model.embed_tokens.weight_quantizer TensorQuantizer(disabled) model.embed_tokens.input_quantizer HardDisabledTensorQuantizer(disabled) model.embed_tokens.output_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.q_proj.input_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.q_proj.weight_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.k_proj.input_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.k_proj.weight_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.v_proj.input_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.v_proj.weight_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.o_proj.input_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.o_proj.weight_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.0.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.0.mlp.gate_proj.input_quantizer TensorQuantizer(disabled) model.layers.0.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.0.mlp.gate_proj.weight_quantizer TensorQuantizer(disabled) model.layers.0.mlp.up_proj.input_quantizer TensorQuantizer(disabled) model.layers.0.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.0.mlp.up_proj.weight_quantizer TensorQuantizer(disabled) model.layers.0.mlp.down_proj.input_quantizer TensorQuantizer(disabled) model.layers.0.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.0.mlp.down_proj.weight_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.q_proj.input_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.q_proj.weight_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.k_proj.input_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.k_proj.weight_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.v_proj.input_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.v_proj.weight_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.o_proj.input_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.o_proj.weight_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.1.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.1.mlp.gate_proj.input_quantizer TensorQuantizer(disabled) model.layers.1.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.1.mlp.gate_proj.weight_quantizer TensorQuantizer(disabled) model.layers.1.mlp.up_proj.input_quantizer TensorQuantizer(disabled) model.layers.1.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.1.mlp.up_proj.weight_quantizer TensorQuantizer(disabled) model.layers.1.mlp.down_proj.input_quantizer TensorQuantizer(disabled) model.layers.1.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.1.mlp.down_proj.weight_quantizer TensorQuantizer(disabled) model.layers.2.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.32e+01 calibrator=MaxCalibrator quant) model.layers.2.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.2.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.66e-01 calibrator=MaxCalibrator quant) model.layers.2.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.32e+01 calibrator=MaxCalibrator quant) model.layers.2.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.2.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.41e-01 calibrator=MaxCalibrator quant) model.layers.2.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.32e+01 calibrator=MaxCalibrator quant) model.layers.2.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.2.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.41e-01 calibrator=MaxCalibrator quant) model.layers.2.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.17e+00 calibrator=MaxCalibrator quant) model.layers.2.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.2.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=9.92e-01 calibrator=MaxCalibrator quant) model.layers.2.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.2.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.2.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.2.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.2.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.77e+00 calibrator=MaxCalibrator quant) model.layers.2.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.2.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.42e-01 calibrator=MaxCalibrator quant) model.layers.2.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.77e+00 calibrator=MaxCalibrator quant) model.layers.2.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.2.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.89e-01 calibrator=MaxCalibrator quant) model.layers.2.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.09e+01 calibrator=MaxCalibrator quant) model.layers.2.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.2.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.04e+00 calibrator=MaxCalibrator quant) model.layers.3.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.32e+01 calibrator=MaxCalibrator quant) model.layers.3.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.3.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.76e-01 calibrator=MaxCalibrator quant) model.layers.3.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.32e+01 calibrator=MaxCalibrator quant) model.layers.3.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.3.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.99e-01 calibrator=MaxCalibrator quant) model.layers.3.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.32e+01 calibrator=MaxCalibrator quant) model.layers.3.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.3.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.32e-01 calibrator=MaxCalibrator quant) model.layers.3.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.45e+00 calibrator=MaxCalibrator quant) model.layers.3.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.3.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.03e-01 calibrator=MaxCalibrator quant) model.layers.3.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.3.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.3.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.3.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.3.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.69e+00 calibrator=MaxCalibrator quant) model.layers.3.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.3.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.62e-01 calibrator=MaxCalibrator quant) model.layers.3.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.69e+00 calibrator=MaxCalibrator quant) model.layers.3.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.3.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.03e+00 calibrator=MaxCalibrator quant) model.layers.3.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.64e+01 calibrator=MaxCalibrator quant) model.layers.3.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.3.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=8.91e-01 calibrator=MaxCalibrator quant) model.layers.4.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.18e+01 calibrator=MaxCalibrator quant) model.layers.4.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.4.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.02e+00 calibrator=MaxCalibrator quant) model.layers.4.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.18e+01 calibrator=MaxCalibrator quant) model.layers.4.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.4.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.96e-01 calibrator=MaxCalibrator quant) model.layers.4.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.18e+01 calibrator=MaxCalibrator quant) model.layers.4.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.4.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.94e-01 calibrator=MaxCalibrator quant) model.layers.4.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.61e+00 calibrator=MaxCalibrator quant) model.layers.4.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.4.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.85e-01 calibrator=MaxCalibrator quant) model.layers.4.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.4.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.4.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.4.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.4.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.15e+01 calibrator=MaxCalibrator quant) model.layers.4.