layers.0.ffn.experts.241.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0101e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.159.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8407e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.90.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8831e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.31.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5652e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.134.w2: m=-13 lossless=100.0000% mse(v3)=2.6171e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.162.w2: m=-13 lossless=100.0000% mse(v3)=2.6153e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.224.w2: m=-13 lossless=100.0000% mse(v3)=2.6220e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.141.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9668e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.230.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9822e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.106.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7812e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.249.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9330e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.133.w2: m=-13 lossless=100.0000% mse(v3)=2.6495e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.191.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5970e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.249.w2: m=-13 lossless=100.0000% mse(v3)=2.6658e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.244.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7004e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.66.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8919e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.240.w2: m=-12 lossless=100.0000% mse(v3)=2.5948e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.54.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9822e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.136.w2: m=-13 lossless=100.0000% mse(v3)=2.6253e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.209.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9724e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.90.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0438e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.168.w2: m=-13 lossless=100.0000% mse(v3)=2.6296e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.91.w2: m=-13 lossless=100.0000% mse(v3)=2.6626e-06 mse(v3.5)=0.0000e+00 layers.0.ffn.experts.62.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5893e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.23.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5449e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.34.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5978e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.178.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5837e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.53.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5498e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.105.w2: m=-13 lossless=100.0000% mse(v3)=2.8396e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.126.w2: m=-13 lossless=100.0000% mse(v3)=2.5861e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.173.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5864e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.67.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5992e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.208.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8588e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.153.w2: m=-13 lossless=100.0000% mse(v3)=2.8412e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.166.w2: m=-13 lossless=100.0000% mse(v3)=2.5658e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.201.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5580e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.42.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5810e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.247.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5971e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.73.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5528e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.157.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5726e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.224.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5699e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.89.w2: m=-13 lossless=100.0000% mse(v3)=2.5596e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.36.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5595e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.198.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5832e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.102.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5631e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.7.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5523e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.11.w2: m=-13 lossless=100.0000% mse(v3)=2.8358e-06 mse(v3.5)=0.0000e+00 layers.1.ffn.experts.70.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5869e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.210.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7158e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.94.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6963e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.44.w2: m=-13 lossless=100.0000% mse(v3)=2.6139e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.223.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6972e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.170.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7087e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.221.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6155e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.1.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6059e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.72.w2: m=-13 lossless=100.0000% mse(v3)=2.9061e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.89.w2: m=-13 lossless=100.0000% mse(v3)=2.6174e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.137.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6990e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.165.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7019e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.218.w2: m=-13 lossless=100.0000% mse(v3)=2.6334e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.254.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7063e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.102.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6198e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.141.w2: m=-13 lossless=100.0000% mse(v3)=2.6296e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.118.w2: m=-13 lossless=100.0000% mse(v3)=2.6109e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.36.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5959e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.249.w2: m=-13 lossless=100.0000% mse(v3)=2.6177e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.8.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6122e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.29.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6118e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.239.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7191e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.143.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6159e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.22.w2: m=-13 lossless=100.0000% mse(v3)=2.6037e-06 mse(v3.5)=0.0000e+00 layers.2.ffn.experts.139.w2: m=-13 lossless=100.0000% mse(v3)=2.6117e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.12.w2: m=-12 lossless=100.0000% mse(v3)=2.4813e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.155.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5319e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.232.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6746e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.91.w2: m=-12 lossless=100.0000% mse(v3)=2.5000e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.231.w2: m=-12 lossless=100.0000% mse(v3)=2.4861e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.202.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6649e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.58.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5510e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.201.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4141e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.20.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5063e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.182.w2: m=-12 lossless=100.0000% mse(v3)=2.3482e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.22.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4925e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.78.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5150e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.198.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7608e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.1.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5426e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.144.w2: m=-12 lossless=100.0000% mse(v3)=2.7948e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.95.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4873e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.168.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5791e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.109.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7394e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.253.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7182e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.174.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4679e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.33.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4621e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.241.w2: m=-12 lossless=100.0000% mse(v3)=2.4063e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.203.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5983e-06 mse(v3.5)=0.0000e+00 layers.3.ffn.experts.208.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5419e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.32.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4249e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.111.w2: m=-12 lossless=100.0000% mse(v3)=2.4311e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.198.w2: m=-13 lossless=100.0000% mse(v3)=2.3852e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.132.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6394e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.109.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5536e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.190.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6537e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.131.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8550e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.128.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2950e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.52.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3189e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.113.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7586e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.117.w2: m=-13 lossless=100.0000% mse(v3)=2.5582e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.103.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3327e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.240.w2: m=-13 lossless=100.0000% mse(v3)=2.3092e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.15.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7621e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.171.w2: m=-13 lossless=100.0000% mse(v3)=2.3499e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.214.w2: m=-13 lossless=100.0000% mse(v3)=2.4935e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.54.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5910e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.51.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5823e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.170.w2: m=-11 lossless=100.0000% mse(v3)=2.2927e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.90.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5251e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.64.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5665e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.60.w2: m=-13 lossless=100.0000% mse(v3)=2.4139e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.79.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9909e-06 mse(v3.5)=0.0000e+00 layers.4.ffn.experts.94.w2: m=-13 lossless=100.0000% mse(v3)=2.4184e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.158.w2: m=-13 lossless=100.0000% mse(v3)=2.7471e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.68.w2: m=-13 lossless=100.0000% mse(v3)=2.5044e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.38.w2: m=-13 lossless=100.0000% mse(v3)=2.4487e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.98.w2: m=-13 lossless=100.0000% mse(v3)=2.4002e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.247.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5256e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.176.w2: m=-13 lossless=100.0000% mse(v3)=2.5622e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.177.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5391e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.79.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6012e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.52.w2: m=-12 lossless=100.0000% mse(v3)=2.4751e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.157.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4279e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.85.w2: m=-13 lossless=100.0000% mse(v3)=2.4808e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.165.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5148e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.135.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2339e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.229.