| { | |
| "apply_residual_connection_post_layernorm": false, | |
| "architectures": [ | |
| "BloomForSequenceClassification" | |
| ], | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 1, | |
| "dtype": "float32", | |
| "eos_token_id": 2, | |
| "hidden_dropout": 0.0, | |
| "hidden_size": 32, | |
| "initializer_range": 0.02, | |
| "layer_norm_epsilon": 1e-05, | |
| "model_type": "bloom", | |
| "n_head": 2, | |
| "n_layer": 2, | |
| "pad_token_id": 3, | |
| "pretraining_tp": 1, | |
| "slow_but_exact": false, | |
| "tie_word_embeddings": false, | |
| "transformers_version": "5.16.0.dev0", | |
| "use_cache": true, | |
| "vocab_size": 250680 | |
| } | |