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.4.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.19e-01 calibrator=MaxCalibrator quant) model.layers.4.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.15e+01 calibrator=MaxCalibrator quant) model.layers.4.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.4.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.73e-01 calibrator=MaxCalibrator quant) model.layers.4.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.18e+03 calibrator=MaxCalibrator quant) model.layers.4.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.4.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.03e+00 calibrator=MaxCalibrator quant) model.layers.5.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.35e+01 calibrator=MaxCalibrator quant) model.layers.5.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.5.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.79e-01 calibrator=MaxCalibrator quant) model.layers.5.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.35e+01 calibrator=MaxCalibrator quant) model.layers.5.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.5.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.34e-01 calibrator=MaxCalibrator quant) model.layers.5.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.35e+01 calibrator=MaxCalibrator quant) model.layers.5.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.5.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.90e-01 calibrator=MaxCalibrator quant) model.layers.5.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.89e+00 calibrator=MaxCalibrator quant) model.layers.5.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.5.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.70e-01 calibrator=MaxCalibrator quant) model.layers.5.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.5.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.5.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.5.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.5.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.14e+01 calibrator=MaxCalibrator quant) model.layers.5.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.5.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.21e-01 calibrator=MaxCalibrator quant) model.layers.5.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.14e+01 calibrator=MaxCalibrator quant) model.layers.5.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.5.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.90e-01 calibrator=MaxCalibrator quant) model.layers.5.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.25e+01 calibrator=MaxCalibrator quant) model.layers.5.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.5.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.31e+00 calibrator=MaxCalibrator quant) model.layers.6.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.22e+01 calibrator=MaxCalibrator quant) model.layers.6.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.6.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.00e-01 calibrator=MaxCalibrator quant) model.layers.6.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.22e+01 calibrator=MaxCalibrator quant) model.layers.6.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.6.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.07e-01 calibrator=MaxCalibrator quant) model.layers.6.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.22e+01 calibrator=MaxCalibrator quant) model.layers.6.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.6.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.55e-01 calibrator=MaxCalibrator quant) model.layers.6.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.34e+00 calibrator=MaxCalibrator quant) model.layers.6.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.6.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.90e-01 calibrator=MaxCalibrator quant) model.layers.6.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.6.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.6.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.6.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.6.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.44e+01 calibrator=MaxCalibrator quant) model.layers.6.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.6.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.77e-01 calibrator=MaxCalibrator quant) model.layers.6.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.44e+01 calibrator=MaxCalibrator quant) model.layers.6.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.6.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.41e-01 calibrator=MaxCalibrator quant) model.layers.6.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.71e+01 calibrator=MaxCalibrator quant) model.layers.6.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.6.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.66e-01 calibrator=MaxCalibrator quant) model.layers.7.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.30e+01 calibrator=MaxCalibrator quant) model.layers.7.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.7.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.16e-01 calibrator=MaxCalibrator quant) model.layers.7.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.30e+01 calibrator=MaxCalibrator quant) model.layers.7.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.7.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.20e-01 calibrator=MaxCalibrator quant) model.layers.7.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.30e+01 calibrator=MaxCalibrator quant) model.layers.7.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.7.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.19e-01 calibrator=MaxCalibrator quant) model.layers.7.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.12e+00 calibrator=MaxCalibrator quant) model.layers.7.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.7.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.27e-01 calibrator=MaxCalibrator quant) model.layers.7.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.7.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.7.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.7.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.7.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.29e+01 calibrator=MaxCalibrator quant) model.layers.7.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.7.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.92e-01 calibrator=MaxCalibrator quant) model.layers.7.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.29e+01 calibrator=MaxCalibrator quant) model.layers.7.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.7.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.78e-01 calibrator=MaxCalibrator quant) model.layers.7.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.19e+01 calibrator=MaxCalibrator quant) model.layers.7.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.7.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.82e-01 calibrator=MaxCalibrator quant) model.layers.8.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.80e+01 calibrator=MaxCalibrator quant) model.layers.8.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.8.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.09e-01 calibrator=MaxCalibrator quant) model.layers.8.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.80e+01 calibrator=MaxCalibrator quant) model.layers.8.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.8.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.73e-01 calibrator=MaxCalibrator quant) model.layers.8.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.80e+01 calibrator=MaxCalibrator quant) model.layers.8.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.8.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.32e-01 calibrator=MaxCalibrator quant) model.layers.8.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=8.19e+00 calibrator=MaxCalibrator quant) model.layers.8.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.8.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.16e-01 calibrator=MaxCalibrator quant) model.layers.8.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.8.