k32-gpt2-exp6-seed2/h.6.mlp/cfg.json

{"expansion_factor": 6, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 768}