-04 15:45:16,652][BaseTrainer][INFO] - [Epoch 1/10, Iter 210/3560] 3.547818183898926, cross_entropy: 3.547818183898926
|
| 698 |
+
[2025-05-04 15:45:17,687][BaseTrainer][INFO] - [Epoch 1/10, Iter 211/3560] 3.799422264099121, cross_entropy: 3.799422264099121
|
| 699 |
+
[2025-05-04 15:45:20,857][BaseTrainer][INFO] - [Epoch 1/10, Iter 212/3560] 3.9624199867248535, cross_entropy: 3.9624199867248535
|
| 700 |
+
[2025-05-04 15:45:21,986][BaseTrainer][INFO] - [Epoch 1/10, Iter 213/3560] 4.425595283508301, cross_entropy: 4.425595283508301
|
| 701 |
+
[2025-05-04 15:45:25,283][BaseTrainer][INFO] - [Epoch 1/10, Iter 214/3560] 3.5650367736816406, cross_entropy: 3.5650367736816406
|
| 702 |
+
[2025-05-04 15:45:26,335][BaseTrainer][INFO] - [Epoch 1/10, Iter 215/3560] 3.262298107147217, cross_entropy: 3.262298107147217
|
| 703 |
+
[2025-05-04 15:45:29,442][BaseTrainer][INFO] - [Epoch 1/10, Iter 216/3560] 4.325966835021973, cross_entropy: 4.325966835021973
|
| 704 |
+
[2025-05-04 15:45:30,496][BaseTrainer][INFO] - [Epoch 1/10, Iter 217/3560] 3.4640183448791504, cross_entropy: 3.4640183448791504
|
| 705 |
+
[2025-05-04 15:45:33,518][BaseTrainer][INFO] - [Epoch 1/10, Iter 218/3560] 3.00398588180542, cross_entropy: 3.00398588180542
|
| 706 |
+
[2025-05-04 15:45:34,571][BaseTrainer][INFO] - [Epoch 1/10, Iter 219/3560] 4.193729400634766, cross_entropy: 4.193729400634766
|
| 707 |
+
[2025-05-04 15:45:38,244][BaseTrainer][INFO] - [Epoch 1/10, Iter 220/3560] 4.179624557495117, cross_entropy: 4.179624557495117
|
| 708 |
+
[2025-05-04 15:45:39,346][BaseTrainer][INFO] - [Epoch 1/10, Iter 221/3560] 3.7965245246887207, cross_entropy: 3.7965245246887207
|
| 709 |
+
[2025-05-04 15:45:42,460][BaseTrainer][INFO] - [Epoch 1/10, Iter 222/3560] 3.20782470703125, cross_entropy: 3.20782470703125
|
| 710 |
+
[2025-05-04 15:45:43,509][BaseTrainer][INFO] - [Epoch 1/10, Iter 223/3560] 4.157542705535889, cross_entropy: 4.157542705535889
|
| 711 |
+
[2025-05-04 15:45:45,875][BaseTrainer][INFO] - [Epoch 1/10, Iter 224/3560] 3.989438533782959, cross_entropy: 3.989438533782959
|
| 712 |
+
[2025-05-04 15:45:47,171][BaseTrainer][INFO] - [Epoch 1/10, Iter 225/3560] 3.6186747550964355, cross_entropy: 3.6186747550964355
|
| 713 |
+
[2025-05-04 15:45:50,109][BaseTrainer][INFO] - [Epoch 1/10, Iter 226/3560] 3.1917433738708496, cross_entropy: 3.1917433738708496
|
| 714 |
+
[2025-05-04 15:45:51,730][BaseTrainer][INFO] - [Epoch 1/10, Iter 227/3560] 3.5871357917785645, cross_entropy: 3.5871357917785645
|
| 715 |
+
[2025-05-04 15:45:54,999][BaseTrainer][INFO] - [Epoch 1/10, Iter 228/3560] 3.428403377532959, cross_entropy: 3.428403377532959
|
| 716 |
+
[2025-05-04 15:45:56,089][BaseTrainer][INFO] - [Epoch 1/10, Iter 229/3560] 3.632758617401123, cross_entropy: 3.632758617401123