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.8.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.8.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.8.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.56e+01 calibrator=MaxCalibrator quant) model.layers.8.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.8.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=8.09e-01 calibrator=MaxCalibrator quant) model.layers.8.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.56e+01 calibrator=MaxCalibrator quant) model.layers.8.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.8.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.66e-01 calibrator=MaxCalibrator quant) model.layers.8.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.12e+01 calibrator=MaxCalibrator quant) model.layers.8.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.8.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.00e-01 calibrator=MaxCalibrator quant) model.layers.9.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.88e+01 calibrator=MaxCalibrator quant) model.layers.9.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.9.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.16e-01 calibrator=MaxCalibrator quant) model.layers.9.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.88e+01 calibrator=MaxCalibrator quant) model.layers.9.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.9.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.62e-01 calibrator=MaxCalibrator quant) model.layers.9.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.88e+01 calibrator=MaxCalibrator quant) model.layers.9.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.9.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.67e-01 calibrator=MaxCalibrator quant) model.layers.9.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.59e+00 calibrator=MaxCalibrator quant) model.layers.9.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.9.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.74e-01 calibrator=MaxCalibrator quant) model.layers.9.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.9.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.9.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.9.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.9.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.96e+01 calibrator=MaxCalibrator quant) model.layers.9.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.9.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=8.71e-01 calibrator=MaxCalibrator quant) model.layers.9.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.96e+01 calibrator=MaxCalibrator quant) model.layers.9.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.9.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=9.77e-01 calibrator=MaxCalibrator quant) model.layers.9.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.00e+01 calibrator=MaxCalibrator quant) model.layers.9.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.9.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.07e+00 calibrator=MaxCalibrator quant) model.layers.10.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.38e+01 calibrator=MaxCalibrator quant) model.layers.10.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.10.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.04e-01 calibrator=MaxCalibrator quant) model.layers.10.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.38e+01 calibrator=MaxCalibrator quant) model.layers.10.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.10.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.81e-01 calibrator=MaxCalibrator quant) model.layers.10.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.38e+01 calibrator=MaxCalibrator quant) model.layers.10.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.10.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.59e-01 calibrator=MaxCalibrator quant) model.layers.10.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.50e+00 calibrator=MaxCalibrator quant) model.layers.10.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.10.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.05e-01 calibrator=MaxCalibrator quant) model.layers.10.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.10.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.10.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.10.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.10.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.39e+01 calibrator=MaxCalibrator quant) model.layers.10.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.10.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.17e-01 calibrator=MaxCalibrator quant) model.layers.10.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.39e+01 calibrator=MaxCalibrator quant) model.layers.10.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.10.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.76e-01 calibrator=MaxCalibrator quant) model.layers.10.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.55e+01 calibrator=MaxCalibrator quant) model.layers.10.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.10.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.10e+00 calibrator=MaxCalibrator quant) model.layers.11.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.45e+01 calibrator=MaxCalibrator quant) model.layers.11.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.11.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.73e-01 calibrator=MaxCalibrator quant) model.layers.11.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.45e+01 calibrator=MaxCalibrator quant) model.layers.11.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.11.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.81e-01 calibrator=MaxCalibrator quant) model.layers.11.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.45e+01 calibrator=MaxCalibrator quant) model.layers.11.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.11.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.01e-01 calibrator=MaxCalibrator quant) model.layers.11.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.22e+00 calibrator=MaxCalibrator quant) model.layers.11.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.11.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.50e-01 calibrator=MaxCalibrator quant) model.layers.11.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.11.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.11.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.11.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.11.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.96e+01 calibrator=MaxCalibrator quant) model.layers.11.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.11.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=8.12e-01 calibrator=MaxCalibrator quant) model.layers.11.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.96e+01 calibrator=MaxCalibrator quant) model.layers.11.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.11.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.95e-01 calibrator=MaxCalibrator quant) model.layers.11.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.25e+01 calibrator=MaxCalibrator quant) model.layers.11.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.11.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.28e+00 calibrator=MaxCalibrator quant) model.layers.12.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.28e+01 calibrator=MaxCalibrator quant) model.layers.12.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.12.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.12e-01 calibrator=MaxCalibrator quant) model.layers.12.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.28e+01 calibrator=MaxCalibrator quant) model.layers.12.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.12.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.01e-01 calibrator=MaxCalibrator quant) model.layers.12.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.28e+01 calibrator=MaxCalibrator quant) model.layers.12.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.12.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.07e-01 calibrator=MaxCalibrator quant) model.layers.12.