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3162e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.92.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4914e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.25.w2: m=-12 lossless=100.0000% mse(v3)=2.4079e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.11.w2: m=-12 lossless=100.0000% mse(v3)=2.3924e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.130.w2: m=-12 lossless=100.0000% mse(v3)=2.3835e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.77.w2: m=-13 lossless=100.0000% mse(v3)=2.4364e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.216.w2: m=-13 lossless=100.0000% mse(v3)=2.5032e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.52.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4471e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.118.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6481e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.144.w2: m=-12 lossless=100.0000% mse(v3)=2.2282e-06 mse(v3.5)=0.0000e+00 layers.5.ffn.experts.172.w2: m=-12 lossless=100.0000% mse(v3)=2.4104e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.163.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6331e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.206.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7223e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.118.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3696e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.8.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5363e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.54.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5140e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.249.w2: m=-13 lossless=100.0000% mse(v3)=2.3326e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.18.w2: m=-13 lossless=100.0000% mse(v3)=2.5689e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.80.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4656e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.4.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4821e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.49.w2: m=-13 lossless=100.0000% mse(v3)=2.2974e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.211.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4274e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.16.w2: m=-13 lossless=100.0000% mse(v3)=2.5758e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.229.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6096e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.8.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5030e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.229.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5499e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.207.w2: m=-13 lossless=100.0000% mse(v3)=2.5369e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.253.w2: m=-13 lossless=100.0000% mse(v3)=2.3864e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.162.w2: m=-13 lossless=100.0000% mse(v3)=2.4713e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.50.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3979e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.124.w2: m=-13 lossless=100.0000% mse(v3)=2.4060e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.144.w2: m=-13 lossless=100.0000% mse(v3)=2.4397e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.243.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4493e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.169.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4559e-06 mse(v3.5)=0.0000e+00 layers.6.ffn.experts.74.w2: m=-13 lossless=100.0000% mse(v3)=2.4161e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.26.w2: m=-13 lossless=100.0000% mse(v3)=2.4094e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.65.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7760e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.249.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5800e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.39.w3: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3442e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.135.w2: m=-13 lossless=100.0000% mse(v3)=2.6660e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.44.w2: m=-13 lossless=100.0000% mse(v3)=2.4096e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.210.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5581e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.1.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5144e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.89.w2: m=-13 lossless=100.0000% mse(v3)=2.5758e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.43.w2: m=-12 lossless=100.0000% mse(v3)=2.5753e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.196.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2528e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.64.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5574e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.158.w2: m=-13 lossless=100.0000% mse(v3)=2.7614e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.184.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6042e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.29.w2: m=-12 lossless=100.0000% mse(v3)=2.4270e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.166.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4858e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.199.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5456e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.95.w2: m=-13 lossless=100.0000% mse(v3)=2.3503e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.5.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4235e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.168.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4599e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.210.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5377e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.67.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5102e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.172.w2: m=-11 lossless=100.0000% mse(v3)=2.3588e-06 mse(v3.5)=0.0000e+00 layers.7.ffn.experts.190.w2: m=-13 lossless=100.0000% mse(v3)=2.5020e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.175.w2: m=-13 lossless=100.0000% mse(v3)=2.8009e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.109.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4431e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.205.w2: m=-13 lossless=100.0000% mse(v3)=2.4805e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.151.w2: m=-13 lossless=100.0000% mse(v3)=2.4975e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.75.w2: m=-13 lossless=100.0000% mse(v3)=2.5532e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.8.w2: m=-13 lossless=100.0000% mse(v3)=2.4340e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.147.w2: m=-13 lossless=100.0000% mse(v3)=2.2113e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.19.w2: m=-13 lossless=100.0000% mse(v3)=2.5271e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.63.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3264e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.91.w2: m=-13 lossless=100.0000% mse(v3)=2.3436e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.200.w2: m=-13 lossless=100.0000% mse(v3)=2.4294e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.74.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3453e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.90.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6404e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.171.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4124e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.172.w2: m=-13 lossless=100.0000% mse(v3)=2.1933e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.233.w2: m=-13 lossless=100.0000% mse(v3)=2.4396e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.131.w2: m=-13 lossless=100.0000% mse(v3)=2.3524e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.176.w2: m=-13 lossless=100.0000% mse(v3)=2.3932e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.233.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4484e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.54.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5950e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.91.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3923e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.198.w2: m=-13 lossless=100.0000% mse(v3)=2.6032e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.63.w2: m=-13 lossless=100.0000% mse(v3)=2.3043e-06 mse(v3.5)=0.0000e+00 layers.8.ffn.experts.47.w2: m=-13 lossless=100.0000% mse(v3)=2.5107e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.157.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7920e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.242.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7620e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.228.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4416e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.211.w2: m=-12 lossless=100.0000% mse(v3)=2.3339e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.240.w2: m=-13 lossless=100.0000% mse(v3)=2.4092e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.34.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6219e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.135.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6450e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.182.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7658e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.129.w2: m=-13 lossless=100.0000% mse(v3)=2.3416e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.228.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4311e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.34.w2: m=-13 lossless=100.0000% mse(v3)=2.4618e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.234.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0049e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.110.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5089e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.193.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3936e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.220.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5946e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.45.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4455e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.55.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4220e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.213.w2: m=-13 lossless=100.0000% mse(v3)=2.2992e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.102.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5079e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.172.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5216e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.107.w2: m=-13 lossless=100.0000% mse(v3)=2.1865e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.207.w2: m=-13 lossless=100.0000% mse(v3)=2.7215e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.245.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4158e-06 mse(v3.5)=0.0000e+00 layers.9.ffn.experts.75.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4916e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.81.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4298e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.26.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6152e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.32.w2: m=-12 lossless=100.0000% mse(v3)=2.5197e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.35.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5691e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.103.w2: m=-12 lossless=100.0000% mse(v3)=2.3329e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.173.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.1912e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.251.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2408e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.96.w2: m=-12 lossless=100.0000% mse(v3)=2.2710e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.74.w2: m=-13 lossless=100.0000% mse(v3)=2.4387e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.184.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7768e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.114.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5953e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.45.w2: m=-13 lossless=100.0000% mse(v3)=2.1682e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.29.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5245e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.149.w2: m=-13 lossless=100.0000% mse(v3)=2.3914e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.46.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3841e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.173.w2: m=-13 lossless=100.0000% mse(v3)=2.1050e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.240.w2: m=-13 lossless=100.0000% mse(v3)=2.4298e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.15.w2: m=-13 lossless=100.0000% mse(v3)=2.2399e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.49.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3863e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.191.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7000e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.233.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3015e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.41.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6758e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.110.w2: m=-13 lossless=100.0000% mse(v3)=2.3544e-06 mse(v3.5)=0.0000e+00 layers.10.ffn.experts.175.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4479e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.187.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3822e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.213.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5024e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.123.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2938e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.209.