|
| 717 |
+
[2025-05-04 15:45:58,615][BaseTrainer][INFO] - [Epoch 1/10, Iter 230/3560] 3.972907066345215, cross_entropy: 3.972907066345215
|
| 718 |
+
[2025-05-04 15:45:59,732][BaseTrainer][INFO] - [Epoch 1/10, Iter 231/3560] 3.840569257736206, cross_entropy: 3.840569257736206
|
| 719 |
+
[2025-05-04 15:46:03,151][BaseTrainer][INFO] - [Epoch 1/10, Iter 232/3560] 3.692924976348877, cross_entropy: 3.692924976348877
|
| 720 |
+
[2025-05-04 15:46:04,208][BaseTrainer][INFO] - [Epoch 1/10, Iter 233/3560] 3.61606502532959, cross_entropy: 3.61606502532959
|
| 721 |
+
[2025-05-04 15:46:08,442][BaseTrainer][INFO] - [Epoch 1/10, Iter 234/3560] 3.608592987060547, cross_entropy: 3.608592987060547
|
| 722 |
+
[2025-05-04 15:46:09,543][BaseTrainer][INFO] - [Epoch 1/10, Iter 235/3560] 3.3924107551574707, cross_entropy: 3.3924107551574707
|
| 723 |
+
[2025-05-04 15:46:12,410][BaseTrainer][INFO] - [Epoch 1/10, Iter 236/3560] 3.389481544494629, cross_entropy: 3.389481544494629
|
| 724 |
+
[2025-05-04 15:46:13,458][BaseTrainer][INFO] - [Epoch 1/10, Iter 237/3560] 3.601651906967163, cross_entropy: 3.601651906967163
|
| 725 |
+
[2025-05-04 15:46:16,231][BaseTrainer][INFO] - [Epoch 1/10, Iter 238/3560] 3.541386604309082, cross_entropy: 3.541386604309082
|
| 726 |
+
[2025-05-04 15:46:17,281][BaseTrainer][INFO] - [Epoch 1/10, Iter 239/3560] 3.364543914794922, cross_entropy: 3.364543914794922
|
| 727 |
+
[2025-05-04 15:46:20,548][BaseTrainer][INFO] - [Epoch 1/10, Iter 240/3560] 3.4454920291900635, cross_entropy: 3.4454920291900635
|
| 728 |
+
[2025-05-04 15:46:21,604][BaseTrainer][INFO] - [Epoch 1/10, Iter 241/3560] 3.2512612342834473, cross_entropy: 3.2512612342834473
|
| 729 |
+
[2025-05-04 15:46:25,509][BaseTrainer][INFO] - [Epoch 1/10, Iter 242/3560] 3.982595443725586, cross_entropy: 3.982595443725586
|
| 730 |
+
[2025-05-04 15:46:26,564][BaseTrainer][INFO] - [Epoch 1/10, Iter 243/3560] 3.024195671081543, cross_entropy: 3.024195671081543
|
| 731 |
+
[2025-05-04 15:46:30,189][BaseTrainer][INFO] - [Epoch 1/10, Iter 244/3560] 3.38856840133667, cross_entropy: 3.38856840133667
|
| 732 |
+
[2025-05-04 15:46:31,222][BaseTrainer][INFO] - [Epoch 1/10, Iter 245/3560] 3.58903169631958, cross_entropy: 3.58903169631958
|
| 733 |
+
[2025-05-04 15:46:33,952][BaseTrainer][INFO] - [Epoch 1/10, Iter 246/3560] 3.0575690269470215, cross_entropy: 3.0575690269470215
|
| 734 |
+
[2025-05-04 15:46:35,005][BaseTrainer][INFO] - [Epoch 1/10, Iter 247/3560] 3.803276300430298, cross_entropy: 3.803276300430298
|
| 735 |
+
[2025-05-04 15:46:38,371][BaseTrainer][INFO] - [Epoch 1/10, Iter 248/3560] 4.254518508911133, cross_entropy: 4.254518508911133
|
| 736 |
+
