{ "architectures": [ "Gemma3ForConditionalGeneration" ], "boi_token_index": 255999, "eoi_token_index": 256000, "eos_token_id": [ 1, 106 ], "image_token_index": 262144, "initializer_range": 0.02, "mm_tokens_per_image": 256, "model_type": "gemma3", "pad_token_id": 0, "quantization": { "group_size": 32, "bits": 6, "language_model.model.embed_tokens": true, "language_model.model.layers.0.self_attn.q_proj": true, "language_model.model.layers.0.self_attn.k_proj": true, "language_model.model.layers.0.self_attn.v_proj": true, "language_model.model.layers.0.self_attn.o_proj": true, "language_model.model.layers.0.self_attn.q_norm": false, "language_model.model.layers.0.self_attn.k_norm": false, "language_model.model.layers.0.self_attn.rope": false, "language_model.model.layers.0.mlp.gate_proj": true, "language_model.model.layers.0.mlp.down_proj": true, "language_model.model.layers.0.mlp.up_proj": true, "language_model.model.layers.0.input_layernorm": false, "language_model.model.layers.0.post_attention_layernorm": false, "language_model.model.layers.0.pre_feedforward_layernorm": false, "language_model.model.layers.0.post_feedforward_layernorm": false, "language_model.model.layers.1.self_attn.q_proj": true, "language_model.model.layers.1.self_attn.k_proj": true, "language_model.model.layers.1.self_attn.v_proj": true, "language_model.model.layers.1.self_attn.o_proj": true, "language_model.model.layers.1.self_attn.q_norm": false, "language_model.model.layers.1.self_attn.k_norm": false, "language_model.model.layers.1.self_attn.rope": false, "language_model.model.layers.1.mlp.gate_proj": true, "language_model.model.layers.1.mlp.down_proj": true, "language_model.model.layers.1.mlp.up_proj": true, "language_model.model.layers.1.input_layernorm": false, "language_model.model.layers.1.post_attention_layernorm": false, "language_model.model.layers.1.pre_feedforward_layernorm": false, "language_model.model.layers.1.post_feedforward_layernorm": false, "language_model.model.layers.2.self_attn.q_proj": true, "language_model.model.layers.2.self_attn.k_proj": true, "language_model.model.layers.2.self_attn.v_proj": { "bits": 8, "group_size": 32 }, "language_model.model.layers.2.self_attn.o_proj": true, "language_model.model.layers.2.self_attn.q_norm": false, "language_model.model.layers.2.self_attn.k_norm": false, "language_model.model.layers.2.self_attn.rope": false, "language_model.model.layers.2.mlp.gate_proj": true, "language_model.model.layers.2.mlp.down_proj": true, "language_model.model.layers.2.mlp.up_proj": true, "language_model.model.layers.2.input_layernorm": false, "language_model.model.layers.2.post_attention_layernorm": false, "language_model.model.layers.2.pre_feedforward_layernorm": false, "language_model.model.layers.2.post_feedforward_layernorm": false, "language_model.model.layers.3.self_attn.q_proj": true, "language_model.model.layers.3.self_attn.k_proj": true, "language_model.model.layers.3.self_attn.v_proj": true, "language_model.model.layers.3.self_attn.o_proj": true, "language_model.model.layers.3.self_attn.q_norm": false, "language_model.model.layers.3.self_attn.k_norm": false, "language_model.model.layers.3.self_attn.rope": false, "language_model.model.layers.3.mlp.gate_proj": true, "language_model.model.layers.3.mlp.down_proj": true, "language_model.model.layers.3.mlp.up_proj": true, "language_model.model.layers.3.input_layernorm": false, "language_model.model.layers.3.post_attention_layernorm": false, "language_model.model.layers.3.pre_feedforward_layernorm": false, "language_model.model.layers.3.post_feedforward_layernorm": false, "language_model.model.layers.4.self_attn.q_proj": true, "language_model.model.layers.4.self_attn.k_proj": true, "language_model.model.layers.4.self_attn.v_proj": true, "language_model.model.layers.4.self_attn.o_proj": true, "language_model.model.layers.4.self_attn.q_norm": false, "language_model.model.layers.4.self_attn.k_norm": false, "language_model.model.layers.4.self_attn.rope": false, "language_model.model.layers.4.mlp.gate_proj": true, "language_model.model.layers.4.mlp.down_proj": true, "language_model.model.layers.4.mlp.up_proj": true, "language_model.model.layers.4.input_layernorm": false, "language_model.model.layers.4.post_attention_layernorm": false, "language_model.model.layers.4.pre_feedforward_layernorm": false, "language_model.model.layers.4.post_feedforward_layernorm": false, "language_model.model.layers.5.self_attn.q_proj": true, "language_model.model.layers.5.self_attn.k_proj": true, "language_model.model.layers.5.self_attn.v_proj": true, "language_model.model.layers.5.self_attn.o_proj": true, "language_model.model.layers.5.self_attn.q_norm": false, "language_model.model.layers.5.self_attn.k_norm": false, "language_model.model.layers.5.self_attn.rope": false, "language_model.model.layers.5.mlp.gate_proj": true, "language_model.model.layers.5.mlp.down_proj": true, "language_model.model.layers.5.mlp.up_proj": true, "language_model.model.layers.5.input_layernorm": false, "language_model.model.layers.5.post_attention_layernorm": false, "language_model.model.layers.5.pre_feedforward_layernorm": false, "language_model.model.layers.5.post_feedforward_layernorm": false, "language_model.model.layers.6.self_attn.q_proj": true, "language_model.model.layers.6.self_attn.k_proj": true, "language_model.model.layers.6.self_attn.v_proj": true, "language_model.model.layers.6.self_attn.o_proj": true, "language_model.model.layers.6.self_attn.q_norm": false, "language_model.model.layers.6.self_attn.k_norm": false, "language_model.model.layers.6.self_attn.rope": false, "language_model.model.layers.6.mlp.gate_proj": true, "language_model.model.layers.6.mlp.down_proj": true, "language_model.model.layers.6.mlp.up_proj": true, "language_model.model.layers.6.input_layernorm": false, "language_model.model.layers.6.post_attention_layernorm": false, "language_model.model.layers.6.pre_feedforward_layernorm": false, "language_model.model.layers.6.post_feedforward_layernorm": false, "language_model.model.layers.7.self_attn.q_proj": true, "language_model.model.layers.7.self_attn.k_proj": true, "language_model.model.layers.7.self_attn.v_proj": true, "language_model.model.layers.7.self_attn.o_proj": true, "language_model.model.layers.7.self_attn.q_norm": false, "language_model.model.layers.7.self_attn.k_norm": false, "language_model.model.layers.7.self_attn.rope": false, "language_model.model.layers.7.mlp.gate_proj": true, "language_model.model.layers.7.mlp.down_proj": true, "language_model.model.layers.7.mlp.up_proj": true, "language_model.model.layers.7.input_layernorm": false, "language_model.model.layers.7.post_attention_layernorm": false, "language_model.model.layers.7.pre_feedforward_layernorm": false, "language_model.model.layers.7.post_feedforward_layernorm": false, "language_model.model.layers.8.self_attn.q_proj": true, "language_model.model.layers.8.self_attn.k_proj": true, "language_model.model.layers.8.self_attn.v_proj": true, "language_model.model.layers.8.self_attn.o_proj": true, "language_model.model.layers.8.self_attn.q_norm": false, "language_model