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.78e+00 calibrator=MaxCalibrator quant) model.layers.12.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.12.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.23e-01 calibrator=MaxCalibrator quant) model.layers.12.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.12.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.12.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.12.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.12.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.86e+01 calibrator=MaxCalibrator quant) model.layers.12.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.12.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.72e-01 calibrator=MaxCalibrator quant) model.layers.12.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.86e+01 calibrator=MaxCalibrator quant) model.layers.12.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.12.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.33e-01 calibrator=MaxCalibrator quant) model.layers.12.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.40e+01 calibrator=MaxCalibrator quant) model.layers.12.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.12.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=9.45e-01 calibrator=MaxCalibrator quant) model.layers.13.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.95e+01 calibrator=MaxCalibrator quant) model.layers.13.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.13.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.61e-01 calibrator=MaxCalibrator quant) model.layers.13.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.95e+01 calibrator=MaxCalibrator quant) model.layers.13.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.13.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.07e-01 calibrator=MaxCalibrator quant) model.layers.13.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.95e+01 calibrator=MaxCalibrator quant) model.layers.13.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.13.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.87e-01 calibrator=MaxCalibrator quant) model.layers.13.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.56e+00 calibrator=MaxCalibrator quant) model.layers.13.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.13.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.08e-01 calibrator=MaxCalibrator quant) model.layers.13.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.13.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.13.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.13.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.13.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.51e+01 calibrator=MaxCalibrator quant) model.layers.13.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.13.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=8.67e-01 calibrator=MaxCalibrator quant) model.layers.13.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.51e+01 calibrator=MaxCalibrator quant) model.layers.13.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.13.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.00e+00 calibrator=MaxCalibrator quant) model.layers.13.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.98e+02 calibrator=MaxCalibrator quant) model.layers.13.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.13.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.04e+00 calibrator=MaxCalibrator quant) model.layers.14.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.98e+01 calibrator=MaxCalibrator quant) model.layers.14.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.14.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.49e-01 calibrator=MaxCalibrator quant) model.layers.14.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.98e+01 calibrator=MaxCalibrator quant) model.layers.14.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.14.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.40e-01 calibrator=MaxCalibrator quant) model.layers.14.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.98e+01 calibrator=MaxCalibrator quant) model.layers.14.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.14.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.08e-01 calibrator=MaxCalibrator quant) model.layers.14.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.69e+00 calibrator=MaxCalibrator quant) model.layers.14.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.14.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.51e-01 calibrator=MaxCalibrator quant) model.layers.14.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.14.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.14.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.14.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.14.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.64e+01 calibrator=MaxCalibrator quant) model.layers.14.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.14.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=9.65e-01 calibrator=MaxCalibrator quant) model.layers.14.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.64e+01 calibrator=MaxCalibrator quant) model.layers.14.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.14.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.86e-01 calibrator=MaxCalibrator quant) model.layers.14.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.10e+01 calibrator=MaxCalibrator quant) model.layers.14.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.14.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.13e+00 calibrator=MaxCalibrator quant) model.layers.15.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.05e+01 calibrator=MaxCalibrator quant) model.layers.15.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.15.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.57e-01 calibrator=MaxCalibrator quant) model.layers.15.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.05e+01 calibrator=MaxCalibrator quant) model.layers.15.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.15.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.41e-01 calibrator=MaxCalibrator quant) model.layers.15.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.05e+01 calibrator=MaxCalibrator quant) model.layers.15.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.15.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.96e-01 calibrator=MaxCalibrator quant) model.layers.15.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.19e+00 calibrator=MaxCalibrator quant) model.layers.15.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.15.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.04e-01 calibrator=MaxCalibrator quant) model.layers.15.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.15.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.15.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.15.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.15.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.78e+01 calibrator=MaxCalibrator quant) model.layers.15.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.15.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.00e+00 calibrator=MaxCalibrator quant) model.layers.15.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.78e+01 calibrator=MaxCalibrator quant) model.layers.15.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.15.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.86e-01 calibrator=MaxCalibrator quant) model.layers.15.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.61e+01 calibrator=MaxCalibrator quant) model.layers.15.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.15.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=9.45e-01 calibrator=MaxCalibrator quant) model.layers.16.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.20e+01 calibrator=MaxCalibrator quant) model.layers.16.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.16.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.55e-01 calibrator=MaxCalibrator quant) model.layers.16.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.20e+01 calibrator=MaxCalibrator quant) model.layers.16.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.16.