[2025-05-04 15:46:39,430][BaseTrainer][INFO] - [Epoch 1/10, Iter 249/3560] 3.6299855709075928, cross_entropy: 3.6299855709075928
|
| 737 |
+
[2025-05-04 15:46:42,489][BaseTrainer][INFO] - [Epoch 1/10, Iter 250/3560] 4.0379180908203125, cross_entropy: 4.0379180908203125
|
| 738 |
+
[2025-05-04 15:46:43,543][BaseTrainer][INFO] - [Epoch 1/10, Iter 251/3560] 3.455204486846924, cross_entropy: 3.455204486846924
|
| 739 |
+
[2025-05-04 15:46:47,003][BaseTrainer][INFO] - [Epoch 1/10, Iter 252/3560] 3.641664743423462, cross_entropy: 3.641664743423462
|
| 740 |
+
[2025-05-04 15:46:48,054][BaseTrainer][INFO] - [Epoch 1/10, Iter 253/3560] 3.341184616088867, cross_entropy: 3.341184616088867
|
| 741 |
+
[2025-05-04 15:46:50,934][BaseTrainer][INFO] - [Epoch 1/10, Iter 254/3560] 3.268005609512329, cross_entropy: 3.268005609512329
|
| 742 |
+
[2025-05-04 15:46:51,968][BaseTrainer][INFO] - [Epoch 1/10, Iter 255/3560] 2.850250720977783, cross_entropy: 2.850250720977783
|
| 743 |
+
[2025-05-04 15:46:55,426][BaseTrainer][INFO] - [Epoch 1/10, Iter 256/3560] 3.390626907348633, cross_entropy: 3.390626907348633
|
| 744 |
+
[2025-05-04 15:46:56,480][BaseTrainer][INFO] - [Epoch 1/10, Iter 257/3560] 4.221792221069336, cross_entropy: 4.221792221069336
|
| 745 |
+
[2025-05-04 15:46:59,679][BaseTrainer][INFO] - [Epoch 1/10, Iter 258/3560] 3.77260160446167, cross_entropy: 3.77260160446167
|
| 746 |
+
[2025-05-04 15:47:00,734][BaseTrainer][INFO] - [Epoch 1/10, Iter 259/3560] 3.8018369674682617, cross_entropy: 3.8018369674682617
|
| 747 |
+
[2025-05-04 15:47:04,046][BaseTrainer][INFO] - [Epoch 1/10, Iter 260/3560] 3.230081558227539, cross_entropy: 3.230081558227539
|
| 748 |
+
[2025-05-04 15:47:05,102][BaseTrainer][INFO] - [Epoch 1/10, Iter 261/3560] 2.9646286964416504, cross_entropy: 2.9646286964416504
|
| 749 |
+
[2025-05-04 15:47:07,788][BaseTrainer][INFO] - [Epoch 1/10, Iter 262/3560] 4.133569717407227, cross_entropy: 4.133569717407227
|
| 750 |
+
[2025-05-04 15:47:08,830][BaseTrainer][INFO] - [Epoch 1/10, Iter 263/3560] 3.628683090209961, cross_entropy: 3.628683090209961
|
| 751 |
+
[2025-05-04 15:47:12,809][BaseTrainer][INFO] - [Epoch 1/10, Iter 264/3560] 3.167900562286377, cross_entropy: 3.167900562286377
|
| 752 |
+
[2025-05-04 15:47:13,868][BaseTrainer][INFO] - [Epoch 1/10, Iter 265/3560] 3.3190231323242188, cross_entropy: 3.3190231323242188
|
| 753 |
+
[2025-05-04 15:47:16,945][BaseTrainer][INFO] - [Epoch 1/10, Iter 266/3560] 3.4320666790008545, cross_entropy: 3.4320666790008545
|
| 754 |
+
[2025-05-04 15:47:18,014][BaseTrainer][INFO] - [Epoch 1/10, Iter 267/3560] 3.4276881217956543, cross_entropy: 3.4276881217956543
|
| 755 |
+
[2025-05-04 15:47:21,062][BaseTrainer][INFO] - [Epoch 1/10, Iter 268/3560] 2.7244088649749756, cross_entropy: 2.7244088649749756
