{ "transformers_version": "5.14.0", "architectures": [ "SmolLM3ForCausalLM" ], "output_hidden_states": false, "return_dict": true, "dtype": "bfloat16", "chunk_size_feed_forward": 0, "is_encoder_decoder": false, "id2label": { "0": "LABEL_0", "1": "LABEL_1" }, "label2id": { "LABEL_0": 0, "LABEL_1": 1 }, "problem_type": null, "vocab_size": 128256, "hidden_size": 2048, "intermediate_size": 11008, "num_hidden_layers": 36, "num_attention_heads": 16, "num_key_value_heads": 4, "hidden_act": "silu", "max_position_embeddings": 65536, "initializer_range": 0.02, "rms_norm_eps": 1e-06, "use_cache": true, "pad_token_id": 128004, "bos_token_id": 128000, "eos_token_id": 128012, "rope_parameters": { "rope_theta": 5000000.0, "rope_type": "default" }, "use_sliding_window": false, "sliding_window": null, "no_rope_layers": [ 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0 ], "no_rope_layer_interval": 4, "layer_types": [ "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention" ], "attention_bias": false, "attention_dropout": 0.0, "mlp_bias": false, "tie_word_embeddings": true, "_name_or_path": "HuggingFaceTB/SmolLM3-3B", "max_window_layers": 28, "model_type": "smollm3", "pretraining_tp": 2, "output_attentions": false, "quantization_config": { "quant_method": "glq", "codebook": "trellis", "block_diagonal": true, "codesz": 16, "bpw": 3, "layer_bpw": { "model.layers.0.self_attn.q_proj": 3, "model.layers.0.self_attn.k_proj": 3, "model.layers.0.self_attn.v_proj": 3, "model.layers.0.self_attn.o_proj": 3, "model.layers.0.mlp.gate_proj": 3, "model.layers.0.mlp.up_proj": 3, "model.layers.0.mlp.down_proj": 3, "model.layers.1.self_attn.q_proj": 3, "model.layers.1.self_attn.k_proj": 3, "model.layers.1.self_attn.v_proj": 3, "model.layers.1.self_attn.o_proj": 3, "model.layers.1.mlp.gate_proj": 3, "model.layers.1.mlp.up_proj": 3, "model.layers.1.mlp.down_proj": 3, "model.layers.2.self_attn.q_proj": 3, "model.layers.2.self_attn.k_proj": 3, "model.layers.2.self_attn.v_proj": 3, "model.layers.2.self_attn.o_proj": 3, "model.layers.2.mlp.gate_proj": 3, "model.layers.2.mlp.up_proj": 3, "model.layers.2.mlp.down_proj": 3, "model.layers.3.self_attn.q_proj": 3, "model.layers.3.self_attn.k_proj": 3, "model.layers.3.self_attn.v_proj": 3, "model.layers.3.self_attn.o_proj": 3, "model.layers.3.mlp.gate_proj": 3, "model.layers.3.mlp.up_proj": 3, "model.layers.3.mlp.down_proj": 3, "model.layers.4.self_attn.q_proj": 3, "model.layers.4.self_attn.k_proj": 3, "model.layers.4.self_attn.v_proj": 3, "model.layers.4.self_attn.o_proj": 3, "model.layers.4.mlp.gate_proj": 3, "model.layers.4.mlp.up_proj": 3, "model.layers.4.mlp.down_proj": 3, "model.layers.5.self_attn.q_proj": 3, "model.layers.5.self_attn.k_proj": 3, "model.layers.5.self_attn.v_proj": 3, "model.layers.5.self_attn.o_proj": 3, "model.layers.5.mlp.gate_proj": 3, "model.layers.5.mlp.up_proj": 3, "model.layers.5.mlp.down_proj": 3, "model.layers.6.self_attn.q_proj": 3, "model.layers.6.self_attn.k_proj": 3, "model.layers.6.self_attn.v_proj": 3, "model.layers.6.self_attn.o_proj": 3, "model.layers.6.mlp.gate_proj": 3, "model.layers.6.mlp.up_proj": 3, "model.layers.6.mlp.down_proj": 3, "model.layers.7.self_attn.q_proj": 3, "model.layers.7.self_attn.k_proj": 3, "model.layers.7.self_attn.v_proj": 3, "model.layers.7.self_attn.o_proj": 3, "model.layers.7.mlp.gate_proj": 3, "model.layers.7.mlp.up_proj": 3, "model.layers.7.mlp.down_proj": 3, "model.layers.8.self_attn.q_proj": 3, "model.layers.8.self_attn.k_proj": 3, "model.layers.8.self_attn.v_proj": 3, "model.layers.8.self_attn.o_proj": 3, "model.layers.8.mlp.gate_proj": 3, "model.layers.8.mlp.up_proj": 3, "model.layers.8.mlp.down_proj": 3, "model.layers.9.self_attn.q_proj": 3, "model.layers.9.self_attn.k_proj": 3, "model.layers.9.self_attn.v_proj": 3, "model.layers.9.self_attn.o_proj": 3, "model.layers.9.mlp.gate_proj": 3, "model.layers.9.mlp.up_proj": 3, "model.layers.9.mlp.down_proj": 3, "model.layers.10.self_attn.q_proj": 3, "model.layers.10.self_attn.k_proj": 3, "model.layers.10.self_attn.v_proj": 3, "model.layers.10.self_attn.o_proj": 3, "model.layers.10.mlp.gate_proj": 3, "model.layers.10.mlp.up_proj": 3, "model.layers.10.mlp.down_proj": 3, "model.layers.11.self_attn.q_proj": 3, "model.layers.11.self_attn.k_proj": 3, "model.layers.11.self_attn.v_proj": 3, "model.layers.11.self_attn.o_proj": 3, "model.layers.11.mlp.gate_proj": 3, "model.layers.11.mlp.up_proj": 3, "model.layers.11.mlp.down_proj": 3, "model.layers.12.self_attn.q_proj": 3, "model.layers.12.self_attn.k_proj": 3, "model.layers.12.self_attn.v_proj": 3, "model.layers.12.self_attn.o_proj": 3, "model.layers.12.mlp.gate_proj": 3, "model.layers.12.mlp.up_proj": 3, "model.layers.12.mlp.down_proj": 3, "model.layers.13.self_attn.q_proj": 3, "model.layers.13.self_attn.k_proj": 3, "model.layers.13.self_attn.v_proj": 3, "model.layers.13.self_attn.o_proj": 3, "model.layers.13.mlp.gate_proj": 3, "model.layers.13.mlp.up_proj": 3, "model.layers.13.mlp.down_proj": 3, "model.layers.14.self_attn.q_proj": 3, "model.layers.14.self_attn.k_proj": 3, "model.layers.14.self_attn.v_proj": 3, "model.layers.14.self_attn.o_proj": 3, "model.layers.14.mlp.gate_proj": 3, "model.layers.14.mlp.up_proj": 3, "model.layers.14.mlp.down_proj": 3, "model.layers.15.self_attn.q_proj": 3, "model.layers.15.self_attn.k_proj": 3, "model.layers.15.self_attn.v_proj": 3, "model.layers.15.self_attn.o_proj": 3, "model.layers.15.mlp.gate_proj": 3, "model.layers.15.mlp.up_proj": 3, "model.layers.15.mlp.down_proj": 3, "model.layers.16.self_attn.q_proj": 3, "model.layers.16.self_attn.k_proj": 3, "model.layers.16.self_attn.v_proj": 3, "model.layers.16.self_attn.o_proj": 3, "model.layers.16.mlp.gate_proj": 3, "model.layers.16.mlp.up_proj": 3, "model.layers.16.mlp.down_proj": 3, "model.layers.17.self_attn.q_proj": 3, "model.layers.17.self_attn.k_proj": 3, "model.layers.17.self_attn.v_proj": 3, "model.layers.17.self_attn.o_proj": 3, "model.layers.17.mlp.gate_proj": 3, "model.layers.17.mlp.up_proj": 3, "model.layers.17.mlp.down_proj": 3, "model.layers.18.self_attn.q_proj": 3, "model.layers.18.self_attn.k_proj": 3, "model.layers.18.self_attn.v_proj": 3, "model.layers.18.self_attn.o_proj": 3, "model.layers.18.mlp.gate_proj": 3, "model.layers.18.mlp.up_proj": 3, "model.layers.18.mlp.down_proj": 3, "model.layers.19.self_attn.q_proj": 3, "model.layers.19.self_attn.k_proj": 3, "model.layers.19.self_attn.v_proj": 3, "model.layers.19.self_attn.o_proj": 3, "model.layers.19.mlp.gate_proj": 3, "model.layers.19.mlp.up_proj": 3, "model.layers.19.mlp.down_proj": 3, "model.layers.20.self_attn.q_proj": 3, "model.layers.20.self_attn.k_proj": 3, "model.layers.20.self_attn.v_proj": 3, "model.layers.20.self_attn.o_proj": 3, "model.layers.20.mlp.gate_proj": 3, "model.layers.20.mlp.up_proj": 3, "model.layers.20.mlp.down_proj": 3, "model.layers.21.self_attn.q_proj": 3, "model.layers.21.self_attn.k_proj": 3, "model.layers.21.self_attn.v_proj": 3, "model.layers.21.self_attn.o_proj": 3, "model.layers.21.mlp.gate_proj": 3, "model.layers.21.mlp.up_proj": 3, "model.layers.21.mlp.down_proj": 3, "model.layers.22.self_attn.q_proj": 3, "model.layers.22.self_attn.k_proj": 3, "model.layers.22.self_attn.v_proj": 