.model.layers.8.self_attn.k_norm": false, "language_model.model.layers.8.self_attn.rope": false, "language_model.model.layers.8.mlp.gate_proj": true, "language_model.model.layers.8.mlp.down_proj": true, "language_model.model.layers.8.mlp.up_proj": true, "language_model.model.layers.8.input_layernorm": false, "language_model.model.layers.8.post_attention_layernorm": false, "language_model.model.layers.8.pre_feedforward_layernorm": false, "language_model.model.layers.8.post_feedforward_layernorm": false, "language_model.model.layers.9.self_attn.q_proj": true, "language_model.model.layers.9.self_attn.k_proj": true, "language_model.model.layers.9.self_attn.v_proj": true, "language_model.model.layers.9.self_attn.o_proj": true, "language_model.model.layers.9.self_attn.q_norm": false, "language_model.model.layers.9.self_attn.k_norm": false, "language_model.model.layers.9.self_attn.rope": false, "language_model.model.layers.9.mlp.gate_proj": true, "language_model.model.layers.9.mlp.down_proj": true, "language_model.model.layers.9.mlp.up_proj": true, "language_model.model.layers.9.input_layernorm": false, "language_model.model.layers.9.post_attention_layernorm": false, "language_model.model.layers.9.pre_feedforward_layernorm": false, "language_model.model.layers.9.post_feedforward_layernorm": false, "language_model.model.layers.10.self_attn.q_proj": true, "language_model.model.layers.10.self_attn.k_proj": true, "language_model.model.layers.10.self_attn.v_proj": true, "language_model.model.layers.10.self_attn.o_proj": true, "language_model.model.layers.10.self_attn.q_norm": false, "language_model.model.layers.10.self_attn.k_norm": false, "language_model.model.layers.10.self_attn.rope": false, "language_model.model.layers.10.mlp.gate_proj": true, "language_model.model.layers.10.mlp.down_proj": true, "language_model.model.layers.10.mlp.up_proj": true, "language_model.model.layers.10.input_layernorm": false, "language_model.model.layers.10.post_attention_layernorm": false, "language_model.model.layers.10.pre_feedforward_layernorm": false, "language_model.model.layers.10.post_feedforward_layernorm": false, "language_model.model.layers.11.self_attn.q_proj": true, "language_model.model.layers.11.self_attn.k_proj": true, "language_model.model.layers.11.self_attn.v_proj": true, "language_model.model.layers.11.self_attn.o_proj": true, "language_model.model.layers.11.self_attn.q_norm": false, "language_model.model.layers.11.self_attn.k_norm": false, "language_model.model.layers.11.self_attn.rope": false, "language_model.model.layers.11.mlp.gate_proj": true, "language_model.model.layers.11.mlp.down_proj": true, "language_model.model.layers.11.mlp.up_proj": true, "language_model.model.layers.11.input_layernorm": false, "language_model.model.layers.11.post_attention_layernorm": false, "language_model.model.layers.11.pre_feedforward_layernorm": false, "language_model.model.layers.11.post_feedforward_layernorm": false, "language_model.model.layers.12.self_attn.q_proj": true, "language_model.model.layers.12.self_attn.k_proj": true, "language_model.model.layers.12.self_attn.v_proj": true, "language_model.model.layers.12.self_attn.o_proj": true, "language_model.model.layers.12.self_attn.q_norm": false, "language_model.model.layers.12.self_attn.k_norm": false, "language_model.model.layers.12.self_attn.rope": false, "language_model.model.layers.12.mlp.gate_proj": true, "language_model.model.layers.12.mlp.down_proj": true, "language_model.model.layers.12.mlp.up_proj": true, "language_model.model.layers.12.input_layernorm": false, "language_model.model.layers.12.post_attention_layernorm": false, "language_model.model.layers.12.pre_feedforward_layernorm": false, "language_model.model.layers.12.post_feedforward_layernorm": false, "language_model.model.layers.13.self_attn.q_proj": true, "language_model.model.layers.13.self_attn.k_proj": true, "language_model.model.layers.13.self_attn.v_proj": true, "language_model.model.layers.13.self_attn.o_proj": true, "language_model.model.layers.13.self_attn.q_norm": false, "language_model.model.layers.13.self_attn.k_norm": false, "language_model.model.layers.13.self_attn.rope": false, "language_model.model.layers.13.mlp.gate_proj": true, "language_model.model.layers.13.mlp.down_proj": true, "language_model.model.layers.13.mlp.up_proj": true, "language_model.model.layers.13.input_layernorm": false, "language_model.model.layers.13.post_attention_layernorm": false, "language_model.model.layers.13.pre_feedforward_layernorm": false, "language_model.model.layers.13.post_feedforward_layernorm": false, "language_model.model.layers.14.self_attn.q_proj": true, "language_model.model.layers.14.self_attn.k_proj": true, "language_model.model.layers.14.self_attn.v_proj": true, "language_model.model.layers.14.self_attn.o_proj": true, "language_model.model.layers.14.self_attn.q_norm": false, "language_model.model.layers.14.self_attn.k_norm": false, "language_model.model.layers.14.self_attn.rope": false, "language_model.model.layers.14.mlp.gate_proj": true, "language_model.model.layers.14.mlp.down_proj": true, "language_model.model.layers.14.mlp.up_proj": true, "language_model.model.layers.14.input_layernorm": false, "language_model.model.layers.14.post_attention_layernorm": false, "language_model.model.layers.14.pre_feedforward_layernorm": false, "language_model.model.layers.14.post_feedforward_layernorm": false, "language_model.model.layers.15.self_attn.q_proj": true, "language_model.model.layers.15.self_attn.k_proj": true, "language_model.model.layers.15.self_attn.v_proj": true, "language_model.model.layers.15.self_attn.o_proj": true, "language_model.model.layers.15.self_attn.q_norm": false, "language_model.model.layers.15.self_attn.k_norm": false, "language_model.model.layers.15.self_attn.rope": false, "language_model.model.layers.15.mlp.gate_proj": true, "language_model.model.layers.15.mlp.down_proj": true, "language_model.model.layers.15.mlp.up_proj": true, "language_model.model.layers.15.input_layernorm": false, "language_model.model.layers.15.post_attention_layernorm": false, "language_model.model.layers.15.pre_feedforward_layernorm": false, "language_model.model.layers.15.post_feedforward_layernorm": false, "language_model.model.layers.16.self_attn.q_proj": true, "language_model.model.layers.16.self_attn.k_proj": true, "language_model.model.layers.16.self_attn.v_proj": true, "language_model.model.layers.16.self_attn.o_proj": true, "language_model.model.layers.16.self_attn.q_norm": false, "language_model.model.layers.16.self_attn.k_norm": false, "language_model.model.layers.16.self_attn.rope": false, "language_model.model.layers.16.mlp.gate_proj": true, "language_model.model.layers.16.mlp.down_proj": true, "language_model.model.layers.16.mlp.up_proj": true, "language_model.model.layers.16.input_layernorm": false, "language_model.model.layers.16.post_attention_layernorm": false, "language_model.model.layers.16.pre_feedforward_layernorm": false, "language_model.model.layers.16.post_feedforward_layernorm": false, "language_model.model.layers.17.self_attn.q_proj": true, "language_model.