|
| 756 |
+
[2025-05-04 15:47:22,108][BaseTrainer][INFO] - [Epoch 1/10, Iter 269/3560] 3.503838062286377, cross_entropy: 3.503838062286377
|
| 757 |
+
[2025-05-04 15:47:24,722][BaseTrainer][INFO] - [Epoch 1/10, Iter 270/3560] 4.146647930145264, cross_entropy: 4.146647930145264
|
| 758 |
+
[2025-05-04 15:47:25,778][BaseTrainer][INFO] - [Epoch 1/10, Iter 271/3560] 4.013218402862549, cross_entropy: 4.013218402862549
|
| 759 |
+
[2025-05-04 15:47:28,913][BaseTrainer][INFO] - [Epoch 1/10, Iter 272/3560] 3.860719919204712, cross_entropy: 3.860719919204712
|
| 760 |
+
[2025-05-04 15:47:29,976][BaseTrainer][INFO] - [Epoch 1/10, Iter 273/3560] 3.8546671867370605, cross_entropy: 3.8546671867370605
|
| 761 |
+
[2025-05-04 15:47:33,011][BaseTrainer][INFO] - [Epoch 1/10, Iter 274/3560] 3.229749917984009, cross_entropy: 3.229749917984009
|
| 762 |
+
[2025-05-04 15:47:34,073][BaseTrainer][INFO] - [Epoch 1/10, Iter 275/3560] 4.244211196899414, cross_entropy: 4.244211196899414
|
| 763 |
+
[2025-05-04 15:47:36,797][BaseTrainer][INFO] - [Epoch 1/10, Iter 276/3560] 3.8821959495544434, cross_entropy: 3.8821959495544434
|
| 764 |
+
[2025-05-04 15:47:37,828][BaseTrainer][INFO] - [Epoch 1/10, Iter 277/3560] 3.31406307220459, cross_entropy: 3.31406307220459
|
| 765 |
+
[2025-05-04 15:47:41,369][BaseTrainer][INFO] - [Epoch 1/10, Iter 278/3560] 3.015435218811035, cross_entropy: 3.015435218811035
|
| 766 |
+
[2025-05-04 15:47:42,419][BaseTrainer][INFO] - [Epoch 1/10, Iter 279/3560] 4.190752029418945, cross_entropy: 4.190752029418945
|
| 767 |
+
[2025-05-04 15:47:45,365][BaseTrainer][INFO] - [Epoch 1/10, Iter 280/3560] 3.2984061241149902, cross_entropy: 3.2984061241149902
|
| 768 |
+
[2025-05-04 15:47:46,416][BaseTrainer][INFO] - [Epoch 1/10, Iter 281/3560] 3.5504894256591797, cross_entropy: 3.5504894256591797
|
| 769 |
+
[2025-05-04 15:47:49,376][BaseTrainer][INFO] - [Epoch 1/10, Iter 282/3560] 3.105098009109497, cross_entropy: 3.105098009109497
|
| 770 |
+
[2025-05-04 15:47:50,428][BaseTrainer][INFO] - [Epoch 1/10, Iter 283/3560] 3.6767663955688477, cross_entropy: 3.6767663955688477
|
| 771 |
+
[2025-05-04 15:47:53,486][BaseTrainer][INFO] - [Epoch 1/10, Iter 284/3560] 3.078446865081787, cross_entropy: 3.078446865081787
|
| 772 |
+
[2025-05-04 15:47:54,536][BaseTrainer][INFO] - [Epoch 1/10, Iter 285/3560] 2.937081813812256, cross_entropy: 2.937081813812256
|
| 773 |
+
[2025-05-04 15:47:57,459][BaseTrainer][INFO] - [Epoch 1/10, Iter 286/3560] 3.384705066680908, cross_entropy: 3.384705066680908
|
| 774 |
+
[2025-05-04 15:47:58,496][BaseTrainer][INFO] - [Epoch 1/10, Iter 287/3560] 3.093351125717163, cross_entropy: 3.093351125717163
|
| 775 |
+
