| {"version": 4, "model_name": "EleutherAI/pythia-6.9b-deduped", "layers": [1], "setting": "residual", "tensor_name": "gpt_neox.layers.{layer}.mlp.dense_4h_to_h", "sparsity": 40, "kl": false, "reconstruction": false, "dataset_name": null, "device": "cuda:0", "seed": 0, "model_batch_size": 8, "buffer_mult": 100, "lr": 0.0003, "l1_coeff": 0.4928732931374376, "beta1": 0.9, "beta2": 0.999, "dict_mult": 4, "seq_len": 128, "d_model": 4096, "enc_dtype": "fp32", "remove_rare_dir": false, "max_num_tokens": 1000000000, "d_autoencoder": 16384} |