model.layers.17.self_attn.k_proj": true, "language_model.model.layers.17.self_attn.v_proj": true, "language_model.model.layers.17.self_attn.o_proj": true, "language_model.model.layers.17.self_attn.q_norm": false, "language_model.model.layers.17.self_attn.k_norm": false, "language_model.model.layers.17.self_attn.rope": false, "language_model.model.layers.17.mlp.gate_proj": true, "language_model.model.layers.17.mlp.down_proj": true, "language_model.model.layers.17.mlp.up_proj": true, "language_model.model.layers.17.input_layernorm": false, "language_model.model.layers.17.post_attention_layernorm": false, "language_model.model.layers.17.pre_feedforward_layernorm": false, "language_model.model.layers.17.post_feedforward_layernorm": false, "language_model.model.layers.18.self_attn.q_proj": true, "language_model.model.layers.18.self_attn.k_proj": true, "language_model.model.layers.18.self_attn.v_proj": true, "language_model.model.layers.18.self_attn.o_proj": true, "language_model.model.layers.18.self_attn.q_norm": false, "language_model.model.layers.18.self_attn.k_norm": false, "language_model.model.layers.18.self_attn.rope": false, "language_model.model.layers.18.mlp.gate_proj": true, "language_model.model.layers.18.mlp.down_proj": true, "language_model.model.layers.18.mlp.up_proj": true, "language_model.model.layers.18.input_layernorm": false, "language_model.model.layers.18.post_attention_layernorm": false, "language_model.model.layers.18.pre_feedforward_layernorm": false, "language_model.model.layers.18.post_feedforward_layernorm": false, "language_model.model.layers.19.self_attn.q_proj": true, "language_model.model.layers.19.self_attn.k_proj": true, "language_model.model.layers.19.self_attn.v_proj": true, "language_model.model.layers.19.self_attn.o_proj": true, "language_model.model.layers.19.self_attn.q_norm": false, "language_model.model.layers.19.self_attn.k_norm": false, "language_model.model.layers.19.self_attn.rope": false, "language_model.model.layers.19.mlp.gate_proj": true, "language_model.model.layers.19.mlp.down_proj": true, "language_model.model.layers.19.mlp.up_proj": true, "language_model.model.layers.19.input_layernorm": false, "language_model.model.layers.19.post_attention_layernorm": false, "language_model.model.layers.19.pre_feedforward_layernorm": false, "language_model.model.layers.19.post_feedforward_layernorm": false, "language_model.model.layers.20.self_attn.q_proj": true, "language_model.model.layers.20.self_attn.k_proj": true, "language_model.model.layers.20.self_attn.v_proj": true, "language_model.model.layers.20.self_attn.o_proj": true, "language_model.model.layers.20.self_attn.q_norm": false, "language_model.model.layers.20.self_attn.k_norm": false, "language_model.model.layers.20.self_attn.rope": false, "language_model.model.layers.20.mlp.gate_proj": true, "language_model.model.layers.20.mlp.down_proj": true, "language_model.model.layers.20.mlp.up_proj": true, "language_model.model.layers.20.input_layernorm": false, "language_model.model.layers.20.post_attention_layernorm": false, "language_model.model.layers.20.pre_feedforward_layernorm": false, "language_model.model.layers.20.post_feedforward_layernorm": false, "language_model.model.layers.21.self_attn.q_proj": true, "language_model.model.layers.21.self_attn.k_proj": true, "language_model.model.layers.21.self_attn.v_proj": true, "language_model.model.layers.21.self_attn.o_proj": true, "language_model.model.layers.21.self_attn.q_norm": false, "language_model.model.layers.21.self_attn.k_norm": false, "language_model.model.layers.21.self_attn.rope": false, "language_model.model.layers.21.mlp.gate_proj": true, "language_model.model.layers.21.mlp.down_proj": true, "language_model.model.layers.21.mlp.up_proj": true, "language_model.model.layers.21.input_layernorm": false, "language_model.model.layers.21.post_attention_layernorm": false, "language_model.model.layers.21.pre_feedforward_layernorm": false, "language_model.model.layers.21.post_feedforward_layernorm": false, "language_model.model.layers.22.self_attn.q_proj": true, "language_model.model.layers.22.self_attn.k_proj": true, "language_model.model.layers.22.self_attn.v_proj": true, "language_model.model.layers.22.self_attn.o_proj": true, "language_model.model.layers.22.self_attn.q_norm": false, "language_model.model.layers.22.self_attn.k_norm": false, "language_model.model.layers.22.self_attn.rope": false, "language_model.model.layers.22.mlp.gate_proj": true, "language_model.model.layers.22.mlp.down_proj": true, "language_model.model.layers.22.mlp.up_proj": true, "language_model.model.layers.22.input_layernorm": false, "language_model.model.layers.22.post_attention_layernorm": false, "language_model.model.layers.22.pre_feedforward_layernorm": false, "language_model.model.layers.22.post_feedforward_layernorm": false, "language_model.model.layers.23.self_attn.q_proj": true, "language_model.model.layers.23.self_attn.k_proj": true, "language_model.model.layers.23.self_attn.v_proj": true, "language_model.model.layers.23.self_attn.o_proj": true, "language_model.model.layers.23.self_attn.q_norm": false, "language_model.model.layers.23.self_attn.k_norm": false, "language_model.model.layers.23.self_attn.rope": false, "language_model.model.layers.23.mlp.gate_proj": true, "language_model.model.layers.23.mlp.down_proj": true, "language_model.model.layers.23.mlp.up_proj": true, "language_model.model.layers.23.input_layernorm": false, "language_model.model.layers.23.post_attention_layernorm": false, "language_model.model.layers.23.pre_feedforward_layernorm": false, "language_model.model.layers.23.post_feedforward_layernorm": false, "language_model.model.layers.24.self_attn.q_proj": true, "language_model.model.layers.24.self_attn.k_proj": true, "language_model.model.layers.24.self_attn.v_proj": true, "language_model.model.layers.24.self_attn.o_proj": true, "language_model.model.layers.24.self_attn.q_norm": false, "language_model.model.layers.24.self_attn.k_norm": false, "language_model.model.layers.24.self_attn.rope": false, "language_model.model.layers.24.mlp.gate_proj": true, "language_model.model.layers.24.mlp.down_proj": true, "language_model.model.layers.24.mlp.up_proj": true, "language_model.model.layers.24.input_layernorm": false, "language_model.model.layers.24.post_attention_layernorm": false, "language_model.model.layers.24.pre_feedforward_layernorm": false, "language_model.model.layers.24.post_feedforward_layernorm": false, "language_model.model.layers.25.self_attn.q_proj": true, "language_model.model.layers.25.self_attn.k_proj": true, "language_model.model.layers.25.self_attn.v_proj": true, "language_model.model.layers.25.self_attn.o_proj": true, "language_model.model.layers.25.self_attn.q_norm": false, "language_model.model.layers.25.self_attn.k_norm": false, "language_model.model.layers.25.self_attn.rope": false, "language_model.model.layers.25.mlp.gate_proj": true, "language_model.model.layers.25.mlp.down_proj": true, "language_model.model.layers.25.mlp.up_proj": true, "language_model.model.layers.25.input_layernorm": false, "language_model.model.layers.25.post_attention_layernorm": false, "language_model.model.layers.25.pre_feedforward_layernorm": false, "language_model.model.layers.25.post_feedforward_layernorm": false, "language_model.model.layers.26.self_attn.q_proj": true, "language_model.model.layers.26.self_attn.k_proj": true, "language_model.model.layers.26.self_attn.v_proj": true, "language_model.model.layers.26.self_attn.o_proj": true, "language_model.model.layers.26.self_attn.q_norm": false, "language_model.model.layers.26.self_attn.k_norm": false, "language_model.model.layers.26.self_attn.rope": false, "language_model.model.layers.26.mlp.gate_proj": true, "language_model.model.layers.26.mlp.down_proj": true, "language_model.model.layers.26.mlp.up_proj": true, "language_model.model.layers.26.input_layernorm": false, "language_model.model.layers.26.post_attention_layernorm": false, "language_model.model.layers.26.pre_feedforward_layernorm": false, "language_model.model.layers.26.post_feedforward_layernorm": false, "language_model.model.layers.27.self_attn.q_proj": true, "language_model.model.layers.27.self_attn.k_proj": true, "language_model.model.layers.27.self_attn.v_proj": true, "language_model.model.layers.27.self_attn.o_proj": true, "language_model.model.layers.27.self_attn.q_norm": false, "language_model.model.layers.27.self_attn.k_norm": false, "language_model.model.layers.27.self_attn.rope": false, "language_model.model.layers.27.mlp.gate_proj": true, "language_model.model.layers.27.mlp.down_proj": true, "language_model.model.layers.27.mlp.up_proj": true, "language_model.model.layers.27.input_layernorm": false, "language_model.model.layers.27.post_attention_layernorm": false, "language_model.model.layers.27.pre_feedforward_layernorm": false, "language_model.model.layers.27.post_feedforward_layernorm": false, "language_model.model.layers.28.self_attn.q_proj": true, "language_model.model.layers.28.self_attn.k_proj": true, "language_model.model.layers.28.self_attn.v_proj": true, "language_model.model.layers.28.self_attn.o_proj": true, "language_model.model.layers.28.self_attn.q_norm": false, "language_model.model.layers.28.self_attn.k_norm": false, "language_model.model.layers.28.self_attn.rope": false, "language_model.model.layers.28.mlp.gate_proj": true, "language_model.model.layers.28.mlp.down_proj": true, "language_model.model.layers.28.mlp.up_proj": true, "language_model.model.layers.28.input_layernorm": false, "language_model.model.layers.28.post_attention_layernorm": false, "language_model.model.layers.28.pre_feedforward_layernorm": false, "language_model.model.layers.28.post_feedforward_layernorm": false, "language_model.model.layers.29.self_attn.q_proj": true, "language_model.model.layers.29.self_attn.k_proj": true, "language_model.model.layers.29.self_attn.v_proj": true, "language_model.model.layers.29.self_attn.o_proj": true, "language_model.model.layers.29.self_attn.q_norm": false, "language_model.model.layers.29.self_attn.k_norm": false, "language_model.model.layers.29.self_attn.rope": false, "language_model.model.layers.29.mlp.gate_proj": true, "language_model.model.layers.29.mlp.down_proj": true, "language_model.model.layers.29.mlp.up_proj": true, "language_model.model.layers.29.input_layernorm": false, "language_model.model.layers.29.post_attention_layernorm": false, "language_model.model.layers.29.pre_feedforward_layernorm": false, "language_model.model.layers.29.post_feedforward_layernorm": false, "language_model.model.layers.30.self_attn.q_proj": true, "language_model.model.layers.30.self_attn.k_proj": true, "language_model.model.layers.30.self_attn.v_proj": true, "language_model.model.layers.30.self_attn.o_proj": true, "language_model.model.layers.30.self_attn.q_norm": false, "language_model.model.layers.30.self_attn.k_norm": false, "language_model.model.layers.30.self_attn.rope": false, "language_model.model.layers.30.mlp.gate_proj": true, "language_model.model.layers.30.mlp.down_proj": true, "language_model.model.layers.30.mlp.up_proj": true, "language_model.model.layers.30.input_layernorm": false, "language_model.model.layers.30.post_attention_layernorm": false, "language_model.model.layers.30.pre_feedforward_layernorm": false, "language_model.model.layers.30.post_feedforward_layernorm": false, "language_model.model.layers.31.self_attn.q_proj": true, "language_model.model.layers.31.self_attn.k_proj": true, "language_model.model.layers.31.self_attn.v_proj": true, "language_model.model.layers.31.self_attn.o_proj": true, "language_model.model.layers.31.self_attn.q_norm": false, "language_model.model.layers.31.self_attn.k_norm": false, "language_model.model.layers.31.self_attn.rope": false, "language_model.model.layers.31.mlp.gate_proj": true, "language_model.model.layers.31.mlp.down_proj": true, "language_model.model.layers.31.mlp.up_proj": true, "language_model.model.layers.31.input_layernorm": false, "language_model.model.layers.31.post_attention_layernorm": false, "language_model.model.layers.31.pre_feedforward_layernorm": false, "language_model.model.layers.31.post_feedforward_layernorm": false, "language_model.model.layers.32.self_attn.q_proj": true, "language_model.model.layers.32.self_attn.k_proj": true, "language_model.model.layers.32.self_attn.v_proj": true, "language_model.model.layers.32.self_attn.o_proj": true, "language_model.model.layers.32.self_attn.q_norm": false, "language_model.model.layers.32.self_attn.k_norm": false, "language_model.model.layers.32.self_attn.rope": false, "language_model.model.layers.32.mlp.gate_proj": true, "language_model.model.layers.32.mlp.down_proj": true, "language_model.model.layers.32.mlp.up_proj": true, "language_model.model.layers.32.input_layernorm": false, "language_model.model.layers.32.post_attention_layernorm": false, "language_model.model.layers.32.pre_feedforward_layernorm": false, "language_model.model.layers.32.post_feedforward_layernorm": false, "language_model.model.layers.33.self_attn.q_proj": true, "language_model.model.layers.33.self_attn.k_proj": true, "language_model.model.layers.33.self_attn.v_proj": true, "language_model.model.layers.33.self_attn.o_proj": true, "language_model.model.layers.33.self_attn.q_norm": false, "language_model.model.layers.33.self_attn.k_norm": false, "language_model.model.layers.33.self_attn.rope": false, "language_model.model.layers.33.mlp.gate_proj": true, "language_model.model.layers.33.mlp.down_proj": true, "language_model.model.layers.33.mlp.up_proj": true, "language_model.model.layers.33.input_layernorm": false, "language_model.model.layers.33.post_attention_layernorm": false, "language_model.model.layers.33.pre_feedforward_layernorm": false, "language_model.model.layers.33.post_feedforward_layernorm": false, "language_model.model.norm": false, "language_model.lm_head": true }, "quantization_config": { "group_size": 32, "bits": 