[2025-05-04 15:48:01,395][BaseTrainer][INFO] - [Epoch 1/10, Iter 288/3560] 2.6946282386779785, cross_entropy: 2.6946282386779785
|
| 776 |
+
[2025-05-04 15:48:02,449][BaseTrainer][INFO] - [Epoch 1/10, Iter 289/3560] 3.7547240257263184, cross_entropy: 3.7547240257263184
|
| 777 |
+
[2025-05-04 15:48:06,235][BaseTrainer][INFO] - [Epoch 1/10, Iter 290/3560] 2.8298704624176025, cross_entropy: 2.8298704624176025
|
| 778 |
+
[2025-05-04 15:48:07,269][BaseTrainer][INFO] - [Epoch 1/10, Iter 291/3560] 3.0255863666534424, cross_entropy: 3.0255863666534424
|
| 779 |
+
[2025-05-04 15:48:10,384][BaseTrainer][INFO] - [Epoch 1/10, Iter 292/3560] 3.471223831176758, cross_entropy: 3.471223831176758
|
| 780 |
+
[2025-05-04 15:48:11,439][BaseTrainer][INFO] - [Epoch 1/10, Iter 293/3560] 4.027524948120117, cross_entropy: 4.027524948120117
|
| 781 |
+
[2025-05-04 15:48:13,751][BaseTrainer][INFO] - [Epoch 1/10, Iter 294/3560] 2.9611196517944336, cross_entropy: 2.9611196517944336
|
| 782 |
+
[2025-05-04 15:48:14,788][BaseTrainer][INFO] - [Epoch 1/10, Iter 295/3560] 2.9723658561706543, cross_entropy: 2.9723658561706543
|
| 783 |
+
[2025-05-04 15:48:18,471][BaseTrainer][INFO] - [Epoch 1/10, Iter 296/3560] 3.3414180278778076, cross_entropy: 3.3414180278778076
|
| 784 |
+
[2025-05-04 15:48:19,576][BaseTrainer][INFO] - [Epoch 1/10, Iter 297/3560] 4.208791732788086, cross_entropy: 4.208791732788086
|
| 785 |
+
[2025-05-04 15:48:22,848][BaseTrainer][INFO] - [Epoch 1/10, Iter 298/3560] 2.780437707901001, cross_entropy: 2.780437707901001
|
| 786 |
+
[2025-05-04 15:48:23,886][BaseTrainer][INFO] - [Epoch 1/10, Iter 299/3560] 3.171988010406494, cross_entropy: 3.171988010406494
|
| 787 |
+
[2025-05-04 15:48:26,473][BaseTrainer][INFO] - [Epoch 1/10, Iter 300/3560] 2.800706386566162, cross_entropy: 2.800706386566162
|
| 788 |
+
[2025-05-04 15:48:27,553][BaseTrainer][INFO] - [Epoch 1/10, Iter 301/3560] 2.5326716899871826, cross_entropy: 2.5326716899871826
|
| 789 |
+
[2025-05-04 15:48:31,087][BaseTrainer][INFO] - [Epoch 1/10, Iter 302/3560] 3.0759329795837402, cross_entropy: 3.0759329795837402
|
| 790 |
+
[2025-05-04 15:48:32,170][BaseTrainer][INFO] - [Epoch 1/10, Iter 303/3560] 3.110572576522827, cross_entropy: 3.110572576522827
|
| 791 |
+
[2025-05-04 15:48:35,517][BaseTrainer][INFO] - [Epoch 1/10, Iter 304/3560] 2.567490816116333, cross_entropy: 2.567490816116333
|
| 792 |
+
[2025-05-04 15:48:36,725][BaseTrainer][INFO] - [Epoch 1/10, Iter 305/3560] 3.407499074935913, cross_entropy: 3.407499074935913
|
| 793 |
+
[2025-05-04 15:48:39,501][BaseTrainer][INFO] - [Epoch 1/10, Iter 306/3560] 2.7290828227996826, cross_entropy: 2.7290828227996826
|
| 794 |
+
[2025-05-04 15:48:41,177][BaseTrainer][INFO] - [Epoch 1/10, Iter 307/3560] 3.153337001800537, cross_entropy: 3.153337001800537
