NicoNico6
commited on
Commit
·
db2b361
1
Parent(s):
7a7c71f
update
Browse files- model.safetensors +2 -2
- quant_strategy.json +93 -93
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f49bd33dea3023b79936ea738a31c465f42f0d369f2387e61b8ef6ae3a1a7489
|
| 3 |
+
size 2817405544
|
quant_strategy.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"measurement": {
|
| 3 |
"model.layers.0": {
|
| 4 |
-
"accuracy": 0.
|
| 5 |
-
"total_bits":
|
| 6 |
"q_proj": {
|
| 7 |
"group_size": {
|
| 8 |
"2": 64
|
|
@@ -17,10 +17,10 @@
|
|
| 17 |
},
|
| 18 |
"k_proj": {
|
| 19 |
"group_size": {
|
| 20 |
-
"
|
| 21 |
},
|
| 22 |
"bits": [
|
| 23 |
-
|
| 24 |
],
|
| 25 |
"bits_prop": [
|
| 26 |
1
|
|
@@ -41,10 +41,10 @@
|
|
| 41 |
},
|
| 42 |
"o_proj": {
|
| 43 |
"group_size": {
|
| 44 |
-
"
|
| 45 |
},
|
| 46 |
"bits": [
|
| 47 |
-
|
| 48 |
],
|
| 49 |
"bits_prop": [
|
| 50 |
1
|
|
@@ -77,10 +77,10 @@
|
|
| 77 |
},
|
| 78 |
"down_proj": {
|
| 79 |
"group_size": {
|
| 80 |
-
"
|
| 81 |
},
|
| 82 |
"bits": [
|
| 83 |
-
|
| 84 |
],
|
| 85 |
"bits_prop": [
|
| 86 |
1
|
|
@@ -89,8 +89,8 @@
|
|
| 89 |
}
|
| 90 |
},
|
| 91 |
"model.layers.1": {
|
| 92 |
-
"accuracy": 0.
|
| 93 |
-
"total_bits":
|
| 94 |
"q_proj": {
|
| 95 |
"group_size": {
|
| 96 |
"2": 64
|
|
@@ -165,10 +165,10 @@
|
|
| 165 |
},
|
| 166 |
"down_proj": {
|
| 167 |
"group_size": {
|
| 168 |
-
"
|
| 169 |
},
|
| 170 |
"bits": [
|
| 171 |
-
|
| 172 |
],
|
| 173 |
"bits_prop": [
|
| 174 |
1
|
|
@@ -177,7 +177,7 @@
|
|
| 177 |
}
|
| 178 |
},
|
| 179 |
"model.layers.2": {
|
| 180 |
-
"accuracy": 0.
|
| 181 |
"total_bits": 458124288,
|
| 182 |
"q_proj": {
|
| 183 |
"group_size": {
|
|
@@ -265,8 +265,8 @@
|
|
| 265 |
}
|
| 266 |
},
|
| 267 |
"model.layers.3": {
|
| 268 |
-
"accuracy": 0.
|
| 269 |
-
"total_bits":
|
| 270 |
"q_proj": {
|
| 271 |
"group_size": {
|
| 272 |
"2": 64
|
|
@@ -281,10 +281,10 @@
|
|
| 281 |
},
|
| 282 |
"k_proj": {
|
| 283 |
"group_size": {
|
| 284 |
-
"
|
| 285 |
},
|
| 286 |
"bits": [
|
| 287 |
-
|
| 288 |
],
|
| 289 |
"bits_prop": [
|
| 290 |
1
|
|
@@ -353,7 +353,7 @@
|
|
| 353 |
}
|
| 354 |
},
|
| 355 |
"model.layers.4": {
|
| 356 |
-
"accuracy": 0.
|
| 357 |
"total_bits": 458124288,
|
| 358 |
"q_proj": {
|
| 359 |
"group_size": {
|
|
@@ -441,7 +441,7 @@
|
|
| 441 |
}
|
| 442 |
},
|
| 443 |
"model.layers.5": {
|
| 444 |
-
"accuracy": 0.
|
| 445 |
"total_bits": 458124288,
|
| 446 |
"q_proj": {
|
| 447 |
"group_size": {
|
|
@@ -529,7 +529,7 @@
|
|
| 529 |
}
|
| 530 |
},
|
| 531 |
"model.layers.6": {
|
| 532 |
-
"accuracy": 0.
