| {"architectures": ["GPT2SLMHeadModel"], "activation_function": "gelu_new", "bos_token_id": 50256, "embd_pdrop": 0.1, "eos_token_id": 50256, "layer_norm_epsilon": 1e-05, "model_type": "gpt2", "n_embd": 768, "num_senses": 16, "resid_pdrop": 0.1, "sense_intermediate_scale": 4, "transformers_version": "4.29.0.dev0", "vocab_size": 50264} |