| {"seed": 52, "batch_size": 4096, "buffer_mult": 384, "lr": 0.0001, "num_tokens": 2000000000, "l1_coeff": 0.0003, "beta1": 0.9, "beta2": 0.99, "dict_mult": 8, "seq_len": 128, "d_mlp": 2048, "enc_dtype": "fp32", "model_batch_size": 512, "buffer_size": 1572864, "buffer_batches": 12288} |