3, "model.layers.22.self_attn.o_proj": 3, "model.layers.22.mlp.gate_proj": 3, "model.layers.22.mlp.up_proj": 3, "model.layers.22.mlp.down_proj": 3, "model.layers.23.self_attn.q_proj": 3, "model.layers.23.self_attn.k_proj": 3, "model.layers.23.self_attn.v_proj": 3, "model.layers.23.self_attn.o_proj": 3, "model.layers.23.mlp.gate_proj": 3, "model.layers.23.mlp.up_proj": 3, "model.layers.23.mlp.down_proj": 3, "model.layers.24.self_attn.q_proj": 3, "model.layers.24.self_attn.k_proj": 3, "model.layers.24.self_attn.v_proj": 3, "model.layers.24.self_attn.o_proj": 3, "model.layers.24.mlp.gate_proj": 3, "model.layers.24.mlp.up_proj": 3, "model.layers.24.mlp.down_proj": 3, "model.layers.25.self_attn.q_proj": 3, "model.layers.25.self_attn.k_proj": 3, "model.layers.25.self_attn.v_proj": 3, "model.layers.25.self_attn.o_proj": 3, "model.layers.25.mlp.gate_proj": 3, "model.layers.25.mlp.up_proj": 3, "model.layers.25.mlp.down_proj": 3, "model.layers.26.self_attn.q_proj": 3, "model.layers.26.self_attn.k_proj": 3, "model.layers.26.self_attn.v_proj": 3, "model.layers.26.self_attn.o_proj": 3, "model.layers.26.mlp.gate_proj": 3, "model.layers.26.mlp.up_proj": 3, "model.layers.26.mlp.down_proj": 3, "model.layers.27.self_attn.q_proj": 3, "model.layers.27.self_attn.k_proj": 3, "model.layers.27.self_attn.v_proj": 3, "model.layers.27.self_attn.o_proj": 3, "model.layers.27.mlp.gate_proj": 3, "model.layers.27.mlp.up_proj": 3, "model.layers.27.mlp.down_proj": 3, "model.layers.28.self_attn.q_proj": 3, "model.layers.28.self_attn.k_proj": 3, "model.layers.28.self_attn.v_proj": 3, "model.layers.28.self_attn.o_proj": 3, "model.layers.28.mlp.gate_proj": 3, "model.layers.28.mlp.up_proj": 3, "model.layers.28.mlp.down_proj": 3, "model.layers.29.self_attn.q_proj": 3, "model.layers.29.self_attn.k_proj": 3, "model.layers.29.self_attn.v_proj": 3, "model.layers.29.self_attn.o_proj": 3, "model.layers.29.mlp.gate_proj": 3, "model.layers.29.mlp.up_proj": 3, "model.layers.29.mlp.down_proj": 3, "model.layers.30.self_attn.q_proj": 3, "model.layers.30.self_attn.k_proj": 3, "model.layers.30.self_attn.v_proj": 3, "model.layers.30.self_attn.o_proj": 3, "model.layers.30.mlp.gate_proj": 3, "model.layers.30.mlp.up_proj": 3, "model.layers.30.mlp.down_proj": 3, "model.layers.31.self_attn.q_proj": 3, "model.layers.31.self_attn.k_proj": 3, "model.layers.31.self_attn.v_proj": 3, "model.layers.31.self_attn.o_proj": 3, "model.layers.31.mlp.gate_proj": 3, "model.layers.31.mlp.up_proj": 3, "model.layers.31.mlp.down_proj": 3, "model.layers.32.self_attn.q_proj": 3, "model.layers.32.self_attn.k_proj": 3, "model.layers.32.self_attn.v_proj": 3, "model.layers.32.self_attn.o_proj": 3, "model.layers.32.mlp.gate_proj": 3, "model.layers.32.mlp.up_proj": 3, "model.layers.32.mlp.down_proj": 3, "model.layers.33.self_attn.q_proj": 3, "model.layers.33.self_attn.k_proj": 3, "model.layers.33.self_attn.v_proj": 3, "model.layers.33.self_attn.o_proj": 3, "model.layers.33.mlp.gate_proj": 3, "model.layers.33.mlp.up_proj": 3, "model.layers.33.mlp.down_proj": 3, "model.layers.34.self_attn.q_proj": 3, "model.layers.34.self_attn.k_proj": 3, "model.layers.34.self_attn.v_proj": 3, "model.layers.34.self_attn.o_proj": 3, "model.layers.34.mlp.gate_proj": 3, "model.layers.34.mlp.up_proj": 3, "model.layers.34.mlp.down_proj": 3, "model.layers.35.self_attn.q_proj": 3, "model.layers.35.self_attn.k_proj": 3, "model.layers.35.self_attn.v_proj": 3, "model.layers.35.self_attn.o_proj": 3, "model.layers.35.mlp.gate_proj": 3, "model.layers.35.mlp.up_proj": 3, "model.layers.35.mlp.down_proj": 3 }, "variant": "3inst", "trellis_layout": "kernel" } }