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.03e-01 calibrator=MaxCalibrator quant) model.layers.16.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.20e+01 calibrator=MaxCalibrator quant) model.layers.16.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.16.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.33e-01 calibrator=MaxCalibrator quant) model.layers.16.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.19e+00 calibrator=MaxCalibrator quant) model.layers.16.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.16.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.41e-01 calibrator=MaxCalibrator quant) model.layers.16.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.16.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.16.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.16.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.16.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.81e+01 calibrator=MaxCalibrator quant) model.layers.16.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.16.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.97e-01 calibrator=MaxCalibrator quant) model.layers.16.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.81e+01 calibrator=MaxCalibrator quant) model.layers.16.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.16.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.27e-01 calibrator=MaxCalibrator quant) model.layers.16.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.51e+01 calibrator=MaxCalibrator quant) model.layers.16.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.16.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.01e+00 calibrator=MaxCalibrator quant) model.layers.17.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.85e+01 calibrator=MaxCalibrator quant) model.layers.17.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.17.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.51e-01 calibrator=MaxCalibrator quant) model.layers.17.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.85e+01 calibrator=MaxCalibrator quant) model.layers.17.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.17.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.89e-01 calibrator=MaxCalibrator quant) model.layers.17.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.85e+01 calibrator=MaxCalibrator quant) model.layers.17.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.17.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.83e-01 calibrator=MaxCalibrator quant) model.layers.17.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.22e+00 calibrator=MaxCalibrator quant) model.layers.17.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.17.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.90e-01 calibrator=MaxCalibrator quant) model.layers.17.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.17.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.17.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.17.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.17.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.71e+01 calibrator=MaxCalibrator quant) model.layers.17.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.17.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.73e-01 calibrator=MaxCalibrator quant) model.layers.17.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.71e+01 calibrator=MaxCalibrator quant) model.layers.17.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.17.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.86e-01 calibrator=MaxCalibrator quant) model.layers.17.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.58e+01 calibrator=MaxCalibrator quant) model.layers.17.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.17.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.01e+00 calibrator=MaxCalibrator quant) model.layers.18.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.50e+01 calibrator=MaxCalibrator quant) model.layers.18.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.18.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.46e-01 calibrator=MaxCalibrator quant) model.layers.18.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.50e+01 calibrator=MaxCalibrator quant) model.layers.18.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.18.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.54e-01 calibrator=MaxCalibrator quant) model.layers.18.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.50e+01 calibrator=MaxCalibrator quant) model.layers.18.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.18.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.92e-01 calibrator=MaxCalibrator quant) model.layers.18.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.78e+00 calibrator=MaxCalibrator quant) model.layers.18.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.18.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.37e-01 calibrator=MaxCalibrator quant) model.layers.18.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.18.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.18.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.18.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.18.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.56e+01 calibrator=MaxCalibrator quant) model.layers.18.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.18.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.36e+00 calibrator=MaxCalibrator quant) model.layers.18.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.56e+01 calibrator=MaxCalibrator quant) model.layers.18.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.18.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.84e-01 calibrator=MaxCalibrator quant) model.layers.18.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.40e+01 calibrator=MaxCalibrator quant) model.layers.18.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.18.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.01e+00 calibrator=MaxCalibrator quant) model.layers.19.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=8.20e+01 calibrator=MaxCalibrator quant) model.layers.19.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.19.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.73e-01 calibrator=MaxCalibrator quant) model.layers.19.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=8.20e+01 calibrator=MaxCalibrator quant) model.layers.19.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.19.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.30e-01 calibrator=MaxCalibrator quant) model.layers.19.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=8.20e+01 calibrator=MaxCalibrator quant) model.layers.19.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.19.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.14e-01 calibrator=MaxCalibrator quant) model.layers.19.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.84e+00 calibrator=MaxCalibrator quant) model.layers.19.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.19.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.05e-01 calibrator=MaxCalibrator quant) model.layers.19.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.19.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.19.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.19.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.19.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.50e+01 calibrator=MaxCalibrator quant) model.layers.19.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.19.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.76e-01 calibrator=MaxCalibrator quant) model.layers.19.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.50e+01 calibrator=MaxCalibrator quant) model.layers.19.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.19.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.52e-01 calibrator=MaxCalibrator quant) model.layers.19.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.60e+01 calibrator=MaxCalibrator quant) model.layers.19.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.19.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=9.18e-01 calibrator=MaxCalibrator quant) model.layers.20.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.15e+01 calibrator=MaxCalibrator quant) model.layers.20.