|
| 533 |
"total_bits": 458124288,
|
| 534 |
"q_proj": {
|
| 535 |
"group_size": {
|
|
@@ -617,7 +617,7 @@
|
|
| 617 |
}
|
| 618 |
},
|
| 619 |
"model.layers.7": {
|
| 620 |
-
"accuracy": 0.
|
| 621 |
"total_bits": 458124288,
|
| 622 |
"q_proj": {
|
| 623 |
"group_size": {
|
|
@@ -705,8 +705,8 @@
|
|
| 705 |
}
|
| 706 |
},
|
| 707 |
"model.layers.8": {
|
| 708 |
-
"accuracy": 0.
|
| 709 |
-
"total_bits":
|
| 710 |
"q_proj": {
|
| 711 |
"group_size": {
|
| 712 |
"2": 64
|
|
@@ -745,10 +745,10 @@
|
|
| 745 |
},
|
| 746 |
"o_proj": {
|
| 747 |
"group_size": {
|
| 748 |
-
"
|
| 749 |
},
|
| 750 |
"bits": [
|
| 751 |
-
|
| 752 |
],
|
| 753 |
"bits_prop": [
|
| 754 |
1
|
|
@@ -793,7 +793,7 @@
|
|
| 793 |
}
|
| 794 |
},
|
| 795 |
"model.layers.9": {
|
| 796 |
-
"accuracy": 0.
|
| 797 |
"total_bits": 458124288,
|
| 798 |
"q_proj": {
|
| 799 |
"group_size": {
|
|
@@ -881,7 +881,7 @@
|
|
| 881 |
}
|
| 882 |
},
|
| 883 |
"model.layers.10": {
|
| 884 |
-
"accuracy": 0.
|
| 885 |
"total_bits": 458124288,
|
| 886 |
"q_proj": {
|
| 887 |
"group_size": {
|
|
@@ -969,7 +969,7 @@
|
|
| 969 |
}
|
| 970 |
},
|
| 971 |
"model.layers.11": {
|
| 972 |
-
"accuracy": 0.
|
| 973 |
"total_bits": 458124288,
|
| 974 |
"q_proj": {
|
| 975 |
"group_size": {
|
|
@@ -1057,14 +1057,14 @@
|
|
| 1057 |
}
|
| 1058 |
},
|
| 1059 |
"model.layers.12": {
|
| 1060 |
-
"accuracy": 0.
|
| 1061 |
-
"total_bits":
|
| 1062 |
"q_proj": {
|
| 1063 |
"group_size": {
|
| 1064 |
-
"
|
| 1065 |
},
|
| 1066 |
"bits": [
|
| 1067 |
-
|
| 1068 |
],
|
| 1069 |
"bits_prop": [
|
| 1070 |
1
|
|
@@ -1073,10 +1073,10 @@
|
|
| 1073 |
},
|
| 1074 |
"k_proj": {
|
| 1075 |
"group_size": {
|
| 1076 |
-
"
|
| 1077 |
},
|
| 1078 |
"bits": [
|
| 1079 |
-
|
| 1080 |
],
|
| 1081 |
"bits_prop": [
|
| 1082 |
1
|
|
@@ -1145,7 +1145,7 @@
|
|
| 1145 |
}
|
| 1146 |
},
|
| 1147 |
"model.layers.13": {
|
| 1148 |
-
"accuracy": 0.
|
| 1149 |
"total_bits": 458124288,
|
| 1150 |
"q_proj": {
|
| 1151 |
"group_size": {
|
|
@@ -1233,7 +1233,7 @@
|
|
| 1233 |
}
|
| 1234 |
},
|
| 1235 |
"model.layers.14": {
|
| 1236 |
-
"accuracy": 0.
|
| 1237 |
"total_bits": 458124288,
|
| 1238 |
"q_proj": {
|
| 1239 |
"group_size": {
|
|
@@ -1321,7 +1321,7 @@
|
|
| 1321 |
}
|
| 1322 |
},
|
| 1323 |
"model.layers.15": {
|
| 1324 |
-
"accuracy": 0.
|
| 1325 |
"total_bits": 458124288,
|
| 1326 |
"q_proj": {
|
| 1327 |
"group_size": {
|
|
@@ -1409,7 +1409,7 @@
|
|
| 1409 |
}
|
| 1410 |
},
|
| 1411 |
"model.layers.16": {
|
| 1412 |
-
"accuracy": 0.
|
| 1413 |
"total_bits": 458124288,
|
| 1414 |
"q_proj": {
|
| 1415 |
"group_size": {
|
|
@@ -1497,8 +1497,8 @@
|
|
| 1497 |
}
|
| 1498 |
},
|
| 1499 |
"model.layers.17": {
|
| 1500 |
-
"accuracy": 0.
