default.yaml

defaults:
  - cluster: example
  - dataset: flores200.full.v4.4
  - model_type: dense
  - _self_

fairseq_root: ???
output_dir: ???
log_dir: null

train_prefix: "dense"
seed: 2
arch: "transformer_24_24"
max_updates: 100000
max_update_str: null
resume_finished: false
synchronize_checkpoints_before_copy: false
validate_interval_updates: 20000
keep_interval_updates: 10
symlink_best_and_last_checkpoints: false
save_interval_updates: 20000
save_interval: 1
best_checkpoint_metric: "loss"
encoder_langtok: "src"
ddp_backend: "fully_sharded"
fp16: true
lr: 0.001
warmup: 8000
max_tokens: 4096
update_freq: 2
num_nodes: 16
num_gpus_per_node: 8
temp: 1
dropout: 0
module_name: "examples.nllb.modeling.sweep.sweep_mmt"
num_trials: 1
max_time_mins: 4320
mem: 0
moe_eval_cap: 1.0
checkpoint_activations: false
zero2: false
ssl_task: null
dae_mask: 0.5
train_subset: train
finetune_dict_specs: null
finetune_from_model: null
restore_file: null
no_save: false
log_interval: 100
eval_lang_pairs: null
reset_dataloader: false
replication_count: 1
reset_all: false