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.20.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.73e-01 calibrator=MaxCalibrator quant) model.layers.20.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.15e+01 calibrator=MaxCalibrator quant) model.layers.20.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.20.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.11e-01 calibrator=MaxCalibrator quant) model.layers.20.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.15e+01 calibrator=MaxCalibrator quant) model.layers.20.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.20.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.71e-01 calibrator=MaxCalibrator quant) model.layers.20.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=8.81e+00 calibrator=MaxCalibrator quant) model.layers.20.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.20.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.54e-01 calibrator=MaxCalibrator quant) model.layers.20.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.20.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.20.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.20.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.20.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.48e+01 calibrator=MaxCalibrator quant) model.layers.20.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.20.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.21e-01 calibrator=MaxCalibrator quant) model.layers.20.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.48e+01 calibrator=MaxCalibrator quant) model.layers.20.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.20.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.20e-01 calibrator=MaxCalibrator quant) model.layers.20.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.10e+01 calibrator=MaxCalibrator quant) model.layers.20.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.20.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=8.44e-01 calibrator=MaxCalibrator quant) model.layers.21.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.75e+01 calibrator=MaxCalibrator quant) model.layers.21.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.21.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.22e-01 calibrator=MaxCalibrator quant) model.layers.21.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.75e+01 calibrator=MaxCalibrator quant) model.layers.21.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.21.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.46e-01 calibrator=MaxCalibrator quant) model.layers.21.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.75e+01 calibrator=MaxCalibrator quant) model.layers.21.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.21.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.70e-01 calibrator=MaxCalibrator quant) model.layers.21.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=9.38e+00 calibrator=MaxCalibrator quant) model.layers.21.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.21.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.60e-01 calibrator=MaxCalibrator quant) model.layers.21.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.21.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.21.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.21.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.21.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.48e+01 calibrator=MaxCalibrator quant) model.layers.21.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.21.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.77e-01 calibrator=MaxCalibrator quant) model.layers.21.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.48e+01 calibrator=MaxCalibrator quant) model.layers.21.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.21.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.48e-01 calibrator=MaxCalibrator quant) model.layers.21.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.22e+02 calibrator=MaxCalibrator quant) model.layers.21.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.21.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.41e+00 calibrator=MaxCalibrator quant) model.layers.22.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.85e+01 calibrator=MaxCalibrator quant) model.layers.22.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.22.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.42e-01 calibrator=MaxCalibrator quant) model.layers.22.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.85e+01 calibrator=MaxCalibrator quant) model.layers.22.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.22.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.11e-01 calibrator=MaxCalibrator quant) model.layers.22.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.85e+01 calibrator=MaxCalibrator quant) model.layers.22.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.22.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.54e-01 calibrator=MaxCalibrator quant) model.layers.22.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=8.62e+00 calibrator=MaxCalibrator quant) model.layers.22.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.22.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.96e-01 calibrator=MaxCalibrator quant) model.layers.22.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.22.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.22.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.22.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.22.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.55e+01 calibrator=MaxCalibrator quant) model.layers.22.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.22.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.19e-01 calibrator=MaxCalibrator quant) model.layers.22.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.55e+01 calibrator=MaxCalibrator quant) model.layers.22.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.22.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.66e-01 calibrator=MaxCalibrator quant) model.layers.22.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=9.90e+01 calibrator=MaxCalibrator quant) model.layers.22.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.22.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.09e+00 calibrator=MaxCalibrator quant) model.layers.23.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.10e+01 calibrator=MaxCalibrator quant) model.layers.23.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.23.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.32e-01 calibrator=MaxCalibrator quant) model.layers.23.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.10e+01 calibrator=MaxCalibrator quant) model.layers.23.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.23.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.63e-01 calibrator=MaxCalibrator quant) model.layers.23.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.10e+01 calibrator=MaxCalibrator quant) model.layers.23.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.23.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.48e-01 calibrator=MaxCalibrator quant) model.layers.23.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.19e+01 calibrator=MaxCalibrator quant) model.layers.23.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.23.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=7.73e-01 calibrator=MaxCalibrator quant) model.layers.23.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.23.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.23.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.23.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.23.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.89e+01 calibrator=MaxCalibrator quant) model.layers.23.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.23.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.70e-01 calibrator=MaxCalibrator quant) model.layers.23.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.89e+01 calibrator=MaxCalibrator quant) model.layers.23.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.23.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.43e-01 calibrator=MaxCalibrator quant) model.layers.23.