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4383e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.174.w2: m=-13 lossless=100.0000% mse(v3)=2.2477e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.153.w2: m=-13 lossless=100.0000% mse(v3)=2.4264e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.198.w2: m=-13 lossless=100.0000% mse(v3)=2.3789e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.234.w2: m=-13 lossless=100.0000% mse(v3)=2.5053e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.68.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6237e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.166.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3778e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.52.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3697e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.172.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7273e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.18.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6198e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.175.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7709e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.144.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3634e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.81.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4916e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.100.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5353e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.189.w2: m=-13 lossless=100.0000% mse(v3)=2.5291e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.15.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3232e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.84.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4068e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.33.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3848e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.87.w2: m=-13 lossless=100.0000% mse(v3)=2.2578e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.192.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5838e-06 mse(v3.5)=0.0000e+00 layers.11.ffn.experts.165.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3108e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.191.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3501e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.181.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4130e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.110.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5596e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.41.w2: m=-13 lossless=100.0000% mse(v3)=2.2902e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.187.w2: m=-13 lossless=100.0000% mse(v3)=2.2468e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.242.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6411e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.185.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2260e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.244.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4285e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.112.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3601e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.195.w2: m=-13 lossless=100.0000% mse(v3)=2.3316e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.99.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3220e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.232.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4160e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.102.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5714e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.50.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5807e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.169.w2: m=-12 lossless=100.0000% mse(v3)=2.4360e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.249.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3182e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.70.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5162e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.178.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4042e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.150.w2: m=-13 lossless=100.0000% mse(v3)=2.2556e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.127.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3937e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.239.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3595e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.19.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3352e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.226.w2: m=-13 lossless=100.0000% mse(v3)=2.6141e-06 mse(v3.5)=0.0000e+00 layers.12.ffn.experts.139.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6472e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.0.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4276e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.36.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3357e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.50.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.1953e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.205.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5724e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.222.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.1005e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.11.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4357e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.74.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5018e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.100.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4219e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.211.w2: m=-13 lossless=100.0000% mse(v3)=2.3792e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.105.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4222e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.40.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3151e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.75.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2927e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.113.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3705e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.89.w2: m=-13 lossless=100.0000% mse(v3)=2.4889e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.241.w2: m=-11 lossless=100.0000% mse(v3)=2.2340e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.237.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3889e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.26.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4755e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.250.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3559e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.129.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4043e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.174.w2: m=-13 lossless=100.0000% mse(v3)=2.4520e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.24.w2: m=-13 lossless=100.0000% mse(v3)=2.5280e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.194.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4350e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.145.w2: m=-13 lossless=100.0000% mse(v3)=2.2333e-06 mse(v3.5)=0.0000e+00 layers.13.ffn.experts.160.w2: m=-13 lossless=100.0000% mse(v3)=2.3605e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.158.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3577e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.194.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3630e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.3.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5959e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.226.w2: m=-13 lossless=100.0000% mse(v3)=2.2728e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.80.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4881e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.151.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4009e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.19.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4150e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.18.w2: m=-13 lossless=100.0000% mse(v3)=2.3058e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.131.w2: m=-13 lossless=100.0000% mse(v3)=2.2310e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.171.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3867e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.114.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8032e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.153.w2: m=-13 lossless=100.0000% mse(v3)=2.2493e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.162.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3739e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.186.w2: m=-13 lossless=100.0000% mse(v3)=2.5052e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.220.w2: m=-13 lossless=100.0000% mse(v3)=2.5090e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.34.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4786e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.209.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3366e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.21.w2: m=-13 lossless=100.0000% mse(v3)=2.4127e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.218.w2: m=-13 lossless=100.0000% mse(v3)=2.2656e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.191.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6334e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.115.w2: m=-13 lossless=100.0000% mse(v3)=2.3437e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.131.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3991e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.25.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5472e-06 mse(v3.5)=0.0000e+00 layers.14.ffn.experts.250.w2: m=-13 lossless=100.0000% mse(v3)=2.3611e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.41.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5697e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.122.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2735e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.179.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4847e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.42.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6518e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.170.w2: m=-13 lossless=100.0000% mse(v3)=2.2729e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.187.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3514e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.36.w2: m=-13 lossless=100.0000% mse(v3)=2.4296e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.146.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5036e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.228.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3870e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.206.w2: m=-13 lossless=100.0000% mse(v3)=2.2889e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.153.w2: m=-13 lossless=100.0000% mse(v3)=2.2676e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.98.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5558e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.148.w2: m=-13 lossless=100.0000% mse(v3)=2.2304e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.107.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3476e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.183.w2: m=-13 lossless=100.0000% mse(v3)=2.2659e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.217.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.1604e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.22.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6258e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.44.w2: m=-12 lossless=100.0000% mse(v3)=2.2760e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.246.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5238e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.250.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2427e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.223.w2: m=-13 lossless=100.0000% mse(v3)=2.3837e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.102.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3000e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.123.w2: m=-13 lossless=100.0000% mse(v3)=2.2513e-06 mse(v3.5)=0.0000e+00 layers.15.ffn.experts.135.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8266e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.255.w2: m=-11 lossless=100.0000% mse(v3)=2.3348e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.195.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6443e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.143.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4886e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.239.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4292e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.166.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=1.9101e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.214.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3506e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.127.w2: m=-12 lossless=100.0000% mse(v3)=2.3800e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.126.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3061e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.109.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9480e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.17.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.1987e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.173.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.1679e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.213.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3574e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.151.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3342e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.231.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3115e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.183.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4819e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.158.