|
| 1501 |
-
"total_bits":
|
| 1502 |
"q_proj": {
|
| 1503 |
"group_size": {
|
| 1504 |
"2": 64
|
|
@@ -1513,10 +1513,10 @@
|
|
| 1513 |
},
|
| 1514 |
"k_proj": {
|
| 1515 |
"group_size": {
|
| 1516 |
-
"
|
| 1517 |
},
|
| 1518 |
"bits": [
|
| 1519 |
-
|
| 1520 |
],
|
| 1521 |
"bits_prop": [
|
| 1522 |
1
|
|
@@ -1537,10 +1537,10 @@
|
|
| 1537 |
},
|
| 1538 |
"o_proj": {
|
| 1539 |
"group_size": {
|
| 1540 |
-
"
|
| 1541 |
},
|
| 1542 |
"bits": [
|
| 1543 |
-
|
| 1544 |
],
|
| 1545 |
"bits_prop": [
|
| 1546 |
1
|
|
@@ -1585,8 +1585,8 @@
|
|
| 1585 |
}
|
| 1586 |
},
|
| 1587 |
"model.layers.18": {
|
| 1588 |
-
"accuracy": 0.
|
| 1589 |
-
"total_bits":
|
| 1590 |
"q_proj": {
|
| 1591 |
"group_size": {
|
| 1592 |
"2": 64
|
|
@@ -1601,10 +1601,10 @@
|
|
| 1601 |
},
|
| 1602 |
"k_proj": {
|
| 1603 |
"group_size": {
|
| 1604 |
-
"
|
| 1605 |
},
|
| 1606 |
"bits": [
|
| 1607 |
-
|
| 1608 |
],
|
| 1609 |
"bits_prop": [
|
| 1610 |
1
|
|
@@ -1625,10 +1625,10 @@
|
|
| 1625 |
},
|
| 1626 |
"o_proj": {
|
| 1627 |
"group_size": {
|
| 1628 |
-
"
|
| 1629 |
},
|
| 1630 |
"bits": [
|
| 1631 |
-
|
| 1632 |
],
|
| 1633 |
"bits_prop": [
|
| 1634 |
1
|
|
@@ -1673,8 +1673,8 @@
|
|
| 1673 |
}
|
| 1674 |
},
|
| 1675 |
"model.layers.19": {
|
| 1676 |
-
"accuracy": 0.
|
| 1677 |
-
"total_bits":
|
| 1678 |
"q_proj": {
|
| 1679 |
"group_size": {
|
| 1680 |
"2": 64
|
|
@@ -1713,10 +1713,10 @@
|
|
| 1713 |
},
|
| 1714 |
"o_proj": {
|
| 1715 |
"group_size": {
|
| 1716 |
-
"
|
| 1717 |
},
|
| 1718 |
"bits": [
|
| 1719 |
-
|
| 1720 |
],
|
| 1721 |
"bits_prop": [
|
| 1722 |
1
|
|
@@ -1761,8 +1761,8 @@
|
|
| 1761 |
}
|
| 1762 |
},
|
| 1763 |
"model.layers.20": {
|
| 1764 |
-
"accuracy": 0.
|
| 1765 |
-
"total_bits":
|
| 1766 |
"q_proj": {
|
| 1767 |
"group_size": {
|
| 1768 |
"2": 64
|
|
@@ -1777,10 +1777,10 @@
|
|
| 1777 |
},
|
| 1778 |
"k_proj": {
|
| 1779 |
"group_size": {
|
| 1780 |
-
"
|
| 1781 |
},
|
| 1782 |
"bits": [
|
| 1783 |
-
|
| 1784 |
],
|
| 1785 |
"bits_prop": [
|
| 1786 |
1
|
|
@@ -1849,7 +1849,7 @@
|
|
| 1849 |
}
|
| 1850 |
},
|
| 1851 |
"model.layers.21": {
|
| 1852 |
-
"accuracy": 0.
|
| 1853 |
"total_bits": 458124288,
|
| 1854 |
"q_proj": {
|
| 1855 |
"group_size": {
|
|
@@ -1937,7 +1937,7 @@
|
|
| 1937 |
}
|
| 1938 |
},
|
| 1939 |
"model.layers.22": {
|
| 1940 |
-
"accuracy": 0.