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.14e+02 calibrator=MaxCalibrator quant) model.layers.23.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.23.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.08e+00 calibrator=MaxCalibrator quant) model.layers.24.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.72e+01 calibrator=MaxCalibrator quant) model.layers.24.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.24.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.89e-01 calibrator=MaxCalibrator quant) model.layers.24.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.72e+01 calibrator=MaxCalibrator quant) model.layers.24.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.24.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.36e-01 calibrator=MaxCalibrator quant) model.layers.24.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.72e+01 calibrator=MaxCalibrator quant) model.layers.24.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.24.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.26e-01 calibrator=MaxCalibrator quant) model.layers.24.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.19e+01 calibrator=MaxCalibrator quant) model.layers.24.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.24.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.80e-01 calibrator=MaxCalibrator quant) model.layers.24.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.24.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.24.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.24.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.24.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.08e+01 calibrator=MaxCalibrator quant) model.layers.24.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.24.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.47e-01 calibrator=MaxCalibrator quant) model.layers.24.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.08e+01 calibrator=MaxCalibrator quant) model.layers.24.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.24.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.04e+00 calibrator=MaxCalibrator quant) model.layers.24.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.43e+02 calibrator=MaxCalibrator quant) model.layers.24.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.24.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.09e+00 calibrator=MaxCalibrator quant) model.layers.25.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.40e+01 calibrator=MaxCalibrator quant) model.layers.25.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.25.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.04e-01 calibrator=MaxCalibrator quant) model.layers.25.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.40e+01 calibrator=MaxCalibrator quant) model.layers.25.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.25.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.09e-01 calibrator=MaxCalibrator quant) model.layers.25.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.40e+01 calibrator=MaxCalibrator quant) model.layers.25.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.25.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.28e-01 calibrator=MaxCalibrator quant) model.layers.25.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.29e+01 calibrator=MaxCalibrator quant) model.layers.25.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.25.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=9.06e-01 calibrator=MaxCalibrator quant) model.layers.25.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.25.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.25.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.25.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.25.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.19e+01 calibrator=MaxCalibrator quant) model.layers.25.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.25.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=7.62e-01 calibrator=MaxCalibrator quant) model.layers.25.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.19e+01 calibrator=MaxCalibrator quant) model.layers.25.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.25.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.69e-01 calibrator=MaxCalibrator quant) model.layers.25.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.11e+02 calibrator=MaxCalibrator quant) model.layers.25.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.25.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.09e+00 calibrator=MaxCalibrator quant) model.layers.26.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.92e+01 calibrator=MaxCalibrator quant) model.layers.26.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.26.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.28e-01 calibrator=MaxCalibrator quant) model.layers.26.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.92e+01 calibrator=MaxCalibrator quant) model.layers.26.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.26.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.64e-01 calibrator=MaxCalibrator quant) model.layers.26.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.92e+01 calibrator=MaxCalibrator quant) model.layers.26.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.26.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.93e-01 calibrator=MaxCalibrator quant) model.layers.26.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.38e+01 calibrator=MaxCalibrator quant) model.layers.26.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.26.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=6.48e-01 calibrator=MaxCalibrator quant) model.layers.26.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.26.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.26.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.26.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.26.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.22e+01 calibrator=MaxCalibrator quant) model.layers.26.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.26.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.17e-01 calibrator=MaxCalibrator quant) model.layers.26.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.22e+01 calibrator=MaxCalibrator quant) model.layers.26.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.26.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.88e-01 calibrator=MaxCalibrator quant) model.layers.26.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.96e+02 calibrator=MaxCalibrator quant) model.layers.26.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.26.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.51e+00 calibrator=MaxCalibrator quant) model.layers.27.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.10e+01 calibrator=MaxCalibrator quant) model.layers.27.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.27.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.08e-01 calibrator=MaxCalibrator quant) model.layers.27.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.10e+01 calibrator=MaxCalibrator quant) model.layers.27.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.27.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.93e-01 calibrator=MaxCalibrator quant) model.layers.27.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.10e+01 calibrator=MaxCalibrator quant) model.layers.27.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.27.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.55e-01 calibrator=MaxCalibrator quant) model.layers.27.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.26e+01 calibrator=MaxCalibrator quant) model.layers.27.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.27.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.31e+00 calibrator=MaxCalibrator quant) model.layers.27.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.27.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.27.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.27.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.27.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.48e+01 calibrator=MaxCalibrator quant) model.layers.27.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.27.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=6.76e-01 calibrator=MaxCalibrator quant) model.layers.27.