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6454e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.169.w2: m=-13 lossless=100.0000% mse(v3)=2.2568e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.54.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5907e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.102.w2: m=-12 lossless=100.0000% mse(v3)=2.3090e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.59.w2: m=-13 lossless=100.0000% mse(v3)=2.3633e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.73.w2: m=-13 lossless=100.0000% mse(v3)=2.4309e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.106.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7705e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.76.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4993e-06 mse(v3.5)=0.0000e+00 layers.16.ffn.experts.189.w1: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2364e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.143.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4672e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.235.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5646e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.70.w2: m=-11 lossless=100.0000% mse(v3)=2.4070e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.26.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3411e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.134.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4395e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.176.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6043e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.224.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3052e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.29.w2: m=-13 lossless=100.0000% mse(v3)=2.7094e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.135.w2: m=-13 lossless=100.0000% mse(v3)=2.4760e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.154.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4715e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.164.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8623e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.80.w2: m=-13 lossless=100.0000% mse(v3)=2.5271e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.189.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6006e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.158.w2: m=-13 lossless=100.0000% mse(v3)=2.6398e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.60.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3833e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.193.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5447e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.40.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3182e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.123.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3892e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.162.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9092e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.85.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3851e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.248.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4283e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.240.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3490e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.255.w2: m=-13 lossless=100.0000% mse(v3)=2.3107e-06 mse(v3.5)=0.0000e+00 layers.17.ffn.experts.94.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6659e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.211.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5505e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.105.w2: m=-12 lossless=100.0000% mse(v3)=2.4791e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.190.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5675e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.180.w2: m=-11 lossless=100.0000% mse(v3)=2.5585e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.251.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7531e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.112.w2: m=-12 lossless=100.0000% mse(v3)=2.5012e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.197.w1: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2821e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.122.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8670e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.107.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3781e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.78.w2: m=-13 lossless=100.0000% mse(v3)=2.7514e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.143.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3232e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.58.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2496e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.135.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4286e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.223.w2: m=-13 lossless=100.0000% mse(v3)=2.6021e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.0.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6720e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.166.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7289e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.159.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4195e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.75.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3703e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.76.w2: m=-13 lossless=100.0000% mse(v3)=2.5465e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.238.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5586e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.33.w1: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2710e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.75.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5113e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.79.w2: m=-11 lossless=100.0000% mse(v3)=2.6245e-06 mse(v3.5)=0.0000e+00 layers.18.ffn.experts.103.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.2396e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.23.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8013e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.245.w1: m=-10 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8432e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.207.w1: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9172e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.254.w2: m=-13 lossless=100.0000% mse(v3)=2.3815e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.231.w2: m=-11 lossless=100.0000% mse(v3)=2.6619e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.139.w2: m=-13 lossless=100.0000% mse(v3)=2.2523e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.97.w2: m=-12 lossless=100.0000% mse(v3)=2.5059e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.47.w3: m=-10 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5896e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.74.w2: m=-13 lossless=100.0000% mse(v3)=2.5356e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.237.w2: m=-12 lossless=100.0000% mse(v3)=2.6556e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.58.w3: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6302e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.11.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7316e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.111.w2: m=-12 lossless=100.0000% mse(v3)=2.5455e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.42.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0276e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.23.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7928e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.113.w3: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3488e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.191.w2: m=-12 lossless=100.0000% mse(v3)=2.5106e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.248.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5799e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.99.w2: m=-11 lossless=100.0000% mse(v3)=2.7890e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.205.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6837e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.128.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8354e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.54.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1115e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.68.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5941e-06 mse(v3.5)=0.0000e+00 layers.19.ffn.experts.233.w2: m=-13 lossless=100.0000% mse(v3)=2.2566e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.145.w2: m=-13 lossless=100.0000% mse(v3)=2.4709e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.109.w3: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9969e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.16.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5347e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.144.w2: m=-13 lossless=100.0000% mse(v3)=4.3121e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.102.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.3437e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.252.w2: m=-13 lossless=100.0000% mse(v3)=2.4196e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.94.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6810e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.156.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4825e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.85.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7258e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.198.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5689e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.200.w2: m=-11 lossless=100.0000% mse(v3)=2.7172e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.11.w2: m=-13 lossless=100.0000% mse(v3)=2.4230e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.62.w2: m=-13 lossless=100.0000% mse(v3)=2.5429e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.28.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9823e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.176.w2: m=-12 lossless=100.0000% mse(v3)=2.4490e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.70.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0639e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.210.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7833e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.6.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.1986e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.186.w2: m=-11 lossless=100.0000% mse(v3)=2.4826e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.140.w1: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5398e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.126.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6530e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.211.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6755e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.23.w2: m=-13 lossless=100.0000% mse(v3)=2.6140e-06 mse(v3.5)=0.0000e+00 layers.20.ffn.experts.142.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4667e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.32.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5173e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.224.w2: m=-13 lossless=100.0000% mse(v3)=2.6008e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.146.w2: m=-13 lossless=100.0000% mse(v3)=2.5355e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.113.w2: m=-13 lossless=100.0000% mse(v3)=2.4267e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.123.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4127e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.36.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3130e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.96.w2: m=-13 lossless=100.0000% mse(v3)=2.5709e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.56.w2: m=-13 lossless=100.0000% mse(v3)=2.4425e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.152.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6404e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.31.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5372e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.78.w2: m=-13 lossless=100.0000% mse(v3)=2.8273e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.207.w2: m=-13 lossless=100.0000% mse(v3)=2.3800e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.64.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4657e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.26.w2: m=-12 lossless=100.0000% mse(v3)=2.4216e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.100.w2: m=-13 lossless=100.0000% mse(v3)=2.5954e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.72.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5211e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.51.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8444e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.70.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6673e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.206.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5347e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.156.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7199e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.251.w2: m=-12 lossless=100.0000% mse(v3)=3.1345e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.212.w2: m=-13 lossless=100.0000% mse(v3)=2.5365e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.186.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4676e-06 mse(v3.5)=0.0000e+00 layers.21.ffn.experts.142.w2: m=-12 lossless=100.0000% mse(v3)=2.6139e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.15.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8666e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.196.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8004e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.119.w2: m=-13 lossless=100.0000% mse(v3)=2.6797e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.31.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7298e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.110.