|
| 795 |
+
[2025-05-04 15:48:43,159][BaseTrainer][INFO] - [Epoch 1/10, Iter 308/3560] 3.1342296600341797, cross_entropy: 3.1342296600341797
|
| 796 |
+
[2025-05-04 15:48:45,243][BaseTrainer][INFO] - [Epoch 1/10, Iter 309/3560] 2.606759548187256, cross_entropy: 2.606759548187256
|
| 797 |
+
[2025-05-04 15:48:46,543][BaseTrainer][INFO] - [Epoch 1/10, Iter 310/3560] 2.7439932823181152, cross_entropy: 2.7439932823181152
|
| 798 |
+
[2025-05-04 15:48:49,531][BaseTrainer][INFO] - [Epoch 1/10, Iter 311/3560] 3.740546226501465, cross_entropy: 3.740546226501465
|
| 799 |
+
[2025-05-04 15:48:50,634][BaseTrainer][INFO] - [Epoch 1/10, Iter 312/3560] 3.289177417755127, cross_entropy: 3.289177417755127
|
| 800 |
+
[2025-05-04 15:48:53,534][BaseTrainer][INFO] - [Epoch 1/10, Iter 313/3560] 3.084707736968994, cross_entropy: 3.084707736968994
|
| 801 |
+
[2025-05-04 15:48:54,815][BaseTrainer][INFO] - [Epoch 1/10, Iter 314/3560] 3.2798752784729004, cross_entropy: 3.2798752784729004
|
| 802 |
+
[2025-05-04 15:48:57,330][BaseTrainer][INFO] - [Epoch 1/10, Iter 315/3560] 2.9585819244384766, cross_entropy: 2.9585819244384766
|
| 803 |
+
[2025-05-04 15:48:59,217][BaseTrainer][INFO] - [Epoch 1/10, Iter 316/3560] 3.504563808441162, cross_entropy: 3.504563808441162
|
| 804 |
+
[2025-05-04 15:49:01,274][BaseTrainer][INFO] - [Epoch 1/10, Iter 317/3560] 2.982103109359741, cross_entropy: 2.982103109359741
|
| 805 |
+
[2025-05-04 15:49:03,260][BaseTrainer][INFO] - [Epoch 1/10, Iter 318/3560] 3.3061461448669434, cross_entropy: 3.3061461448669434
|
| 806 |
+
[2025-05-04 15:49:05,403][BaseTrainer][INFO] - [Epoch 1/10, Iter 319/3560] 3.3542447090148926, cross_entropy: 3.3542447090148926
|
| 807 |
+
[2025-05-04 15:49:07,758][BaseTrainer][INFO] - [Epoch 1/10, Iter 320/3560] 2.830873966217041, cross_entropy: 2.830873966217041
|
| 808 |
+
[2025-05-04 15:49:09,518][BaseTrainer][INFO] - [Epoch 1/10, Iter 321/3560] 3.878793954849243, cross_entropy: 3.878793954849243
|
| 809 |
+
[2025-05-04 15:49:11,693][BaseTrainer][INFO] - [Epoch 1/10, Iter 322/3560] 3.0356221199035645, cross_entropy: 3.0356221199035645
|
| 810 |
+
[2025-05-04 15:49:13,799][BaseTrainer][INFO] - [Epoch 1/10, Iter 323/3560] 2.5573325157165527, cross_entropy: 2.5573325157165527
|
| 811 |
+
[2025-05-04 15:49:15,908][BaseTrainer][INFO] - [Epoch 1/10, Iter 324/3560] 3.7650721073150635, cross_entropy: 3.7650721073150635
|
| 812 |
+
[2025-05-04 15:49:17,897][BaseTrainer][INFO] - [Epoch 1/10, Iter 325/3560] 3.0595507621765137, cross_entropy: 3.0595507621765137
|
| 813 |
+
[2025-05-04 15:49:20,665][BaseTrainer][INFO] - [Epoch 1/10, Iter 326/3560] 2.892014741897583, cross_entropy: 2.892014741897583
|
| 814 |
+
