coreconfig.json

{
  "out_name": "flan-t5-small-gigatrue-layercut-D5",
  "remove_layers": true,
  "model_archetype": "flan-t5",
  "model_name": "google/flan-t5-small",
  "tokenizer_name": "google/flan-t5-small",
  "model_torch_dtype": "bfloat16",
  "dataset_lang": "en",
  "dataset_name": "Plasmoxy/gigatrue",
  "dataset_tokenized_cache_name": "gigatrue_tokenized_flan_t5_110-35",
  "use_half_val_dataset": true,
  "max_input_length": 110,
  "max_target_length": 35,
  "batch_size": 256,
  "learning_rate": 0.0003,
  "num_train_epochs": 3,
  "pkg_versions": {
    "optimum": "1.23.3",
    "transformers": "4.45.2",
    "openvino": "2024.6.0",
    "nncf": "2.14.1",
    "torch": "2.5.1",
    "datasets": "3.2.0",
    "peft": "0.13.3.dev0"
  }
}