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5356e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.41.w2: m=-11 lossless=100.0000% mse(v3)=2.6334e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.115.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6219e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.69.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6666e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.71.w2: m=-13 lossless=100.0000% mse(v3)=3.0357e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.90.w2: m=-13 lossless=100.0000% mse(v3)=2.5075e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.72.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0285e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.241.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3698e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.203.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8918e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.214.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8997e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.227.w2: m=-12 lossless=100.0000% mse(v3)=2.9262e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.53.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6907e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.125.w3: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8088e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.102.w2: m=-13 lossless=100.0000% mse(v3)=2.8166e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.48.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4516e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.130.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8414e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.64.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1044e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.101.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7364e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.40.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5565e-06 mse(v3.5)=0.0000e+00 layers.22.ffn.experts.170.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6237e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.178.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0648e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.87.w2: m=-13 lossless=100.0000% mse(v3)=2.4394e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.155.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8808e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.86.w2: m=-13 lossless=100.0000% mse(v3)=2.5971e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.228.w2: m=-13 lossless=100.0000% mse(v3)=2.7629e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.197.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7813e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.251.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8684e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.145.w2: m=-13 lossless=100.0000% mse(v3)=2.5946e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.91.w2: m=-13 lossless=100.0000% mse(v3)=2.7175e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.154.w2: m=-13 lossless=100.0000% mse(v3)=2.5561e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.127.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5404e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.20.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7081e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.74.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5071e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.192.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6317e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.33.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6705e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.84.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7463e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.195.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7665e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.88.w2: m=-12 lossless=100.0000% mse(v3)=2.6235e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.91.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6115e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.54.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5772e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.235.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4949e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.249.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6431e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.38.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8098e-06 mse(v3.5)=0.0000e+00 layers.23.ffn.experts.175.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7352e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.183.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7642e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.229.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3139e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.35.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4885e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.181.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3847e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.115.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7107e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.245.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6774e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.200.w2: m=-13 lossless=100.0000% mse(v3)=2.6071e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.40.w2: m=-13 lossless=100.0000% mse(v3)=2.9092e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.164.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7005e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.11.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4838e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.41.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2608e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.38.w2: m=-13 lossless=100.0000% mse(v3)=2.6117e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.75.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7788e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.94.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6970e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.160.w2: m=-13 lossless=100.0000% mse(v3)=2.4949e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.194.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6312e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.119.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5659e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.12.w2: m=-13 lossless=100.0000% mse(v3)=2.4943e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.175.w2: m=-13 lossless=100.0000% mse(v3)=2.6595e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.36.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4926e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.72.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1926e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.32.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5788e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.11.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4223e-06 mse(v3.5)=0.0000e+00 layers.24.ffn.experts.230.w2: m=-13 lossless=100.0000% mse(v3)=2.5001e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.205.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0040e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.16.w2: m=-13 lossless=100.0000% mse(v3)=2.5250e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.199.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7550e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.159.w2: m=-13 lossless=100.0000% mse(v3)=2.5802e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.23.w2: m=-13 lossless=100.0000% mse(v3)=2.5255e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.153.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5230e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.219.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6978e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.128.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6510e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.111.w2: m=-13 lossless=100.0000% mse(v3)=2.8297e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.42.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5869e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.238.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5302e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.228.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5106e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.155.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7368e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.5.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4489e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.130.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5856e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.61.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5066e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.32.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6155e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.178.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7924e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.251.w2: m=-13 lossless=100.0000% mse(v3)=2.5335e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.104.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5270e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.148.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6736e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.64.w2: m=-12 lossless=100.0000% mse(v3)=2.5530e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.150.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7777e-06 mse(v3.5)=0.0000e+00 layers.25.ffn.experts.59.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6375e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.12.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.2605e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.60.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4310e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.22.w2: m=-13 lossless=100.0000% mse(v3)=2.6066e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.149.w2: m=-13 lossless=100.0000% mse(v3)=2.5273e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.1.w2: m=-13 lossless=100.0000% mse(v3)=2.4861e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.20.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6574e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.9.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5322e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.7.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4336e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.110.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6270e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.75.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4872e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.81.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7201e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.56.w2: m=-13 lossless=100.0000% mse(v3)=2.6933e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.121.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4939e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.163.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7775e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.211.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6229e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.207.w2: m=-13 lossless=100.0000% mse(v3)=2.6034e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.37.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6551e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.228.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7817e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.177.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3487e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.171.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0696e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.146.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5241e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.18.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6829e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.111.w2: m=-13 lossless=100.0000% mse(v3)=2.5032e-06 mse(v3.5)=0.0000e+00 layers.26.ffn.experts.148.w2: m=-13 lossless=100.0000% mse(v3)=2.7882e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.18.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7081e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.61.w2: m=-13 lossless=100.0000% mse(v3)=2.5727e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.105.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7319e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.91.w2: m=-13 lossless=100.0000% mse(v3)=2.4235e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.123.w2: m=-13 lossless=100.0000% mse(v3)=2.4299e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.48.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5676e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.75.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7091e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.29.w2: m=-13 lossless=100.0000% mse(v3)=2.4291e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.36.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5563e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.85.w2: m=-13 lossless=100.0000% mse(v3)=2.4709e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.104.w2: m=-13 lossless=100.0000% mse(v3)=2.4438e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.240.w2: m=-13 lossless=100.0000% mse(v3)=2.3494e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.213.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7439e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.32.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8251e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.71.w2: m=-13 lossless=100.0000% mse(v3)=2.5377e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.139.w2: m=-13 lossless=100.0000% mse(v3)=2.4840e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.189.