| {"seed": 51, "batch_size": 2048, "buffer_mult": 512, "lr": 2e-05, "num_tokens": 400000000, "l1_coeff": 2, "beta1": 0.9, "beta2": 0.999, "dict_size": 65536, "seq_len": 1024, "enc_dtype": "fp32", "model_name": "gpt2-small", "site": "resid_post", "device": "cuda:0", "model_batch_size": 32, "log_every": 100, "save_every": 100000, "dec_init_norm": 0.005, "name": "gpt2-small_65536_resid_post", "n_layers": 12, "d_model": 768} |