| { | |
| "architectures": [ | |
| "BabyLMForCausalLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "hf_wrapper.BabyLMConfig", | |
| "AutoModelForCausalLM": "hf_wrapper.BabyLMForCausalLM" | |
| }, | |
| "dropout": 0.1, | |
| "dtype": "float32", | |
| "ffn_intermediate_size": 1380, | |
| "hidden_size": 512, | |
| "ipa_dim": 48, | |
| "ipa_lambda": 1.0, | |
| "ipa_mask_ratio": 0.15, | |
| "max_seq_len": 512, | |
| "model_type": "babylm_ipa", | |
| "num_heads": 8, | |
| "num_layers": 6, | |
| "transformers_version": "5.4.0", | |
| "variant": "baseline", | |
| "vocab_size": 32000 | |
| } | |