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8831e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.83.w2: m=-13 lossless=100.0000% mse(v3)=2.5290e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.252.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7253e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.134.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6702e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.138.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9752e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.3.w2: m=-13 lossless=100.0000% mse(v3)=2.5445e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.211.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7818e-06 mse(v3.5)=0.0000e+00 layers.27.ffn.experts.43.w2: m=-13 lossless=100.0000% mse(v3)=2.4880e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.213.w2: m=-13 lossless=100.0000% mse(v3)=2.6697e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.207.w2: m=-13 lossless=100.0000% mse(v3)=2.6096e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.250.w2: m=-13 lossless=100.0000% mse(v3)=2.6040e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.130.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5122e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.171.w2: m=-13 lossless=100.0000% mse(v3)=2.4775e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.188.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5117e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.36.w1: m=-11 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7533e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.209.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6137e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.240.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7019e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.147.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5091e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.65.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5927e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.112.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5109e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.28.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5912e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.142.w2: m=-13 lossless=100.0000% mse(v3)=2.6190e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.224.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7040e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.11.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6529e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.144.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4558e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.195.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6599e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.129.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6504e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.118.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7915e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.144.w2: m=-13 lossless=100.0000% mse(v3)=2.4832e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.156.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5369e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.224.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5838e-06 mse(v3.5)=0.0000e+00 layers.28.ffn.experts.37.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5748e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.71.w2: m=-13 lossless=100.0000% mse(v3)=2.5108e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.112.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6580e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.119.w2: m=-13 lossless=100.0000% mse(v3)=2.5064e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.207.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7036e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.81.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6591e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.220.w2: m=-13 lossless=100.0000% mse(v3)=2.5893e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.15.w2: m=-13 lossless=100.0000% mse(v3)=2.3806e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.57.w2: m=-13 lossless=100.0000% mse(v3)=2.4157e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.236.w2: m=-13 lossless=100.0000% mse(v3)=2.5221e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.157.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5082e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.130.w2: m=-11 lossless=100.0000% mse(v3)=2.6748e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.101.w2: m=-12 lossless=100.0000% mse(v3)=2.6594e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.149.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5788e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.245.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5733e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.44.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7748e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.218.w2: m=-13 lossless=100.0000% mse(v3)=2.5375e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.93.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5214e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.130.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7975e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.167.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1856e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.125.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6519e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.79.w2: m=-13 lossless=100.0000% mse(v3)=2.5744e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.249.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5120e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.100.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6040e-06 mse(v3.5)=0.0000e+00 layers.29.ffn.experts.140.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4800e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.133.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.2455e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.64.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8470e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.247.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6190e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.229.w2: m=-13 lossless=100.0000% mse(v3)=2.6059e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.141.w2: m=-13 lossless=100.0000% mse(v3)=2.4636e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.175.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5957e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.28.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7472e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.99.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6321e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.83.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7465e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.117.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4428e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.228.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3498e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.37.w2: m=-13 lossless=100.0000% mse(v3)=2.7007e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.0.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7673e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.213.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5821e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.102.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6505e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.217.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5698e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.206.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6755e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.96.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5077e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.56.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6619e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.211.w2: m=-13 lossless=100.0000% mse(v3)=2.5676e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.102.w2: m=-13 lossless=100.0000% mse(v3)=2.4899e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.151.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7274e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.32.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8296e-06 mse(v3.5)=0.0000e+00 layers.30.ffn.experts.254.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4663e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.44.w2: m=-13 lossless=100.0000% mse(v3)=2.3710e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.20.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5624e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.183.w2: m=-13 lossless=100.0000% mse(v3)=2.5396e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.46.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6102e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.19.w2: m=-13 lossless=100.0000% mse(v3)=2.6403e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.249.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4509e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.188.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5157e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.111.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7002e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.160.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6723e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.232.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6221e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.122.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6831e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.74.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5765e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.8.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8158e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.31.w2: m=-13 lossless=100.0000% mse(v3)=2.7239e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.93.w2: m=-13 lossless=100.0000% mse(v3)=2.5269e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.166.w2: m=-12 lossless=100.0000% mse(v3)=2.6162e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.35.w2: m=-13 lossless=100.0000% mse(v3)=2.5121e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.204.w2: m=-11 lossless=100.0000% mse(v3)=3.5425e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.89.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5250e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.118.w2: m=-13 lossless=100.0000% mse(v3)=2.7835e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.160.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8362e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.238.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5611e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.13.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7116e-06 mse(v3.5)=0.0000e+00 layers.31.ffn.experts.161.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.2017e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.27.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8181e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.79.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4866e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.98.w2: m=-13 lossless=100.0000% mse(v3)=2.5354e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.108.w2: m=-13 lossless=100.0000% mse(v3)=2.9828e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.34.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0712e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.203.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8617e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.200.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5044e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.201.w2: m=-13 lossless=100.0000% mse(v3)=2.6942e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.73.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4714e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.87.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7690e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.31.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7714e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.90.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6743e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.207.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5288e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.79.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7206e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.13.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4130e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.221.w2: m=-13 lossless=100.0000% mse(v3)=2.4695e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.136.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6607e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.12.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4916e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.207.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7527e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.36.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.2709e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.17.w2: m=-13 lossless=100.0000% mse(v3)=2.5425e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.240.w2: m=-13 lossless=100.0000% mse(v3)=2.6381e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.198.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4154e-06 mse(v3.5)=0.0000e+00 layers.32.ffn.experts.116.w2: m=-13 lossless=100.0000% mse(v3)=2.5555e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.215.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1511e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.16.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7428e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.108.w2: m=-13 lossless=100.0000% mse(v3)=2.8744e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.198.w2: m=-13 lossless=100.0000% mse(v3)=2.5130e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.51.w2: m=-13 lossless=100.0000% mse(v3)=2.4792e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.234.w2: m=-13 lossless=100.0000% mse(v3)=2.5505e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.40.