{ "apply_residual_connection_post_layernorm": false, "architectures": [ "BloomForCausalLM" ], "attention_dropout": 0.0, "attention_softmax_in_fp32": true, "bias_dropout_fusion": true, "bos_token_id": 1, "depth_downsampling_rate": 0.5, "eos_token_id": 2, "hidden_dropout": 0.0, "initializer_range": 0.02, "layer_norm_epsilon": 1e-05, "layer_selection_strategy": "step", "masked_softmax_fusion": true, "model_type": "bloom", "n_embed": 1024, "n_inner": null, "n_layer": 12, "num_attention_heads": 8, "offset_alibi": 100, "pad_token_id": 3, "pretraining_tp": 2, "seq_length": 4096, "skip_bias_add": true, "skip_bias_add_qkv": false, "slow_but_exact": false, "transformers_version": "4.21.0.dev0", "unk_token_id": 0, "use_cache": true, "vocab_size": 250880, "weights_aggregation_strategy": "mean", "width_downsampling_rate": 0.5 }