6, "language_model.model.embed_tokens": true, "language_model.model.layers.0.self_attn.q_proj": true, "language_model.model.layers.0.self_attn.k_proj": true, "language_model.model.layers.0.self_attn.v_proj": true, "language_model.model.layers.0.self_attn.o_proj": true, "language_model.model.layers.0.self_attn.q_norm": false, "language_model.model.layers.0.self_attn.k_norm": false, "language_model.model.layers.0.self_attn.rope": false, "language_model.model.layers.0.mlp.gate_proj": true, "language_model.model.layers.0.mlp.down_proj": true, "language_model.model.layers.0.mlp.up_proj": true, "language_model.model.layers.0.input_layernorm": false, "language_model.model.layers.0.post_attention_layernorm": false, "language_model.model.layers.0.pre_feedforward_layernorm": false, "language_model.model.layers.0.post_feedforward_layernorm": false, "language_model.model.layers.1.self_attn.q_proj": true, "language_model.model.layers.1.self_attn.k_proj": true, "language_model.model.layers.1.self_attn.v_proj": true, "language_model.model.layers.1.self_attn.o_proj": true, "language_model.model.layers.1.self_attn.q_norm": false, "language_model.model.layers.1.self_attn.k_norm": false, "language_model.model.layers.1.self_attn.rope": false, "language_model.model.layers.1.mlp.gate_proj": true, "language_model.model.layers.1.mlp.down_proj": true, "language_model.model.layers.1.mlp.up_proj": true, "language_model.model.layers.1.input_layernorm": false, "language_model.model.layers.1.post_attention_layernorm": false, "language_model.model.layers.1.pre_feedforward_layernorm": false, "language_model.model.layers.1.post_feedforward_layernorm": false, "language_model.model.layers.2.self_attn.q_proj": true, "language_model.model.layers.2.self_attn.k_proj": true, "language_model.model.layers.2.self_attn.v_proj": { "bits": 8, "group_size": 32 }, "language_model.model.layers.2.self_attn.o_proj": true, "language_model.model.layers.2.self_attn.q_norm": false, "language_model.model.layers.2.self_attn.k_norm": false, "language_model.model.layers.2.self_attn.rope": false, "language_model.model.layers.2.mlp.gate_proj": true, "language_model.model.layers.2.mlp.down_proj": true, "language_model.model.layers.2.mlp.up_proj": true, "language_model.model.layers.2.input_layernorm": false, "language_model.model.layers.2.post_attention_layernorm": false, "language_model.model.layers.2.pre_feedforward_layernorm": false, "language_model.model.layers.2.post_feedforward_layernorm": false, "language_model.model.layers.3.self_attn.q_proj": true, "language_model.model.layers.3.self_attn.k_proj": true, "language_model.model.layers.3.self_attn.v_proj": true, "language_model.model.layers.3.self_attn.o_proj": true, "language_model.model.layers.3.self_attn.q_norm": false, "language_model.model.layers.3.self_attn.k_norm": false, "language_model.model.layers.3.self_attn.rope": false, "language_model.model.layers.3.mlp.gate_proj": true, "language_model.model.layers.3.mlp.down_proj": true, "language_model.model.layers.3.mlp.up_proj": true, "language_model.model.layers.3.input_layernorm": false, "language_model.model.layers.3.post_attention_layernorm": false, "language_model.model.layers.3.pre_feedforward_layernorm": false, "language_model.model.layers.3.post_feedforward_layernorm": false, "language_model.model.layers.4.self_attn.q_proj": true, "language_model.model.layers.4.self_attn.k_proj": true, "language_model.model.layers.4.self_attn.v_proj": true, "language_model.model.layers.4.self_attn.o_proj": true, "language_model.model.layers.4.self_attn.q_norm": false, "language_model.model.layers.4.self_attn.k_norm": false, "language_model.model.layers.4.self_attn.rope": false, "language_model.model.layers.4.mlp.gate_proj": true, "language_model.model.layers.4.mlp.down_proj": true, "language_model.model.layers.4.mlp.up_proj": true, "language_model.model.layers.4.input_layernorm": false, "language_model.model.layers.4.post_attention_layernorm": false, "language_model.model.layers.4.pre_feedforward_layernorm": false, "language_model.model.layers.4.post_feedforward_layernorm": false, "language_model.model.layers.5.self_attn.q_proj": true, "language_model.model.layers.5.self_attn.k_proj": true, "language_model.model.layers.5.self_attn.v_proj": true, "language_model.model.layers.5.self_attn.o_proj": true, "language_model.model.layers.5.self_attn.q_norm": false, "language_model.model.layers.5.self_attn.k_norm": false, "language_model.model.layers.5.self_attn.rope": false, "language_model.model.layers.5.mlp.gate_proj": true, "language_model.model.layers.5.mlp.down_proj": true, "language_model.model.layers.5.mlp.up_proj": true, "language_model.model.layers.5.input_layernorm": false, "language_model.model.layers.5.post_attention_layernorm": false, "language_model.model.layers.5.pre_feedforward_layernorm": false, "language_model.model.layers.5.post_feedforward_layernorm": false, "language_model.model.layers.6.self_attn.q_proj": true, "language_model.model.layers.6.self_attn.k_proj": true, "language_model.model.layers.6.self_attn.v_proj": true, "language_model.model.layers.6.self_attn.o_proj": true, "language_model.model.layers.6.self_attn.q_norm": false, "language_model.model.layers.6.self_attn.k_norm": false, "language_model.model.layers.6.self_attn.rope": false, "language_model.model.layers.6.mlp.gate_proj": true, "language_model.model.layers.6.mlp.down_proj": true, "language_model.model.layers.6.mlp.up_proj": true, "language_model.model.layers.6.input_layernorm": false, "language_model.model.layers.6.post_attention_layernorm": false, "language_model.model.layers.6.pre_feedforward_layernorm": false, "language_model.model.layers.6.post_feedforward_layernorm": false, "language_model.model.layers.7.self_attn.q_proj": true, "language_model.model.layers.7.self_attn.k_proj": true, "language_model.model.layers.7.self_attn.v_proj": true, "language_model.model.layers.7.self_attn.o_proj": true, "language_model.model.layers.7.self_attn.q_norm": false, "language_model.model.layers.7.self_attn.k_norm": false, "language_model.model.layers.7.self_attn.rope": false, "language_model.model.layers.7.mlp.gate_proj": true, "language_model.model.layers.7.mlp.down_proj": true, "language_model.model.layers.7.mlp.up_proj": true, "language_model.model.layers.7.input_layernorm": false, "language_model.model.layers.7.post_attention_layernorm": false, "language_model.model.layers.7.pre_feedforward_layernorm": false, "language_model.model.layers.7.post_feedforward_layernorm": false, "language_model.model.layers.8.self_attn.q_proj": true, "language_model.model.layers.8.self_attn.k_proj": true, "language_model.model.layers.8.self_attn.v_proj": true, "language_model.model.layers.8.self_attn.o_proj": true, "language_model.model.layers.8.self_attn.q_norm": false, "language_model.model.layers.8.self_attn.k_norm": false, "language_model.model.layers.8.self_attn.rope": false, "language_model.model.layers.8.mlp.gate_proj": true, "language_model.model.layers.8.mlp.down_proj": true, "language_model.model.layers.8.mlp.up_proj": true, "language_model.model.layers.8.input_layernorm": false, "language_model.model.layers.8.post_attention_layernorm": false, "language_model.model.layers.8.pre_feedforward_layernorm": false, "language_model.model.layers.8.post_feedforward_layernorm": false, "language_model.model.layers.9.self_attn.q_proj": true, "language_model