{ "architectures": [ "KamboForCausalLM" ], "auto_map": { "AutoConfig": "configuration_kambo.KamboConfig", "AutoModel": "modeling_kambo.KamboModel", "AutoModelForCausalLM": "modeling_kambo.KamboForCausalLM" }, "bos_token_id": 151643, "conv_kernel": 3, "d_ff": 1152, "dtype": "bfloat16", "eos_token_id": 151645, "gqa_layers": [ 3, 7, 11, 15, 19, 23 ], "head_dim": 64, "hidden_size": 1024, "intermediate_size": 1152, "max_position_embeddings": 16384, "model_type": "kambo", "n_experts": 16, "num_attention_heads": 16, "num_experts": 16, "num_experts_per_tok": 2, "num_hidden_layers": 24, "num_key_value_heads": 4, "pad_token_id": 151643, "quantization_config": { "checkpoint_format": "dynquant-packed", "group_size": 128, "lm_head_quantized": false, "modules": { "model.embed_tokens": { "bits": 8, "out_features": 151936 }, "model.layers.0.conv.in_proj": { "bits": 8, "out_features": 3072 }, "model.layers.0.conv.out_proj": { "bits": 8, "out_features": 1024 }, "model.layers.0.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.0.moe.sw2": { "bits": 8, "out_features": 1024 }, "model.layers.0.moe.sw3": { "bits": 8, "out_features": 1152 }, "model.layers.0.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.0.moe.w2": { "bits": 2, "out_features": 18432 }, "model.layers.0.moe.w3": { "bits": 4, "out_features": 16384 }, "model.layers.1.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.1.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.1.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.1.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.1.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.1.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.1.moe.w2": { "bits": 2, "out_features": 18432 }, "model.layers.1.moe.w3": { "bits": 2, "out_features": 16384 }, "model.layers.10.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.10.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.10.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.10.moe.sw2": { "bits": 8, "out_features": 1024 }, "model.layers.10.moe.sw3": { "bits": 8, "out_features": 1152 }, "model.layers.10.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.10.moe.w2": { "bits": 4, "out_features": 18432 }, "model.layers.10.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.11.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.11.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.11.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.11.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.11.moe.w2": { "bits": 4, "out_features": 18432 }, "model.layers.11.moe.w3": { "bits": 4, "out_features": 16384 }, "model.layers.11.self_attn.k_proj": { "bits": 8, "out_features": 256 }, "model.layers.11.self_attn.o_proj": { "bits": 8, "out_features": 1024 }, "model.layers.11.self_attn.q_proj": { "bits": 8, "out_features": 1024 }, "model.layers.11.self_attn.v_proj": { "bits": 8, "out_features": 256 }, "model.layers.12.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.12.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.12.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.12.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.12.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.12.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.12.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.12.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.13.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.13.conv.out_proj": { "bits": 8, "out_features": 1024 }, "model.layers.13.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.13.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.13.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.13.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.13.moe.w2": { "bits": 4, "out_features": 18432 }, "model.layers.13.moe.w3": { "bits": 4, "out_features": 16384 }, "model.layers.14.conv.in_proj": { "bits": 8, "out_features": 3072 }, "model.layers.14.conv.out_proj": { "bits": 8, "out_features": 1024 }, "model.layers.14.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.14.moe.sw2": { "bits": 8, "out_features": 1024 }, "model.layers.14.moe.sw3": { "bits": 8, "out_features": 1152 }, "model.layers.14.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.14.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.14.moe.w3": { "bits": 4, "out_features": 16384 }, "model.layers.15.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.15.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.15.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.15.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.15.moe.w2": { "bits": 4, "out_features": 18432 }, "model.layers.15.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.15.self_attn.k_proj": { "bits": 8, "out_features": 256 }, "model.layers.15.self_attn.o_proj": { "bits": 8, "out_features": 1024 }, "model.layers.15.self_attn.q_proj": { "bits": 8, "out_features": 1024 }, "model.layers.15.self_attn.v_proj": { "bits": 8, "out_features": 256 }, "model.layers.16.conv.in_proj": { "bits": 8, "out_features": 3072 }, "model.layers.16.conv.out_proj": { "bits": 8, "out_features": 1024 }, "model.layers.16.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.16.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.16.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.16.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.16.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.16.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.17.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.17.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.17.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.17.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.17.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.17.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.17.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.17.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.18.conv.in_proj": { "bits": 8, "out_features": 3072 }, "model.layers.18.conv.out_proj": { "bits": 8, "out_features": 1024 }, "model.layers.18.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.18.moe.sw2": { "bits": 8, "out_features": 1024 }, "model.layers.18.moe.sw3": { "bits": 8, "out_features": 1152 }, "model.layers.18.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.18.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.18.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.19.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.19.moe.sw2": { "bits": 8, "out_features": 1024 }, "model.layers.19.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.19.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.19.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.19.