[2025-05-04 15:49:22,147][BaseTrainer][INFO] - [Epoch 1/10, Iter 327/3560] 2.8922882080078125, cross_entropy: 2.8922882080078125
|
| 815 |
+
[2025-05-04 15:49:24,585][BaseTrainer][INFO] - [Epoch 1/10, Iter 328/3560] 3.8311381340026855, cross_entropy: 3.8311381340026855
|
| 816 |
+
[2025-05-04 15:49:26,335][BaseTrainer][INFO] - [Epoch 1/10, Iter 329/3560] 3.166689157485962, cross_entropy: 3.166689157485962
|
| 817 |
+
[2025-05-04 15:49:28,684][BaseTrainer][INFO] - [Epoch 1/10, Iter 330/3560] 3.9570493698120117, cross_entropy: 3.9570493698120117
|
| 818 |
+
[2025-05-04 15:49:29,781][BaseTrainer][INFO] - [Epoch 1/10, Iter 331/3560] 3.397995948791504, cross_entropy: 3.397995948791504
|
| 819 |
+
[2025-05-04 15:49:33,002][BaseTrainer][INFO] - [Epoch 1/10, Iter 332/3560] 2.5950450897216797, cross_entropy: 2.5950450897216797
|
| 820 |
+
[2025-05-04 15:49:34,073][BaseTrainer][INFO] - [Epoch 1/10, Iter 333/3560] 3.4235217571258545, cross_entropy: 3.4235217571258545
|
| 821 |
+
[2025-05-04 15:49:36,793][BaseTrainer][INFO] - [Epoch 1/10, Iter 334/3560] 2.8502418994903564, cross_entropy: 2.8502418994903564
|
| 822 |
+
[2025-05-04 15:49:37,986][BaseTrainer][INFO] - [Epoch 1/10, Iter 335/3560] 3.467989921569824, cross_entropy: 3.467989921569824
|
| 823 |
+
[2025-05-04 15:49:40,788][BaseTrainer][INFO] - [Epoch 1/10, Iter 336/3560] 2.778249740600586, cross_entropy: 2.778249740600586
|
| 824 |
+
[2025-05-04 15:49:41,886][BaseTrainer][INFO] - [Epoch 1/10, Iter 337/3560] 2.4447736740112305, cross_entropy: 2.4447736740112305
|
| 825 |
+
[2025-05-04 15:49:45,446][BaseTrainer][INFO] - [Epoch 1/10, Iter 338/3560] 2.629225254058838, cross_entropy: 2.629225254058838
|
| 826 |
+
[2025-05-04 15:49:46,480][BaseTrainer][INFO] - [Epoch 1/10, Iter 339/3560] 2.776583194732666, cross_entropy: 2.776583194732666
|
| 827 |
+
[2025-05-04 15:49:49,778][BaseTrainer][INFO] - [Epoch 1/10, Iter 340/3560] 2.8193931579589844, cross_entropy: 2.8193931579589844
|
| 828 |
+
[2025-05-04 15:49:50,837][BaseTrainer][INFO] - [Epoch 1/10, Iter 341/3560] 3.5558292865753174, cross_entropy: 3.5558292865753174
|
| 829 |
+
[2025-05-04 15:49:54,146][BaseTrainer][INFO] - [Epoch 1/10, Iter 342/3560] 3.3698346614837646, cross_entropy: 3.3698346614837646
|
| 830 |
+
[2025-05-04 15:49:55,200][BaseTrainer][INFO] - [Epoch 1/10, Iter 343/3560] 3.103781223297119, cross_entropy: 3.103781223297119
|
| 831 |
+
[2025-05-04 15:49:58,185][BaseTrainer][INFO] - [Epoch 1/10, Iter 344/3560] 2.4396743774414062, cross_entropy: 2.4396743774414062
|
| 832 |
+
[2025-05-04 15:49:59,219][BaseTrainer][INFO] - [Epoch 1/10, Iter 345/3560] 3.6696176528930664, cross_entropy: 3.6696176528930664
|
| 833 |
+
[2025-05-04 15:50:02,898][BaseTrainer][INFO] - [Epoch 1/10, Iter 346/3560] 3.1565327644348145, cross_entropy: 3.1565327644348145