.model.layers.9.self_attn.k_proj": true, "language_model.model.layers.9.self_attn.v_proj": true, "language_model.model.layers.9.self_attn.o_proj": true, "language_model.model.layers.9.self_attn.q_norm": false, "language_model.model.layers.9.self_attn.k_norm": false, "language_model.model.layers.9.self_attn.rope": false, "language_model.model.layers.9.mlp.gate_proj": true, "language_model.model.layers.9.mlp.down_proj": true, "language_model.model.layers.9.mlp.up_proj": true, "language_model.model.layers.9.input_layernorm": false, "language_model.model.layers.9.post_attention_layernorm": false, "language_model.model.layers.9.pre_feedforward_layernorm": false, "language_model.model.layers.9.post_feedforward_layernorm": false, "language_model.model.layers.10.self_attn.q_proj": true, "language_model.model.layers.10.self_attn.k_proj": true, "language_model.model.layers.10.self_attn.v_proj": true, "language_model.model.layers.10.self_attn.o_proj": true, "language_model.model.layers.10.self_attn.q_norm": false, "language_model.model.layers.10.self_attn.k_norm": false, "language_model.model.layers.10.self_attn.rope": false, "language_model.model.layers.10.mlp.gate_proj": true, "language_model.model.layers.10.mlp.down_proj": true, "language_model.model.layers.10.mlp.up_proj": true, "language_model.model.layers.10.input_layernorm": false, "language_model.model.layers.10.post_attention_layernorm": false, "language_model.model.layers.10.pre_feedforward_layernorm": false, "language_model.model.layers.10.post_feedforward_layernorm": false, "language_model.model.layers.11.self_attn.q_proj": true, "language_model.model.layers.11.self_attn.k_proj": true, "language_model.model.layers.11.self_attn.v_proj": true, "language_model.model.layers.11.self_attn.o_proj": true, "language_model.model.layers.11.self_attn.q_norm": false, "language_model.model.layers.11.self_attn.k_norm": false, "language_model.model.layers.11.self_attn.rope": false, "language_model.model.layers.11.mlp.gate_proj": true, "language_model.model.layers.11.mlp.down_proj": true, "language_model.model.layers.11.mlp.up_proj": true, "language_model.model.layers.11.input_layernorm": false, "language_model.model.layers.11.post_attention_layernorm": false, "language_model.model.layers.11.pre_feedforward_layernorm": false, "language_model.model.layers.11.post_feedforward_layernorm": false, "language_model.model.layers.12.self_attn.q_proj": true, "language_model.model.layers.12.self_attn.k_proj": true, "language_model.model.layers.12.self_attn.v_proj": true, "language_model.model.layers.12.self_attn.o_proj": true, "language_model.model.layers.12.self_attn.q_norm": false, "language_model.model.layers.12.self_attn.k_norm": false, "language_model.model.layers.12.self_attn.rope": false, "language_model.model.layers.12.mlp.gate_proj": true, "language_model.model.layers.12.mlp.down_proj": true, "language_model.model.layers.12.mlp.up_proj": true, "language_model.model.layers.12.input_layernorm": false, "language_model.model.layers.12.post_attention_layernorm": false, "language_model.model.layers.12.pre_feedforward_layernorm": false, "language_model.model.layers.12.post_feedforward_layernorm": false, "language_model.model.layers.13.self_attn.q_proj": true, "language_model.model.layers.13.self_attn.k_proj": true, "language_model.model.layers.13.self_attn.v_proj": true, "language_model.model.layers.13.self_attn.o_proj": true, "language_model.model.layers.13.self_attn.q_norm": false, "language_model.model.layers.13.self_attn.k_norm": false, "language_model.model.layers.13.self_attn.rope": false, "language_model.model.layers.13.mlp.gate_proj": true, "language_model.model.layers.13.mlp.down_proj": true, "language_model.model.layers.13.mlp.up_proj": true, "language_model.model.layers.13.input_layernorm": false, "language_model.model.layers.13.post_attention_layernorm": false, "language_model.model.layers.13.pre_feedforward_layernorm": false, "language_model.model.layers.13.post_feedforward_layernorm": false, "language_model.model.layers.14.self_attn.q_proj": true, "language_model.model.layers.14.self_attn.k_proj": true, "language_model.model.layers.14.self_attn.v_proj": true, "language_model.model.layers.14.self_attn.o_proj": true, "language_model.model.layers.14.self_attn.q_norm": false, "language_model.model.layers.14.self_attn.k_norm": false, "language_model.model.layers.14.self_attn.rope": false, "language_model.model.layers.14.mlp.gate_proj": true, "language_model.model.layers.14.mlp.down_proj": true, "language_model.model.layers.14.mlp.up_proj": true, "language_model.model.layers.14.input_layernorm": false, "language_model.model.layers.14.post_attention_layernorm": false, "language_model.model.layers.14.pre_feedforward_layernorm": false, "language_model.model.layers.14.post_feedforward_layernorm": false, "language_model.model.layers.15.self_attn.q_proj": true, "language_model.model.layers.15.self_attn.k_proj": true, "language_model.model.layers.15.self_attn.v_proj": true, "language_model.model.layers.15.self_attn.o_proj": true, "language_model.model.layers.15.self_attn.q_norm": false, "language_model.model.layers.15.self_attn.k_norm": false, "language_model.model.layers.15.self_attn.rope": false, "language_model.model.layers.15.mlp.gate_proj": true, "language_model.model.layers.15.mlp.down_proj": true, "language_model.model.layers.15.mlp.up_proj": true, "language_model.model.layers.15.input_layernorm": false, "language_model.model.layers.15.post_attention_layernorm": false, "language_model.model.layers.15.pre_feedforward_layernorm": false, "language_model.model.layers.15.post_feedforward_layernorm": false, "language_model.model.layers.16.self_attn.q_proj": true, "language_model.model.layers.16.self_attn.k_proj": true, "language_model.model.layers.16.self_attn.v_proj": true, "language_model.model.layers.16.self_attn.o_proj": true, "language_model.model.layers.16.self_attn.q_norm": false, "language_model.model.layers.16.self_attn.k_norm": false, "language_model.model.layers.16.self_attn.rope": false, "language_model.model.layers.16.mlp.gate_proj": true, "language_model.model.layers.16.mlp.down_proj": true, "language_model.model.layers.16.mlp.up_proj": true, "language_model.model.layers.16.input_layernorm": false, "language_model.model.layers.16.post_attention_layernorm": false, "language_model.model.layers.16.pre_feedforward_layernorm": false, "language_model.model.layers.16.post_feedforward_layernorm": false, "language_model.model.layers.17.self_attn.q_proj": true, "language_model.model.layers.17.self_attn.k_proj": true, "language_model.model.layers.17.self_attn.v_proj": true, "language_model.model.layers.17.self_attn.o_proj": true, "language_model.model.layers.17.self_attn.q_norm": false, "language_model.model.layers.17.self_attn.k_norm": false, "language_model.model.layers.17.self_attn.rope": false, "language_model.model.layers.17.mlp.gate_proj": true, "language_model.model.layers.17.mlp.down_proj": true, "language_model.model.layers.17.mlp.up_proj": true, "language_model.model.layers.17.input_layernorm": false, "language_model.model.layers.17.post