|=>> 12/01 [03:20:32] - mistral - INFO :: Starting Run: shuffle_local10_en_EN_randinit_seed53...
|=>> 12/01 [03:20:32] - mistral - INFO :: Setting Random Seed to 53!
|=>> 12/01 [03:20:32] - mistral - INFO :: Building Tokenize and Initializing `gpt2-small` via AutoModel/AutoConfig...
|=>> 12/01 [03:20:32] - mistral - INFO :: Using Configs For Model From: /local/xiulyang/mission-impossible-language-models/mistral/conf/models/gpt2-small-30000.json ...
|=>> 12/01 [03:20:32] - mistral.models.auto - INFO :: Building Hugging Face GPT2Config from provided configs: {'activation_function': 'gelu_new', 'architectures': ['GPT2LMHeadModel'], 'attn_pdrop': 0.1, 'embd_pdrop': 0.1, 'eos_token_id': 0, 'bos_token_id': 0, 'initializer_range': 0.02, 'layer_norm_epsilon': 1e-05, 'model_type': 'gpt2', 'n_ctx': 1024, 'n_embd': 768, 'n_head': 12, 'n_inner': None, 'n_layer': 12, 'n_positions': 1024, 'reorder_and_upcast_attn': True, 'resid_pdrop': 0.1, 'scale_attn_by_inverse_layer_idx': True, 'scale_attn_weights': True, 'summary_activation': None, 'summary_first_dropout': 0.2, 'summary_proj_to_labels': True, 'summary_type': 'cls_index', 'summary_use_proj': True, 'task_specific_params': {'text-generation': {'do_sample': True, 'max_length': 1024}}, 'torch_dtype': 'float32', 'transformers_version': '4.35.2', 'use_cache': False, 'vocab_size': 30000} ...
|=>> 12/01 [03:20:32] - mistral.models.auto - INFO :: Fetching Hugging Face [Fast] AutoTokenizer for Model: `gpt2`...
|=>> 12/01 [03:20:32] - mistral.models.auto - INFO :: Using a Pretokenized Dataset
|=>> 12/01 [03:20:32] - mistral.models.auto - INFO :: Initializing Custom GPT-2 Model from Configuration: `gpt2`...
|=>> 12/01 [03:20:36] - mistral - INFO :: Setting Training Arguments from Quinfig...
|=>> 12/01 [03:20:36] - mistral.args.training - INFO :: Setting Gradient Accumulation Steps = `64` [BSZ: 512 World Size: 1 Device BSZ: 8]
|=>> 12/01 [03:20:36] - mistral - INFO :: Downloading and Preprocessing Dataset `/local/xiulyang/mission-impossible-language-models/training/babylm_dataset.py`...
|=>> 12/01 [03:20:36] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Generating examples from = /local/xiulyang/mission-impossible-language-models/data/multilingual/multilingual_data_perturbed/shuffle_local10_en/train
|=>> 12/01 [03:20:36] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Total sentences: 1112000
|=>> 12/01 [03:20:36] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Loading pre-tokenized data
|=>> 12/01 [03:20:41] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Concatenating tokenized data using EOS token
|=>> 12/01 [03:20:42] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Chunking tokens into sublists of 1024
|=>> 12/01 [03:20:43] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Writing dataset as space-separated sequences of tokens
|=>> 12/01 [03:20:46] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Generating examples from = /local/xiulyang/mission-impossible-language-models/data/multilingual/multilingual_data_perturbed/shuffle_local10_en/dev
|=>> 12/01 [03:20:46] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Total sentences: 5431
|=>> 12/01 [03:20:46] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Loading pre-tokenized data
|=>> 12/01 [03:20:46] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Concatenating tokenized data using EOS token
|=>> 12/01 [03:20:46] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Chunking tokens into sublists of 1024
|=>> 12/01 [03:20:46] - datasets_modules.datasets.babylm_dataset.635a7d5fded93747e7bbf7059010da97f83d268db9e59629b0409af3d5c2cf6f.babylm_dataset - INFO :: Writing dataset as space-separated sequences of tokens
|=>> 12/01 [03:20:46] - mistral.corpora.auto - INFO :: Building Tokenized Indexed Dataset for {dataset_id}/{dataset_name}...
