{ "base_model": "Qwen/Qwen3.5-9B", "format": "dashq-packed-linear", "format_version": 1, "method": "dashq", "model_class": "image_text_to_text", "params": { "bits": 4, "group_size": 32, "low_memory_optimization": false, "moe_hessian_scope": "shared", "n_samples": 128, "scale_zero_dtype": "float16", "symmetric": false, "use_error_compensation": true, "use_optimal_shrinkage": true, "use_weighted_quantization": true }, "quantized_modules": { "model.language_model.layers.0.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.0.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.0.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.0.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.0.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.0.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.0.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.0.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.1.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.1.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.1.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.1.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.1.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.1.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.1.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.1.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.10.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.10.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.10.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.10.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.10.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.10.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.10.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.10.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.11.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.11.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.11.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.11.self_attn.k_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.11.self_attn.o_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.11.self_attn.q_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.11.self_attn.v_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.12.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.12.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.12.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.12.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.12.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.12.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.12.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.12.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.13.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.13.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.13.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.13.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.13.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.13.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.13.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.13.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.14.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.14.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.14.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.14.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.14.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.14.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.14.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.14.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.15.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.15.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.15.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.15.self_attn.k_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.15.self_attn.o_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.15.self_attn.q_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.15.self_attn.v_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.16.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.16.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.16.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.16.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.16.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.16.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.16.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.16.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.17.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.17.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.17.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.17.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.17.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.17.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.17.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.17.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.18.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.18.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.18.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.18.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.18.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.18.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.18.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.18.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.19.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.19.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.19.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.19.self_attn.k_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.19.self_attn.o_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.19.self_attn.q_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.19.self_attn.v_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.2.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.2.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.2.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.2.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.2.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.2.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.2.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.2.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.20.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.20.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.20.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.20.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.20.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.20.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.20.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.20.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.21.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.21.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.21.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.21.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.21.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.21.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.21.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.21.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.22.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.22.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.22.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.22.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.22.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.22.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.22.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.22.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.23.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.23.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.23.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.23.self_attn.k_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.23.self_attn.o_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.23.self_attn.q_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.23.self_attn.v_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.24.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.24.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.24.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.24.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.24.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.24.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.24.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.24.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.25.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.25.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.25.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.25.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.25.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.25.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.25.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.25.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.26.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.26.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.26.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.26.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.26.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.26.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.26.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.26.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.27.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.27.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.27.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.27.self_attn.k_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.27.self_attn.o_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.27.self_attn.q_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.27.self_attn.v_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.28.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.28.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.28.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.28.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.28.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.28.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.28.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.28.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.29.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.29.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.29.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.29.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.29.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.29.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.29.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.29.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.3.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.3.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.3.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.3.self_attn.k_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.3.self_attn.o_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.3.self_attn.q_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.3.self_attn.v_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.30.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.30.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.30.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.30.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.30.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.30.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.30.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.30.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.31.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.31.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.31.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.31.self_attn.k_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.31.self_attn.o_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.31.self_attn.q_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.31.self_attn.v_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.4.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.4.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.4.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.4.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.4.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.4.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.4.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.4.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.5.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.5.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.5.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.5.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.5.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.5.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.5.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.5.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.6.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.6.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.6.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.6.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.6.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.6.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.6.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.6.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.7.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.7.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.7.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.7.self_attn.k_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.7.self_attn.o_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.7.self_attn.q_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.7.self_attn.v_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 131072, "out_features": 1024, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.8.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.8.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.8.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.8.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.8.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.8.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.8.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.8.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.9.linear_attn.in_proj_a": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.9.linear_attn.in_proj_b": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 4096, "out_features": 32, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.9.linear_attn.in_proj_qkv": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1048576, "out_features": 8192, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.9.linear_attn.in_proj_z": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.9.linear_attn.out_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 524288, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.9.mlp.down_proj": { "group_size": 32, "in_features": 12288, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 4096, "packing": "int4_packed_u32", "quant_in_features": 12288, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.9.mlp.gate_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" }, "model.language_model.layers.9.mlp.up_proj": { "group_size": 32, "in_features": 4096, "linear_dtype": "bfloat16", "nbits": 4, "num_groups": 1572864, "out_features": 12288, "packing": "int4_packed_u32", "quant_in_features": 4096, "runtime_backend": "torch", "scale_zero_dtype": "float16" } }, "results": { "Method": "dashq", "Model": "Qwen/Qwen3.5-9B", "ModelSizeGB": 9.306842152, "OriginalSizeGB": 19.30631088, "PPL": 8.678437232971191, "Params": "{'bits': 4, 'group_size': 32, 'scale_zero_dtype': 'float16', 'n_samples': 128, 'moe_hessian_scope': 'shared', 'use_error_compensation': True, 'use_optimal_shrinkage': True, 'use_weighted_quantization': True, 'symmetric': False, 'low_memory_optimization': False}", "QuantTime": 374.64265871047974, "arc_challenge": 55.71672354948806, "arc_easy": 74.66329966329967, "commonsense_qa": 82.96478296478297, "gsm8k_cot": 89.23426838514025, "hellaswag": 77.63393746265685, "lambada_openai": 70.0950902386959, "mmlu": 78.45748468879077, "openbookqa": 42.0, "piqa": 80.14145810663766, "truthfulqa_mc2": 53.08014480902932, "winogrande": 71.42857142857143, "zeroshot_avg": 67.52488980257354 } }