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1664e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.74.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6128e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.12.w2: m=-13 lossless=100.0000% mse(v3)=2.6041e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.27.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1570e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.10.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7933e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.176.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5739e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.179.w2: m=-13 lossless=100.0000% mse(v3)=2.5443e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.118.w2: m=-13 lossless=100.0000% mse(v3)=2.5438e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.241.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7782e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.194.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6612e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.246.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6677e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.33.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6651e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.208.w2: m=-13 lossless=100.0000% mse(v3)=2.4248e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.194.w2: m=-13 lossless=100.0000% mse(v3)=2.5350e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.29.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3640e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.122.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6134e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.22.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.2964e-06 mse(v3.5)=0.0000e+00 layers.33.ffn.experts.101.w2: m=-13 lossless=100.0000% mse(v3)=2.7560e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.116.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6772e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.91.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8352e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.1.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9173e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.97.w2: m=-13 lossless=100.0000% mse(v3)=2.5626e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.110.w2: m=-13 lossless=100.0000% mse(v3)=2.4596e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.55.w2: m=-13 lossless=100.0000% mse(v3)=2.5122e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.251.w2: m=-13 lossless=100.0000% mse(v3)=2.6979e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.252.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1240e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.3.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6599e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.127.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4279e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.74.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6373e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.229.w2: m=-13 lossless=100.0000% mse(v3)=2.7057e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.51.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6979e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.5.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0547e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.173.w2: m=-13 lossless=100.0000% mse(v3)=2.5471e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.221.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9691e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.68.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6581e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.246.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0673e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.138.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7060e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.219.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6610e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.113.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4777e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.133.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7421e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.49.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6172e-06 mse(v3.5)=0.0000e+00 layers.34.ffn.experts.127.w2: m=-13 lossless=100.0000% mse(v3)=2.4596e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.234.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6545e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.88.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6213e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.61.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5892e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.2.w2: m=-13 lossless=100.0000% mse(v3)=2.7121e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.213.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1891e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.23.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6698e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.248.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7660e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.13.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9492e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.159.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4350e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.92.w2: m=-13 lossless=100.0000% mse(v3)=2.5893e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.4.w2: m=-13 lossless=100.0000% mse(v3)=2.8338e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.87.w2: m=-13 lossless=100.0000% mse(v3)=2.5394e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.252.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5166e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.33.w2: m=-13 lossless=100.0000% mse(v3)=2.4231e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.80.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9771e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.253.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5109e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.81.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0519e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.113.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5136e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.52.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0766e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.197.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6673e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.46.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5264e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.255.w2: m=-12 lossless=100.0000% mse(v3)=2.8465e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.240.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7564e-06 mse(v3.5)=0.0000e+00 layers.35.ffn.experts.119.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8105e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.129.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5098e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.148.w2: m=-13 lossless=100.0000% mse(v3)=2.6730e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.109.w2: m=-13 lossless=100.0000% mse(v3)=2.5374e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.35.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6900e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.29.w2: m=-13 lossless=100.0000% mse(v3)=2.6099e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.62.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7545e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.193.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3500e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.126.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6217e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.181.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5528e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.98.w2: m=-13 lossless=100.0000% mse(v3)=2.7490e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.39.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.5384e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.70.w2: m=-13 lossless=100.0000% mse(v3)=2.5916e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.147.w2: m=-13 lossless=100.0000% mse(v3)=2.6442e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.209.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0440e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.5.w2: m=-13 lossless=100.0000% mse(v3)=2.4658e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.254.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6733e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.107.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5598e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.34.w2: m=-13 lossless=100.0000% mse(v3)=2.6296e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.144.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6569e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.230.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7327e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.209.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6773e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.198.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5076e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.57.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7784e-06 mse(v3.5)=0.0000e+00 layers.36.ffn.experts.43.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6725e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.194.w2: m=-13 lossless=100.0000% mse(v3)=2.5538e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.31.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1380e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.47.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6840e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.122.w2: m=-13 lossless=100.0000% mse(v3)=2.4842e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.17.w2: m=-13 lossless=100.0000% mse(v3)=2.6803e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.135.w2: m=-13 lossless=100.0000% mse(v3)=2.6029e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.112.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8959e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.195.w2: m=-13 lossless=100.0000% mse(v3)=3.2402e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.161.w2: m=-13 lossless=100.0000% mse(v3)=2.4720e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.211.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4045e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.196.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7633e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.188.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7322e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.116.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8278e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.254.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9008e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.87.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8493e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.192.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7034e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.138.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.2867e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.137.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6902e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.249.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6409e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.253.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7879e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.201.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8067e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.179.w2: m=-13 lossless=100.0000% mse(v3)=2.6987e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.227.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6481e-06 mse(v3.5)=0.0000e+00 layers.37.ffn.experts.148.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1002e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.117.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.3015e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.205.w2: m=-13 lossless=100.0000% mse(v3)=2.6521e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.53.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7761e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.16.w2: m=-13 lossless=100.0000% mse(v3)=2.7865e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.93.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7305e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.98.w2: m=-12 lossless=100.0000% mse(v3)=3.0904e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.79.w2: m=-13 lossless=100.0000% mse(v3)=2.5579e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.251.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5875e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.250.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9094e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.122.w2: m=-13 lossless=100.0000% mse(v3)=2.5314e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.148.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5759e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.145.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9868e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.162.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6152e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.47.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1884e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.159.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6681e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.80.