|
| 834 |
+
[2025-05-04 15:50:03,953][BaseTrainer][INFO] - [Epoch 1/10, Iter 347/3560] 2.7521986961364746, cross_entropy: 2.7521986961364746
|
| 835 |
+
[2025-05-04 15:50:07,638][BaseTrainer][INFO] - [Epoch 1/10, Iter 348/3560] 2.695011854171753, cross_entropy: 2.695011854171753
|
| 836 |
+
[2025-05-04 15:50:08,676][BaseTrainer][INFO] - [Epoch 1/10, Iter 349/3560] 2.9952139854431152, cross_entropy: 2.9952139854431152
|
| 837 |
+
[2025-05-04 15:50:11,707][BaseTrainer][INFO] - [Epoch 1/10, Iter 350/3560] 2.670276641845703, cross_entropy: 2.670276641845703
|
| 838 |
+
[2025-05-04 15:50:12,764][BaseTrainer][INFO] - [Epoch 1/10, Iter 351/3560] 2.838888168334961, cross_entropy: 2.838888168334961
|
| 839 |
+
[2025-05-04 15:50:15,545][BaseTrainer][INFO] - [Epoch 1/10, Iter 352/3560] 2.5220344066619873, cross_entropy: 2.5220344066619873
|
| 840 |
+
[2025-05-04 15:50:16,593][BaseTrainer][INFO] - [Epoch 1/10, Iter 353/3560] 2.882241725921631, cross_entropy: 2.882241725921631
|
| 841 |
+
[2025-05-04 15:50:19,281][BaseTrainer][INFO] - [Epoch 1/10, Iter 354/3560] 2.993210792541504, cross_entropy: 2.993210792541504
|
| 842 |
+
[2025-05-04 15:50:20,313][BaseTrainer][INFO] - [Epoch 1/10, Iter 355/3560] 3.0053420066833496, cross_entropy: 3.0053420066833496
|
| 843 |
+
[2025-05-04 15:50:33,914][BaseTrainer][INFO] - [Epoch 1/10, Iter 356/3560] 3.3669323921203613, cross_entropy: 3.3669323921203613
|
| 844 |
+
[2025-05-04 15:53:41,940][BaseTrainer][INFO] - [Epoch 1/10] (train) 3.984675884246826, cross_entropy: 3.984675884246826
|
| 845 |
+
[2025-05-04 15:53:41,940][BaseTrainer][INFO] - [Epoch 1/10] (validation) 2.4288501739501953, cross_entropy: 2.4288501739501953
|
| 846 |
+
[2025-05-04 15:53:41,941][BaseTrainer][INFO] - [Epoch 1/10] (metrics) roc_auc: 0.9189022779464722
|
| 847 |
+
[2025-05-04 15:53:42,436][BaseTrainer][INFO] - Save model: ./exp/20250504-153714/model/best_epoch.pth.
|
| 848 |
+
[2025-05-04 15:53:42,919][BaseTrainer][INFO] - Save model: ./exp/20250504-153714/model/epoch1.pth.
|
| 849 |
+
[2025-05-04 15:53:43,395][BaseTrainer][INFO] - Save model: ./exp/20250504-153714/model/last.pth.
|
| 850 |
+
[2025-05-04 15:53:48,172][BaseTrainer][INFO] - [Epoch 2/10, Iter 357/3560] 2.757335662841797, cross_entropy: 2.757335662841797
|
| 851 |
+
[2025-05-04 15:53:49,263][BaseTrainer][INFO] - [Epoch 2/10, Iter 358/3560] 3.5638482570648193, cross_entropy: 3.5638482570648193
|
| 852 |
+
[2025-05-04 15:53:52,078][BaseTrainer][INFO] - [Epoch 2/10, Iter 359/3560] 3.5854949951171875, cross_entropy: 3.5854949951171875
|
| 853 |
+
[2025-05-04 15:53:53,266][BaseTrainer][INFO] - [Epoch 2/10, Iter 360/3560] 2.808537006378174, cross_entropy: 2.808537006378174
|