mrc_uit_squadv2 / configs /train_vi_electra_base.yaml
annt's picture
Add application file
daeb223
raw
history blame contribute delete
No virus
1.3 kB
RetroDataModelArguments:
# DataArguments
max_seq_length: 512
max_answer_length: 30
doc_stride: 128
return_token_type_ids: True
pad_to_max_length: True
preprocessing_num_workers: 5
overwrite_cache: False
version_2_with_negative: True
null_score_diff_threshold: 0.0
rear_threshold: 0.0
n_best_size: 20
use_choice_logits: False
start_n_top: -1
end_n_top: -1
beta1: 1
beta2: 1
best_cof: 1
# SketchModelArguments
sketch_model_name: NlpHUST/electra-base-vn
sketch_architectures: ElectraForSequenceClassification
# IntensiveModelArguments
intensive_model_name: NlpHUST/electra-base-vn
intensive_architectures: ElectraForQuestionAnsweringAVPool
TrainingArguments:
report_to: wandb
run_name: squadv2-electra-large-sketch,squadv2-electra-large-intensive
output_dir: outputs
overwrite_output_dir: False
learning_rate: 2e-5
evaluation_strategy: epoch
save_strategy: epoch
per_device_train_batch_size: 12
per_device_eval_batch_size: 12
num_train_epochs: 10.0
# save_steps: 5000
save_total_limit: 2
fp16: True
warmup_ratio: 0.1
weight_decay: 0.01
load_best_model_at_end: True
metric_for_best_model: f1,exact
logging_dir: logs