|=>> 12/01 [03:20:46] - mistral.corpora.auto - INFO :: Building Indexed Dataset for train
|=>> 12/01 [03:21:15] - mistral.corpora.auto - INFO :: Building Indexed Dataset for validation
|=>> 12/01 [03:21:15] - mistral - INFO :: Initializing Model Trainer...
|=>> 12/01 [03:21:15] - mistral - INFO :: Training Arguments: TrainingArguments(
_n_gpu=1,
adafactor=False,
adam_beta1=0.9,
adam_beta2=0.999,
adam_epsilon=1e-08,
bf16=False,
bf16_full_eval=False,
data_seed=53,
dataloader_drop_last=False,
dataloader_num_workers=0,
dataloader_pin_memory=True,
ddp_bucket_cap_mb=None,
ddp_find_unused_parameters=None,
debug=[],
deepspeed=None,
disable_tqdm=False,
do_eval=True,
do_predict=False,
do_train=True,
eval_accumulation_steps=None,
eval_delay=0,
eval_steps=1000,
evaluation_strategy=IntervalStrategy.STEPS,
fp16=True,
fp16_backend=auto,
fp16_full_eval=False,
fp16_opt_level=O1,
gradient_accumulation_steps=64,
gradient_checkpointing=False,
greater_is_better=None,
group_by_length=False,
half_precision_backend=auto,
hub_model_id=None,
hub_strategy=HubStrategy.EVERY_SAVE,
hub_token=<HUB_TOKEN>,
ignore_data_skip=False,
label_names=None,
label_smoothing_factor=0.0,
learning_rate=0.0006,
length_column_name=length,
load_best_model_at_end=False,
local_rank=-1,
log_level=-1,
log_level_replica=-1,
log_on_each_node=True,
logging_dir=logs,
logging_first_step=True,
logging_nan_inf_filter=True,
logging_steps=50,
logging_strategy=IntervalStrategy.STEPS,
lr_scheduler_type=SchedulerType.LINEAR,
max_grad_norm=1.0,
max_steps=1200,
metric_for_best_model=None,
mp_parameters=,
no_cuda=False,
num_train_epochs=3.0,
optim=OptimizerNames.ADAMW_HF,
output_dir=//local/xiulyang/babylm_models/shuffle_local10_en_EN_randinit/babylm_shuffle_local10_en_EN_randinit_seed53/runs/shuffle_local10_en_EN_randinit_seed53,
overwrite_output_dir=False,
past_index=-1,
per_device_eval_batch_size=16,
per_device_train_batch_size=8,
prediction_loss_only=True,
push_to_hub=False,
push_to_hub_model_id=None,
push_to_hub_organization=None,
push_to_hub_token=<PUSH_TO_HUB_TOKEN>,
remove_unused_columns=True,
report_to=[],
resume_from_checkpoint=None,
run_name=shuffle_local10_en_EN_randinit_seed53,
save_on_each_node=False,
save_steps=1000,
save_strategy=IntervalStrategy.STEPS,
save_total_limit=None,
seed=53,
sharded_ddp=[],
skip_memory_metrics=True,
tf32=None,
tpu_metrics_debug=False,
tpu_num_cores=None,
use_legacy_prediction_loop=False,
warmup_ratio=0.0,
warmup_steps=120,
weight_decay=0.1,
xpu_backend=None,
)
|=>> 12/01 [03:21:16] - mistral.core.callbacks - INFO :: Setting W&B Project: xiulin-yang-compling
|=>> 12/01 [03:21:19] - mistral - INFO :: Training...
|=>> 12/01 [03:21:19] - mistral.core.callbacks - INFO :: Automatic Weights & Biases logging enabled, to disable set os.environ["WANDB_DISABLED"] = "true"
|=>> 12/01 [12:01:45] - mistral - INFO :: ...and that's all folks!
|=>> 12/01 [12:01:45] - mistral - INFO :: Running final evaluation...