{
"tokenizer_dir": "NeTSlab/gpt2-10M-parfind-eng",
"datapoint_length" : 512,
"training_type" : "strict_small",
"n_epochs" : 10,
"batch_size" : 16,
"learning_rate" : 0.00005,
"weight_decay" : 0,
"num_training_steps" : 200000,
"num_warmup_steps" : 2000,
"sft_learning_rate" : 0.00005,
"gradient_clip_norm" : 1,
"seed" : -1,
"base_folder" : "03-models",
"experiment_name" : "gpt2_SylliTokFast_10M",
"use_wandb" : false,
"wandb_experiment_name" : "gpt2_SylliTokFast_10M",
"wandb_project_name" : "BabyLM-2025",
"tokenizer_class": "SyllabicTokenizerWrapper",
"model_type": "gpt2",
"vocab_size": 20535
}