_attention_layernorm": false, "language_model.model.layers.17.pre_feedforward_layernorm": false, "language_model.model.layers.17.post_feedforward_layernorm": false, "language_model.model.layers.18.self_attn.q_proj": true, "language_model.model.layers.18.self_attn.k_proj": true, "language_model.model.layers.18.self_attn.v_proj": true, "language_model.model.layers.18.self_attn.o_proj": true, "language_model.model.layers.18.self_attn.q_norm": false, "language_model.model.layers.18.self_attn.k_norm": false, "language_model.model.layers.18.self_attn.rope": false, "language_model.model.layers.18.mlp.gate_proj": true, "language_model.model.layers.18.mlp.down_proj": true, "language_model.model.layers.18.mlp.up_proj": true, "language_model.model.layers.18.input_layernorm": false, "language_model.model.layers.18.post_attention_layernorm": false, "language_model.model.layers.18.pre_feedforward_layernorm": false, "language_model.model.layers.18.post_feedforward_layernorm": false, "language_model.model.layers.19.self_attn.q_proj": true, "language_model.model.layers.19.self_attn.k_proj": true, "language_model.model.layers.19.self_attn.v_proj": true, "language_model.model.layers.19.self_attn.o_proj": true, "language_model.model.layers.19.self_attn.q_norm": false, "language_model.model.layers.19.self_attn.k_norm": false, "language_model.model.layers.19.self_attn.rope": false, "language_model.model.layers.19.mlp.gate_proj": true, "language_model.model.layers.19.mlp.down_proj": true, "language_model.model.layers.19.mlp.up_proj": true, "language_model.model.layers.19.input_layernorm": false, "language_model.model.layers.19.post_attention_layernorm": false, "language_model.model.layers.19.pre_feedforward_layernorm": false, "language_model.model.layers.19.post_feedforward_layernorm": false, "language_model.model.layers.20.self_attn.q_proj": true, "language_model.model.layers.20.self_attn.k_proj": true, "language_model.model.layers.20.self_attn.v_proj": true, "language_model.model.layers.20.self_attn.o_proj": true, "language_model.model.layers.20.self_attn.q_norm": false, "language_model.model.layers.20.self_attn.k_norm": false, "language_model.model.layers.20.self_attn.rope": false, "language_model.model.layers.20.mlp.gate_proj": true, "language_model.model.layers.20.mlp.down_proj": true, "language_model.model.layers.20.mlp.up_proj": true, "language_model.model.layers.20.input_layernorm": false, "language_model.model.layers.20.post_attention_layernorm": false, "language_model.model.layers.20.pre_feedforward_layernorm": false, "language_model.model.layers.20.post_feedforward_layernorm": false, "language_model.model.layers.21.self_attn.q_proj": true, "language_model.model.layers.21.self_attn.k_proj": true, "language_model.model.layers.21.self_attn.v_proj": true, "language_model.model.layers.21.self_attn.o_proj": true, "language_model.model.layers.21.self_attn.q_norm": false, "language_model.model.layers.21.self_attn.k_norm": false, "language_model.model.layers.21.self_attn.rope": false, "language_model.model.layers.21.mlp.gate_proj": true, "language_model.model.layers.21.mlp.down_proj": true, "language_model.model.layers.21.mlp.up_proj": true, "language_model.model.layers.21.input_layernorm": false, "language_model.model.layers.21.post_attention_layernorm": false, "language_model.model.layers.21.pre_feedforward_layernorm": false, "language_model.model.layers.21.post_feedforward_layernorm": false, "language_model.model.layers.22.self_attn.q_proj": true, "language_model.model.layers.22.self_attn.k_proj": true, "language_model.model.layers.22.self_attn.v_proj": true, "language_model.model.layers.22.self_attn.o_proj": true, "language_model.model.layers.22.self_attn.q_norm": false, "language_model.model.layers.22.self_attn.k_norm": false, "language_model.model.layers.22.self_attn.rope": false, "language_model.model.layers.22.mlp.gate_proj": true, "language_model.model.layers.22.mlp.down_proj": true, "language_model.model.layers.22.mlp.up_proj": true, "language_model.model.layers.22.input_layernorm": false, "language_model.model.layers.22.post_attention_layernorm": false, "language_model.model.layers.22.pre_feedforward_layernorm": false, "language_model.model.layers.22.post_feedforward_layernorm": false, "language_model.model.layers.23.self_attn.q_proj": true, "language_model.model.layers.23.self_attn.k_proj": true, "language_model.model.layers.23.self_attn.v_proj": true, "language_model.model.layers.23.self_attn.o_proj": true, "language_model.model.layers.23.self_attn.q_norm": false, "language_model.model.layers.23.self_attn.k_norm": false, "language_model.model.layers.23.self_attn.rope": false, "language_model.model.layers.23.mlp.gate_proj": true, "language_model.model.layers.23.mlp.down_proj": true, "language_model.model.layers.23.mlp.up_proj": true, "language_model.model.layers.23.input_layernorm": false, "language_model.model.layers.23.post_attention_layernorm": false, "language_model.model.layers.23.pre_feedforward_layernorm": false, "language_model.model.layers.23.post_feedforward_layernorm": false, "language_model.model.layers.24.self_attn.q_proj": true, "language_model.model.layers.24.self_attn.k_proj": true, "language_model.model.layers.24.self_attn.v_proj": true, "language_model.model.layers.24.self_attn.o_proj": true, "language_model.model.layers.24.self_attn.q_norm": false, "language_model.model.layers.24.self_attn.k_norm": false, "language_model.model.layers.24.self_attn.rope": false, "language_model.model.layers.24.mlp.gate_proj": true, "language_model.model.layers.24.mlp.down_proj": true, "language_model.model.layers.24.mlp.up_proj": true, "language_model.model.layers.24.input_layernorm": false, "language_model.model.layers.24.post_attention_layernorm": false, "language_model.model.layers.24.pre_feedforward_layernorm": false, "language_model.model.layers.24.post_feedforward_layernorm": false, "language_model.model.layers.25.self_attn.q_proj": true, "language_model.model.layers.25.self_attn.k_proj": true, "language_model.model.layers.25.self_attn.v_proj": true, "language_model.model.layers.25.self_attn.o_proj": true, "language_model.model.layers.25.self_attn.q_norm": false, "language_model.model.layers.25.self_attn.k_norm": false, "language_model.model.layers.25.self_attn.rope": false, "language_model.model.layers.25.mlp.gate_proj": true, "language_model.model.layers.25.mlp.down_proj": true, "language_model.model.layers.25.mlp.up_proj": true, "language_model.model.layers.25.input_layernorm": false, "language_model.model.layers.25.post_attention_layernorm": false, "language_model.model.layers.25.pre_feedforward_layernorm": false, "language_model.model.layers.25.post_feedforward_layernorm": false, "language_model.model.layers.26.self_attn.q_proj": true, "language_model.model.layers.26.self_attn.k_proj": true, "language_model.model.layers.26.self_attn.v_proj": true, "language_model.model.layers.26.self_attn.o_proj": true, "language_model.model.layers.26.self_attn.q_norm": false, "language_model.model.layers.26.self_attn.k_norm": false, "language_model.model.layers.26.self_attn.rope": false, "language