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8694e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.237.w2: m=-12 lossless=100.0000% mse(v3)=3.3253e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.148.w2: m=-13 lossless=100.0000% mse(v3)=2.5658e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.101.w2: m=-13 lossless=100.0000% mse(v3)=2.5122e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.208.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4596e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.7.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8022e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.33.w2: m=-13 lossless=100.0000% mse(v3)=2.7882e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.246.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6634e-06 mse(v3.5)=0.0000e+00 layers.38.ffn.experts.103.w2: m=-13 lossless=100.0000% mse(v3)=2.5919e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.195.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1134e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.217.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7464e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.245.w2: m=-13 lossless=100.0000% mse(v3)=2.7754e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.68.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9561e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.125.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6324e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.208.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8416e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.87.w2: m=-12 lossless=100.0000% mse(v3)=2.3944e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.136.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4313e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.77.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5037e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.10.w2: m=-13 lossless=100.0000% mse(v3)=3.2877e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.171.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4834e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.100.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6918e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.53.w2: m=-13 lossless=100.0000% mse(v3)=3.1072e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.115.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7718e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.123.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6517e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.40.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3350e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.122.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2239e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.125.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9623e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.133.w2: m=-13 lossless=100.0000% mse(v3)=2.8841e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.93.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4729e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.65.w2: m=-12 lossless=100.0000% mse(v3)=2.7256e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.81.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6419e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.24.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.3362e-06 mse(v3.5)=0.0000e+00 layers.39.ffn.experts.5.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9862e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.74.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1462e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.97.w2: m=-12 lossless=100.0000% mse(v3)=2.7359e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.50.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6261e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.53.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9490e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.161.w2: m=-13 lossless=100.0000% mse(v3)=3.2877e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.119.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7009e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.24.w2: m=-13 lossless=100.0000% mse(v3)=3.1270e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.221.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6831e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.100.w2: m=-13 lossless=100.0000% mse(v3)=3.2468e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.29.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.3628e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.178.w2: m=-13 lossless=100.0000% mse(v3)=3.3030e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.194.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.3077e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.197.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6321e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.157.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8012e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.130.w2: m=-13 lossless=100.0000% mse(v3)=3.3695e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.129.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9386e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.147.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5142e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.166.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.8600e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.162.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7654e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.107.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4977e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.181.w2: m=-13 lossless=100.0000% mse(v3)=3.2573e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.105.w2: m=-12 lossless=100.0000% mse(v3)=3.5552e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.14.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4890e-06 mse(v3.5)=0.0000e+00 layers.40.ffn.experts.78.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5154e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.113.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9739e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.187.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9376e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.132.w2: m=-12 lossless=100.0000% mse(v3)=3.0484e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.188.w2: m=-12 lossless=100.0000% mse(v3)=3.1116e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.59.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0565e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.80.w2: m=-13 lossless=100.0000% mse(v3)=3.8103e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.187.w2: m=-12 lossless=100.0000% mse(v3)=3.4355e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.38.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9920e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.162.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7283e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.251.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7375e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.106.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4978e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.60.w2: m=-12 lossless=100.0000% mse(v3)=2.9574e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.19.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0824e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.247.w2: m=-12 lossless=100.0000% mse(v3)=3.9193e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.43.w2: m=-11 lossless=100.0000% mse(v3)=3.4124e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.23.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7401e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.217.w2: m=-12 lossless=100.0000% mse(v3)=2.8673e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.158.w2: m=-12 lossless=100.0000% mse(v3)=2.9780e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.202.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5400e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.44.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.5229e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.64.w2: m=-11 lossless=100.0000% mse(v3)=2.7366e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.128.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0380e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.136.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9243e-06 mse(v3.5)=0.0000e+00 layers.41.ffn.experts.242.w2: m=-13 lossless=100.0000% mse(v3)=3.0659e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.111.w2: m=-11 lossless=100.0000% mse(v3)=4.3578e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.120.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.6182e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.203.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4073e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.208.w2: m=-11 lossless=100.0000% mse(v3)=3.3982e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.218.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9306e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.165.w2: m=-10 lossless=100.0000% mse(v3)=3.8720e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.137.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=4.1931e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.171.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4128e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.174.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4216e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.229.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=4.2014e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.82.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.2326e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.206.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.3971e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.131.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.3146e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.162.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4440e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.59.w2: m=-12 lossless=100.0000% mse(v3)=3.9995e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.140.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6379e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.201.w2: m=-12 lossless=100.0000% mse(v3)=4.0319e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.224.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4803e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.110.w2: m=-11 lossless=100.0000% mse(v3)=3.6738e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.210.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.8273e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.135.w2: m=-11 lossless=100.0000% mse(v3)=3.8635e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.152.w1: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.3736e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.136.w2: m=-12 lossless=100.0000% mse(v3)=3.3677e-06 mse(v3.5)=0.0000e+00 layers.42.ffn.experts.174.w2: m=-11 lossless=100.0000% mse(v3)=4.4448e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.65.w2: m=-11 lossless=100.0000% mse(v3)=3.5967e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.17.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4920e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.123.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.9330e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.235.w2: m=-11 lossless=100.0000% mse(v3)=3.1240e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.20.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.6347e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.247.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.4868e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.27.w2: m=-11 lossless=100.0000% mse(v3)=3.2611e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.208.w2: m=-8 lossless=100.0000% mse(v3)=6.3812e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.89.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.3941e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.113.w2: m=-10 lossless=100.0000% mse(v3)=3.1461e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.230.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.1400e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.125.w2: m=-11 lossless=100.0000% mse(v3)=3.0517e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.104.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.5243e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.118.w2: m=-11 lossless=100.0000% mse(v3)=4.2200e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.232.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.4117e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.43.w2: m=-11 lossless=100.0000% mse(v3)=3.5509e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.200.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7208e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.130.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.0951e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.86.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.3949e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.125.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.7161e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.76.w1: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=2.2871e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.92.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.6225e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.3.w3: m=-13 (w1/w3-shared) lossless=100.0000% mse(v3)=3.5551e-06 mse(v3.5)=0.0000e+00 mtp.0.ffn.experts.66.w3: m=-12 (w1/w3-shared) lossless=100.0000% mse(v3)=3.2592e-06 mse(v3.5)=0.0000e+00 ======================================================================== [cast_mxfp4_to_nvfp4] processed 33792 expert weight tensors across 46 shards [cast_mxfp4_to_nvfp4] lossless layers: 33792/33792 (100.00%) [cast_mxfp4_to_nvfp4] lossless blocks: 17716740096/17716740096 (100.0000%) [cast_mxfp4_to_nvfp4] MSE vs MXFP4: v3 (sampled n=1056): mean=2.6511e-06 max=6.3812e-06 [cast_mxfp4_to_nvfp4] MSE vs MXFP4: v3.5 (n=0 OOR + 33792 lossless): mean=0.0000e+00 max=0.0000e+00 ========================================================================