config.json

{
  "_name_or_path": "./squad_bert_base/",
  "architectures": [
    "BertForQuestionAnswering"
  ],
  "attention_probs_dropout_prob": 0.1,
  "classifier_dropout": null,
  "device_iterations": 2,
  "embedding_serialization_factor": 0,
  "enable_half_first_order_momentum": true,
  "enable_half_partials": true,
  "executable_cache_dir": "./exe_cache",
  "gradient_accumulation_steps": 16,
  "gradient_checkpointing": false,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.1,
  "hidden_size": 768,
  "inference_device_iterations": 4,
  "inference_replication_factor": 4,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "ipus_per_replica": 4,
  "layer_norm_eps": 1e-12,
  "layers_per_ipu": [
    0,
    4,
    4,
    4
  ],
  "matmul_proportion": [
    0.25,
    0.2,
    0.2,
    0.25
  ],
  "max_position_embeddings": 512,
  "model_type": "bert",
  "num_attention_heads": 12,
  "num_hidden_layers": 12,
  "optimizer_state_offchip": false,
  "optimum_version": "0.1.2",
  "pad_token_id": 0,
  "position_embedding_type": "absolute",
  "profile_dir": null,
  "recompute_checkpoint_every_layer": true,
  "replicated_tensor_sharding": true,
  "replication_factor": 4,
  "seed": null,
  "synthetic_data": false,
  "torch_dtype": "float32",
  "transformers_version": "4.17.0.dev0",
  "type_vocab_size": 2,
  "use_cache": true,
  "use_popdist": false,
  "vocab_size": 30522
}