=== Quantization Report === Model: /root/src Date: 2026-09-25 13:15:32 Version: 0.14.3 CLI Arguments: --batch_size 4 --iters 1000 --nsamples 512 --seqlen 2048 --dataset opencode-instruct --output_dir /workspace/out/src-int4-AutoRound Memory Summary: Peak RAM: 28.63 GB Peak VRAM: 42.22 GB Sensitivity Analysis: ───────────────────────────────────────────────────────────────────────────────────────────────────── Layer Cosine Sim PSNR (dB) Iters uLoss Status ───────────────────────────────────────────────────────────────────────────────────────────────────── 🟢 model.language_model.layers.0 0.9999 82.0 999 69 → 4 PASS 🟢 model.language_model.layers.1 0.9999 83.7 988 36 → 7 PASS 🟢 model.language_model.layers.2 0.9999 83.1 799 38 → 12 PASS 🟢 model.language_model.layers.3 0.9999 82.8 682 73 → 16 PASS 🟢 model.language_model.layers.4 0.9999 82.5 821 43 → 19 PASS 🟢 model.language_model.layers.5 0.9999 81.9 959 57 → 23 PASS 🟢 model.language_model.layers.6 0.9999 82.3 859 92 → 32 PASS 🟢 model.language_model.layers.7 0.9999 81.2 944 202 → 42 PASS 🟢 model.language_model.layers.8 0.9999 81.0 783 107 → 49 PASS 🟢 model.language_model.layers.9 0.9999 81.3 700 128 → 59 PASS 🟢 model.language_model.layers.10 0.9999 82.8 979 163 → 66 PASS 🟢 model.language_model.layers.11 0.9999 81.6 960 220 → 85 PASS 🟢 model.language_model.layers.12 0.9999 80.9 759 244 → 96 PASS 🟢 model.language_model.layers.13 0.9999 80.7 795 229 → 115 PASS 🟢 model.language_model.layers.14 0.9999 81.8 715 288 → 131 PASS 🟢 model.language_model.layers.15 0.9999 80.8 890 378 → 154 PASS 🟢 model.language_model.layers.16 0.9998 80.0 979 360 → 170 PASS 🟢 model.language_model.layers.17 0.9998 79.4 907 437 → 196 PASS 🟢 model.language_model.layers.18 0.9998 80.2 986 588 → 226 PASS 🟢 model.language_model.layers.19 0.9997 77.8 987 824 → 338 PASS 🟢 model.language_model.layers.20 0.9996 75.5 713 907 → 384 PASS 🟢 model.language_model.layers.21 0.9996 78.1 601 1146 → 486 PASS 🟢 model.language_model.layers.22 0.9996 77.6 714 1285 → 542 PASS 🟢 model.language_model.layers.23 0.9995 77.2 908 1428 → 630 PASS 🟢 model.language_model.layers.24 0.9995 76.8 932 1468 → 670 PASS 🟢 model.language_model.layers.25 0.9994 76.9 793 1936 → 714 PASS 🟢 model.language_model.layers.26 0.9994 76.3 633 1891 → 901 PASS 🟢 model.language_model.layers.27 0.9993 75.5 726 2695 → 1067 PASS 🟢 model.language_model.layers.28 0.9992 75.0 603 2893 → 1181 PASS 🟢 model.language_model.layers.29 0.9991 74.6 897 2870 → 1315 PASS 🟢 model.language_model.layers.30 0.9991 73.9 531 3240 → 1528 PASS 🟢 model.language_model.layers.31 0.9988 72.8 415 4654 → 2064 PASS 🟢 model.language_model.layers.32 0.9987 72.6 756 4986 → 2218 PASS 🟢 model.language_model.layers.33 0.9986 72.4 639 4537 → 2301 PASS 🟢 model.language_model.layers.34 0.9986 71.9 938 6690 → 2416 PASS 🟢 model.language_model.layers.35 0.9981 70.5 937 8413 → 3276 PASS 🟢 model.language_model.layers.36 0.9979 70.2 328 6633 → 3602 PASS 🟢 model.language_model.layers.37 0.9978 69.8 982 7504 → 4095 PASS 🟢 model.language_model.layers.38 0.9979 69.6 408 6049 → 3881 PASS 🟢 model.language_model.layers.39 0.9975 68.7 282 8702 → 4947 PASS 🟢 model.language_model.layers.40 0.9974 68.5 251 7994 → 5080 PASS 🟢 model.language_model.layers.41 0.9973 68.3 969 9446 → 5402 PASS 🟢 model.language_model.layers.42 0.9967 66.6 728 13536 → 6706 PASS 🟢 model.language_model.layers.43 0.9960 65.5 144 18873 → 8889 PASS 🟢 model.language_model.layers.44 0.9958 65.1 994 18136 → 9515 PASS 🟢 model.language_model.layers.45 0.9957 64.7 817 16926 → 10500 PASS 🟢 model.language_model.layers.46 0.9954 63.5 492 18031 → 12803 PASS 🟢 model.language_model.layers.47 0.9949 61.6 852 27443 → 16631 PASS 🟢 model.language_model.layers.48 0.9943 59.7 391 27729 → 18863 PASS 🟢 model.language_model.layers.49 0.9939 58.9 657 30044 → 21714 PASS 🟢 model.language_model.layers.50 0.9924 57.9 830 40183 → 28802 PASS 🟢 model.language_model.layers.51 0.9906 53.3 397 79753 → 47603 PASS 🟢 model.language_model.layers.52 0.9902 52.9 731 90248 → 54403 PASS 🟠 model.language_model.layers.53 0.9894 53.0 787 101130 → 73957 WARN 🟠 model.language_model.layers.54 0.9859 64.2 275 213543 → 111245 WARN 🟠 model.language_model.layers.55 0.9877 61.3 769 179536 → 99929 WARN 🟠 model.language_model.layers.56 0.9877 60.2 698 189011 → 123630 WARN 🟠 model.language_model.layers.57 0.9874 60.0 158 217096 → 142504 WARN 🟠 model.language_model.layers.58 0.9827 64.5 548 321141 → 205431 WARN 🟠 model.language_model.layers.59 0.9883 54.5 210 286896 → 194315 WARN 🟠 model.language_model.layers.60 0.9888 52.6 995 357349 → 214263 WARN 🟠 model.language_model.layers.61 0.9892 52.9 967 400132 → 259227 WARN 🟠 model.language_model.layers.62 0.9874 52.3 748 459020 → 349781 WARN 🟠 model.language_model.layers.63 0.9888 56.0 461 772124 → 516162 WARN Summary: Total blocks: 64 Passed (🟢): 53 Warning (🟠): 11 Thresholds: Cosine Similarity < 0.99, PSNR < 45.0 dB