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.48e+01 calibrator=MaxCalibrator quant) model.layers.27.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.27.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.67e-01 calibrator=MaxCalibrator quant) model.layers.27.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.97e+02 calibrator=MaxCalibrator quant) model.layers.27.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.27.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.49e+00 calibrator=MaxCalibrator quant) model.layers.28.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.10e+01 calibrator=MaxCalibrator quant) model.layers.28.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.28.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.20e-01 calibrator=MaxCalibrator quant) model.layers.28.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.10e+01 calibrator=MaxCalibrator quant) model.layers.28.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.28.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.28e-01 calibrator=MaxCalibrator quant) model.layers.28.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=4.10e+01 calibrator=MaxCalibrator quant) model.layers.28.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.28.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.16e-01 calibrator=MaxCalibrator quant) model.layers.28.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.39e+01 calibrator=MaxCalibrator quant) model.layers.28.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.28.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=5.47e-01 calibrator=MaxCalibrator quant) model.layers.28.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.28.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.28.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.28.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.28.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.82e+01 calibrator=MaxCalibrator quant) model.layers.28.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.28.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=8.44e-01 calibrator=MaxCalibrator quant) model.layers.28.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=3.82e+01 calibrator=MaxCalibrator quant) model.layers.28.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.28.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.00e-01 calibrator=MaxCalibrator quant) model.layers.28.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.25e+02 calibrator=MaxCalibrator quant) model.layers.28.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.28.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.24e+00 calibrator=MaxCalibrator quant) model.layers.29.self_attn.q_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.68e+01 calibrator=MaxCalibrator quant) model.layers.29.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.29.self_attn.q_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.54e-01 calibrator=MaxCalibrator quant) model.layers.29.self_attn.k_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.68e+01 calibrator=MaxCalibrator quant) model.layers.29.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.29.self_attn.k_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.07e-01 calibrator=MaxCalibrator quant) model.layers.29.self_attn.v_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.68e+01 calibrator=MaxCalibrator quant) model.layers.29.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.29.self_attn.v_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=3.54e-01 calibrator=MaxCalibrator quant) model.layers.29.self_attn.o_proj.input_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=2.69e+01 calibrator=MaxCalibrator quant) model.layers.29.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.29.self_attn.o_proj.weight_quantizer TensorQuantizer((4, 3) bit fake per-tensor amax=1.26e+00 calibrator=MaxCalibrator quant) model.layers.29.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.29.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.29.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.29.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.29.mlp.gate_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.80e+01 calibrator=MaxCalibrator quant) model.layers.29.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.29.mlp.gate_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=5.12e-01 calibrator=MaxCalibrator quant) model.layers.29.mlp.up_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=4.80e+01 calibrator=MaxCalibrator quant) model.layers.29.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.29.mlp.up_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=8.44e-01 calibrator=MaxCalibrator quant) model.layers.29.mlp.down_proj.input_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=2.62e+02 calibrator=MaxCalibrator quant) model.layers.29.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.29.mlp.down_proj.weight_quantizer TensorQuantizer((2, 1) bit fake block_sizes={-1: 16, 'type': 'dynamic', 'scale_bits': (4, 3)}, amax=1.08e+00 calibrator=MaxCalibrator quant) model.layers.30.self_attn.q_proj.input_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.q_proj.weight_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.k_proj.input_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.k_proj.weight_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.v_proj.input_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.v_proj.weight_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.o_proj.input_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.o_proj.weight_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.30.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.30.mlp.gate_proj.input_quantizer TensorQuantizer(disabled) model.layers.30.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.30.mlp.gate_proj.weight_quantizer TensorQuantizer(disabled) model.layers.30.mlp.up_proj.input_quantizer TensorQuantizer(disabled) model.layers.30.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.30.mlp.up_proj.weight_quantizer TensorQuantizer(disabled) model.layers.30.mlp.down_proj.input_quantizer TensorQuantizer(disabled) model.layers.30.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.30.mlp.down_proj.weight_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.q_proj.input_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.q_proj.output_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.q_proj.weight_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.k_proj.input_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.k_proj.output_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.k_proj.weight_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.v_proj.input_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.v_proj.output_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.v_proj.weight_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.o_proj.input_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.o_proj.output_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.o_proj.weight_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.q_bmm_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.k_bmm_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.v_bmm_quantizer TensorQuantizer(disabled) model.layers.31.self_attn.softmax_quantizer TensorQuantizer(disabled) model.layers.31.mlp.gate_proj.input_quantizer TensorQuantizer(disabled) model.layers.31.mlp.gate_proj.output_quantizer TensorQuantizer(disabled) model.layers.31.mlp.gate_proj.weight_quantizer TensorQuantizer(disabled) model.layers.31.mlp.up_proj.input_quantizer TensorQuantizer(disabled) model.layers.31.mlp.up_proj.output_quantizer TensorQuantizer(disabled) model.layers.31.mlp.up_proj.weight_quantizer TensorQuantizer(disabled) model.layers.31.mlp.down_proj.input_quantizer TensorQuantizer(disabled) model.layers.31.mlp.down_proj.output_quantizer TensorQuantizer(disabled) model.layers.31.mlp.down_proj.weight_quantizer TensorQuantizer(disabled) lm_head.input_quantizer TensorQuantizer(disabled) lm_head.output_quantizer TensorQuantizer(disabled) lm_head.weight_quantizer TensorQuantizer(disabled) 806 TensorQuantizers found in model