_model.model.layers.26.mlp.gate_proj": true, "language_model.model.layers.26.mlp.down_proj": true, "language_model.model.layers.26.mlp.up_proj": true, "language_model.model.layers.26.input_layernorm": false, "language_model.model.layers.26.post_attention_layernorm": false, "language_model.model.layers.26.pre_feedforward_layernorm": false, "language_model.model.layers.26.post_feedforward_layernorm": false, "language_model.model.layers.27.self_attn.q_proj": true, "language_model.model.layers.27.self_attn.k_proj": true, "language_model.model.layers.27.self_attn.v_proj": true, "language_model.model.layers.27.self_attn.o_proj": true, "language_model.model.layers.27.self_attn.q_norm": false, "language_model.model.layers.27.self_attn.k_norm": false, "language_model.model.layers.27.self_attn.rope": false, "language_model.model.layers.27.mlp.gate_proj": true, "language_model.model.layers.27.mlp.down_proj": true, "language_model.model.layers.27.mlp.up_proj": true, "language_model.model.layers.27.input_layernorm": false, "language_model.model.layers.27.post_attention_layernorm": false, "language_model.model.layers.27.pre_feedforward_layernorm": false, "language_model.model.layers.27.post_feedforward_layernorm": false, "language_model.model.layers.28.self_attn.q_proj": true, "language_model.model.layers.28.self_attn.k_proj": true, "language_model.model.layers.28.self_attn.v_proj": true, "language_model.model.layers.28.self_attn.o_proj": true, "language_model.model.layers.28.self_attn.q_norm": false, "language_model.model.layers.28.self_attn.k_norm": false, "language_model.model.layers.28.self_attn.rope": false, "language_model.model.layers.28.mlp.gate_proj": true, "language_model.model.layers.28.mlp.down_proj": true, "language_model.model.layers.28.mlp.up_proj": true, "language_model.model.layers.28.input_layernorm": false, "language_model.model.layers.28.post_attention_layernorm": false, "language_model.model.layers.28.pre_feedforward_layernorm": false, "language_model.model.layers.28.post_feedforward_layernorm": false, "language_model.model.layers.29.self_attn.q_proj": true, "language_model.model.layers.29.self_attn.k_proj": true, "language_model.model.layers.29.self_attn.v_proj": true, "language_model.model.layers.29.self_attn.o_proj": true, "language_model.model.layers.29.self_attn.q_norm": false, "language_model.model.layers.29.self_attn.k_norm": false, "language_model.model.layers.29.self_attn.rope": false, "language_model.model.layers.29.mlp.gate_proj": true, "language_model.model.layers.29.mlp.down_proj": true, "language_model.model.layers.29.mlp.up_proj": true, "language_model.model.layers.29.input_layernorm": false, "language_model.model.layers.29.post_attention_layernorm": false, "language_model.model.layers.29.pre_feedforward_layernorm": false, "language_model.model.layers.29.post_feedforward_layernorm": false, "language_model.model.layers.30.self_attn.q_proj": true, "language_model.model.layers.30.self_attn.k_proj": true, "language_model.model.layers.30.self_attn.v_proj": true, "language_model.model.layers.30.self_attn.o_proj": true, "language_model.model.layers.30.self_attn.q_norm": false, "language_model.model.layers.30.self_attn.k_norm": false, "language_model.model.layers.30.self_attn.rope": false, "language_model.model.layers.30.mlp.gate_proj": true, "language_model.model.layers.30.mlp.down_proj": true, "language_model.model.layers.30.mlp.up_proj": true, "language_model.model.layers.30.input_layernorm": false, "language_model.model.layers.30.post_attention_layernorm": false, "language_model.model.layers.30.pre_feedforward_layernorm": false, "language_model.model.layers.30.post_feedforward_layernorm": false, "language_model.model.layers.31.self_attn.q_proj": true, "language_model.model.layers.31.self_attn.k_proj": true, "language_model.model.layers.31.self_attn.v_proj": true, "language_model.model.layers.31.self_attn.o_proj": true, "language_model.model.layers.31.self_attn.q_norm": false, "language_model.model.layers.31.self_attn.k_norm": false, "language_model.model.layers.31.self_attn.rope": false, "language_model.model.layers.31.mlp.gate_proj": true, "language_model.model.layers.31.mlp.down_proj": true, "language_model.model.layers.31.mlp.up_proj": true, "language_model.model.layers.31.input_layernorm": false, "language_model.model.layers.31.post_attention_layernorm": false, "language_model.model.layers.31.pre_feedforward_layernorm": false, "language_model.model.layers.31.post_feedforward_layernorm": false, "language_model.model.layers.32.self_attn.q_proj": true, "language_model.model.layers.32.self_attn.k_proj": true, "language_model.model.layers.32.self_attn.v_proj": true, "language_model.model.layers.32.self_attn.o_proj": true, "language_model.model.layers.32.self_attn.q_norm": false, "language_model.model.layers.32.self_attn.k_norm": false, "language_model.model.layers.32.self_attn.rope": false, "language_model.model.layers.32.mlp.gate_proj": true, "language_model.model.layers.32.mlp.down_proj": true, "language_model.model.layers.32.mlp.up_proj": true, "language_model.model.layers.32.input_layernorm": false, "language_model.model.layers.32.post_attention_layernorm": false, "language_model.model.layers.32.pre_feedforward_layernorm": false, "language_model.model.layers.32.post_feedforward_layernorm": false, "language_model.model.layers.33.self_attn.q_proj": true, "language_model.model.layers.33.self_attn.k_proj": true, "language_model.model.layers.33.self_attn.v_proj": true, "language_model.model.layers.33.self_attn.o_proj": true, "language_model.model.layers.33.self_attn.q_norm": false, "language_model.model.layers.33.self_attn.k_norm": false, "language_model.model.layers.33.self_attn.rope": false, "language_model.model.layers.33.mlp.gate_proj": true, "language_model.model.layers.33.mlp.down_proj": true, "language_model.model.layers.33.mlp.up_proj": true, "language_model.model.layers.33.input_layernorm": false, "language_model.model.layers.33.post_attention_layernorm": false, "language_model.model.layers.33.pre_feedforward_layernorm": false, "language_model.model.layers.33.post_feedforward_layernorm": false, "language_model.model.norm": false, "language_model.lm_head": true }, "text_config": { "_sliding_window_pattern": 6, "attention_bias": false, "attention_dropout": 0.0, "attn_logit_softcapping": null, "cache_implementation": "hybrid", "final_logit_softcapping": null, "head_dim": 256, "hidden_activation": "gelu_pytorch_tanh", "hidden_size": 2560, "initializer_range": 0.02, "intermediate_size": 10240, "layer_types": [ "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention" ], "max_position_embeddings": 131072, "model_type": "gemma3_text", "num_attention_heads": 8, "num_hidden_layers": 34, "num_key_value_heads": 4, "query_pre_attn_scalar": 256, "rms_norm_eps": 1e-06, "rope_local_base_freq": 10000, "rope_scaling": { "factor": 8.0, "rope_type": "linear" }, "rope_theta": 1000000, "sliding_window": 1024, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 262208 }, "torch_dtype": "bfloat16", "transformers_version": "4.53.3", "unsloth_version": "2025.7.7" }