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.19.self_attn.k_proj": { "bits": 8, "out_features": 256 }, "model.layers.19.self_attn.o_proj": { "bits": 8, "out_features": 1024 }, "model.layers.19.self_attn.q_proj": { "bits": 8, "out_features": 1024 }, "model.layers.19.self_attn.v_proj": { "bits": 8, "out_features": 256 }, "model.layers.2.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.2.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.2.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.2.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.2.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.2.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.2.moe.w2": { "bits": 2, "out_features": 18432 }, "model.layers.2.moe.w3": { "bits": 2, "out_features": 16384 }, "model.layers.20.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.20.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.20.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.20.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.20.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.20.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.20.moe.w2": { "bits": 2, "out_features": 18432 }, "model.layers.20.moe.w3": { "bits": 2, "out_features": 16384 }, "model.layers.21.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.21.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.21.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.21.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.21.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.21.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.21.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.21.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.22.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.22.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.22.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.22.moe.sw2": { "bits": 8, "out_features": 1024 }, "model.layers.22.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.22.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.22.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.22.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.23.moe.sw1": { "bits": 8, "out_features": 1152 }, "model.layers.23.moe.sw2": { "bits": 8, "out_features": 1024 }, "model.layers.23.moe.sw3": { "bits": 8, "out_features": 1152 }, "model.layers.23.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.23.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.23.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.23.self_attn.k_proj": { "bits": 8, "out_features": 256 }, "model.layers.23.self_attn.o_proj": { "bits": 4, "out_features": 1024 }, "model.layers.23.self_attn.q_proj": { "bits": 8, "out_features": 1024 }, "model.layers.23.self_attn.v_proj": { "bits": 8, "out_features": 256 }, "model.layers.3.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.3.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.3.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.3.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.3.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.3.moe.w3": { "bits": 2, "out_features": 16384 }, "model.layers.3.self_attn.k_proj": { "bits": 8, "out_features": 256 }, "model.layers.3.self_attn.o_proj": { "bits": 8, "out_features": 1024 }, "model.layers.3.self_attn.q_proj": { "bits": 8, "out_features": 1024 }, "model.layers.3.self_attn.v_proj": { "bits": 8, "out_features": 256 }, "model.layers.4.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.4.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.4.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.4.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.4.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.4.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.4.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.4.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.5.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.5.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.5.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.5.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.5.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.5.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.5.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.5.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.6.conv.in_proj": { "bits": 8, "out_features": 3072 }, "model.layers.6.conv.out_proj": { "bits": 8, "out_features": 1024 }, "model.layers.6.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.6.moe.sw2": { "bits": 8, "out_features": 1024 }, "model.layers.6.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.6.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.6.moe.w2": { "bits": 2, "out_features": 18432 }, "model.layers.6.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.7.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.7.moe.sw2": { "bits": 8, "out_features": 1024 }, "model.layers.7.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.7.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.7.moe.w2": { "bits": 3, "out_features": 18432 }, "model.layers.7.moe.w3": { "bits": 3, "out_features": 16384 }, "model.layers.7.self_attn.k_proj": { "bits": 8, "out_features": 256 }, "model.layers.7.self_attn.o_proj": { "bits": 8, "out_features": 1024 }, "model.layers.7.self_attn.q_proj": { "bits": 8, "out_features": 1024 }, "model.layers.7.self_attn.v_proj": { "bits": 8, "out_features": 256 }, "model.layers.8.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.8.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.8.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.8.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.8.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.8.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.8.moe.w2": { "bits": 4, "out_features": 18432 }, "model.layers.8.moe.w3": { "bits": 4, "out_features": 16384 }, "model.layers.9.conv.in_proj": { "bits": 4, "out_features": 3072 }, "model.layers.9.conv.out_proj": { "bits": 4, "out_features": 1024 }, "model.layers.9.moe.sw1": { "bits": 4, "out_features": 1152 }, "model.layers.9.moe.sw2": { "bits": 4, "out_features": 1024 }, "model.layers.9.moe.sw3": { "bits": 4, "out_features": 1152 }, "model.layers.9.moe.w1": { "bits": 4, "out_features": 16384 }, "model.layers.9.moe.w2": { "bits": 4, "out_features": 18432 }, "model.layers.9.moe.w3": { "bits": 4, "out_features": 16384 } }, "modules_to_not_convert": [], "quant_method": "dynquant", "schema_version": 1, "symmetric": false, "version": "0.5.3" }, "rms_norm_eps": 1e-06, "rope_theta": 40000.0, "tie_word_embeddings": true, "top_k": 2, "transformers_version": "5.14.1", "use_cache": true, "vocab_size": 151936 }