pietrolesci's picture
Upload folder using huggingface_hub
7806f40 verified
[2024-04-23 15:11:06,594][hydra][INFO] -
experiment_group: training
run_name: roberta-base_2024-04-23T15-11-06
seed: 42
model:
name: roberta-base
revision: null
seed: 42
base_model: roberta-base
estimator:
accelerator: gpu
precision: bf16-true
deterministic: true
tf32_mode: high
convert_to_bettertransformer: false
fit:
max_epochs: 20
min_epochs: null
optimizer_kwargs:
name: adamw
lr: 3.0e-05
init_kwargs:
fused: true
scheduler_kwargs:
name: constant_schedule_with_warmup
num_warmup_steps: 2000
log_interval: 100
enable_progress_bar: true
limit_train_batches: null
data:
batch_size: 32
eval_batch_size: 128
shuffle: true
replacement: false
data_seed: 42
drop_last: false
num_workers: 8
pin_memory: true
persistent_workers: false
multiprocessing_context: null
max_length: 512
root_path: /home/pl487/coreset-project
data_path: /home/pl487/coreset-project/data/processed
dataset: mnli
dataset_split: train
evaluation: null
loggers:
tensorboard:
_target_: energizer.loggers.TensorBoardLogger
root_dir: ./
name: tb_logs
version: null
callbacks:
timer:
_target_: energizer.active_learning.callbacks.Timer
lr_monitor:
_target_: energizer.callbacks.lr_monitor.LearningRateMonitor
model_checkpoint:
_target_: energizer.callbacks.model_checkpoint.ModelCheckpoint
dirpath: .checkpoints
stage: train
frequency: 1:epoch
user:
id: pl487
======================================================================
[2024-04-23 15:11:06,595][hydra][INFO] - Seed enabled: 42
[2024-04-23 15:11:06,963][hydra][INFO] - Label distribution:
{<RunningStage.TRAIN: 'train'>: {'0-(entailment)': 130899, '1-(neutral)': 130900, '2-(contradiction)': 130903}}
[2024-04-23 15:11:19,109][hydra][INFO] - Loggers: [<energizer.loggers.tensorboard.TensorBoardLogger object at 0x7f86f05beb00>]
[2024-04-23 15:11:19,110][hydra][INFO] - Callbacks: [<energizer.active_learning.callbacks.Timer object at 0x7f86deae60b0>, <energizer.callbacks.lr_monitor.LearningRateMonitor object at 0x7f86deae6110>, <energizer.callbacks.model_checkpoint.ModelCheckpoint object at 0x7f86deae6620>]
[2024-04-23 15:11:19,113][hydra][INFO] - Model summary:
Total num params: 124.6M
Of which trainable: 124.6M
With a memory footprint of 0.25GB
Total memory allocated 0.77GB
[2024-04-23 15:11:19,754][hydra][INFO] - Dataloading params:
SequenceClassificationDataloaderArgs(batch_size=32, eval_batch_size=128, num_workers=8, pin_memory=True, drop_last=False, persistent_workers=False, shuffle=True, replacement=False, data_seed=42, multiprocessing_context=None, max_length=512)
[2024-04-23 15:11:19,760][hydra][INFO] - Batch:
{<InputKeys.INPUT_IDS: 'input_ids'>: tensor([[ 0, 3056, 37463, 14, 18, 761, 9, 4678, 939, 1266,
51, 214, 190, 3406, 24, 7, 7, 147, 122, 37463,
14, 51, 26112, 15, 1012, 47, 216, 114, 110, 114,
47, 37463, 47, 216, 33, 626, 42, 50, 114, 47,
240, 42, 37463, 37463, 52, 581, 14811, 13, 47, 8,
47, 218, 75, 33, 7, 582, 201, 3867, 47, 53,
172, 99, 51, 218, 75, 1137, 47, 16, 14, 114,
47, 114, 51, 339, 47, 492, 106, 23, 513, 10,
371, 9, 5, 9, 5, 631, 14, 51, 339, 98,
1437, 939, 218, 75, 216, 24, 16, 37463, 24, 18,
562, 7, 28, 55, 265, 122, 1195, 87, 37463, 888,
37463, 4098, 19, 5, 1846, 87, 19, 37463, 7252, 5,
37463, 8653, 51, 5, 5, 3969, 32, 95, 11, 24,
13, 5, 418, 1437, 939, 437, 939, 437, 7013, 939,
216, 939, 939, 2854, 19, 47, 939, 206, 47, 214,
588, 47, 214, 182, 235, 14, 5, 3770, 197, 939,
206, 51, 197, 33, 41, 3871, 1280, 9, 47, 216,
2085, 51, 64, 33, 10, 367, 53, 939, 206, 144,
9, 106, 197, 28, 45, 37463, 3969, 11, 5, 754,
8, 14, 18, 169, 169, 51, 348, 5335, 88, 2302,
24, 18, 142, 9, 5, 5, 488, 383, 47, 216,
5, 37482, 8, 960, 53, 37463, 53, 117, 939, 216,
52, 24, 7252, 52, 214, 11, 4788, 8, 37463, 52,
33, 5, 276, 631, 22002, 154, 8, 8, 37463, 51,
32, 6901, 106, 66, 939, 1266, 95, 5, 1675, 2878,
3645, 631, 51, 905, 106, 66, 142, 9, 51, 218,
75, 33, 143, 317, 7, 489, 7, 342, 106, 98,
4909, 14, 115, 47, 216, 37463, 3867, 37463, 57, 10,
538, 2970, 53, 37463, 37463, 190, 5, 181, 6502, 352,
410, 2682, 939, 1266, 5, 5, 1669, 14, 13585, 452,
11, 730, 16, 5, 1802, 8, 14051, 14, 189, 28,
549, 24, 18, 10, 6279, 50, 2196, 50, 3046, 1493,
7252, 51, 32, 5, 1980, 14, 32, 164, 7, 582,
8, 51, 32, 5, 65, 14, 32, 164, 7, 6297,
8, 5, 97, 621, 47, 216, 114, 51, 582, 114,
51, 114, 51, 37463, 2237, 106, 37463, 8, 37463, 172,
5, 47, 216, 86, 5, 488, 1239, 81, 37463, 457,
5, 86, 37463, 51, 1169, 905, 106, 213, 50, 51,
120, 160, 19, 10, 10, 3645, 142, 51, 348, 56,
10, 2470, 14, 47, 216, 2653, 14, 14, 51, 3559,
75, 70, 561, 77, 51, 222, 24, 2, 2, 100,
206, 14, 89, 197, 28, 41, 3871, 8985, 9, 14218,
11, 84, 3770, 4, 2]]), <InputKeys.ATT_MASK: 'attention_mask'>: tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]]), <InputKeys.LABELS: 'labels'>: tensor([0]), <InputKeys.ON_CPU: 'on_cpu'>: {<SpecialKeys.ID: 'uid'>: [221950]}}
[2024-04-23 20:57:35,566][hydra][INFO] - Training complete