|
| 1941 |
"total_bits": 458124288,
|
| 1942 |
"q_proj": {
|
| 1943 |
"group_size": {
|
|
@@ -2025,8 +2025,8 @@
|
|
| 2025 |
}
|
| 2026 |
},
|
| 2027 |
"model.layers.23": {
|
| 2028 |
-
"accuracy": 0.
|
| 2029 |
-
"total_bits":
|
| 2030 |
"q_proj": {
|
| 2031 |
"group_size": {
|
| 2032 |
"2": 64
|
|
@@ -2041,10 +2041,10 @@
|
|
| 2041 |
},
|
| 2042 |
"k_proj": {
|
| 2043 |
"group_size": {
|
| 2044 |
-
"
|
| 2045 |
},
|
| 2046 |
"bits": [
|
| 2047 |
-
|
| 2048 |
],
|
| 2049 |
"bits_prop": [
|
| 2050 |
1
|
|
@@ -2113,7 +2113,7 @@
|
|
| 2113 |
}
|
| 2114 |
},
|
| 2115 |
"model.layers.24": {
|
| 2116 |
-
"accuracy": 0.
|
| 2117 |
"total_bits": 458124288,
|
| 2118 |
"q_proj": {
|
| 2119 |
"group_size": {
|
|
@@ -2201,7 +2201,7 @@
|
|
| 2201 |
}
|
| 2202 |
},
|
| 2203 |
"model.layers.25": {
|
| 2204 |
-
"accuracy": 0.
|
| 2205 |
"total_bits": 458124288,
|
| 2206 |
"q_proj": {
|
| 2207 |
"group_size": {
|
|
@@ -2289,8 +2289,8 @@
|
|
| 2289 |
}
|
| 2290 |
},
|
| 2291 |
"model.layers.26": {
|
| 2292 |
-
"accuracy": 0.
|
| 2293 |
-
"total_bits":
|
| 2294 |
"q_proj": {
|
| 2295 |
"group_size": {
|
| 2296 |
"2": 64
|
|
@@ -2305,10 +2305,10 @@
|
|
| 2305 |
},
|
| 2306 |
"k_proj": {
|
| 2307 |
"group_size": {
|
| 2308 |
-
"
|
| 2309 |
},
|
| 2310 |
"bits": [
|
| 2311 |
-
|
| 2312 |
],
|
| 2313 |
"bits_prop": [
|
| 2314 |
1
|
|
@@ -2377,8 +2377,8 @@
|
|
| 2377 |
}
|
| 2378 |
},
|
| 2379 |
"model.layers.27": {
|
| 2380 |
-
"accuracy": 0.
|
| 2381 |
-
"total_bits":
|
| 2382 |
"q_proj": {
|
| 2383 |
"group_size": {
|
| 2384 |
"2": 64
|
|
@@ -2393,10 +2393,10 @@
|
|
| 2393 |
},
|
| 2394 |
"k_proj": {
|
| 2395 |
"group_size": {
|
| 2396 |
-
"
|
| 2397 |
},
|
| 2398 |
"bits": [
|
| 2399 |
-
|
| 2400 |
],
|
| 2401 |
"bits_prop": [
|
| 2402 |
1
|
|
@@ -2465,7 +2465,7 @@
|
|
| 2465 |
}
|
| 2466 |
},
|
| 2467 |
"model.layers.28": {
|
| 2468 |
-
"accuracy": 0.
|
| 2469 |
"total_bits": 458124288,
|
| 2470 |
"q_proj": {
|
| 2471 |
"group_size": {
|
|
@@ -2553,8 +2553,8 @@
|
|
| 2553 |
}
|
| 2554 |
},
|
| 2555 |
"model.layers.29": {
|
| 2556 |
-
"accuracy": 0.
|
| 2557 |
-
"total_bits":
|
| 2558 |
"q_proj": {
|
| 2559 |
"group_size": {
|
| 2560 |
"2": 64
|
|
@@ -2629,10 +2629,10 @@
|
|
| 2629 |
},
|
| 2630 |
"down_proj": {
|
| 2631 |
"group_size": {
|
| 2632 |
-
"
|
| 2633 |
},
|
| 2634 |
"bits": [
|
| 2635 |
-
|
| 2636 |
],
|
| 2637 |
"bits_prop": [
|
| 2638 |
1
|
|
@@ -2641,8 +2641,8 @@
|
|
| 2641 |
}
|
| 2642 |
},
|
| 2643 |
"model.layers.30": {
|
| 2644 |
-
"accuracy": 0.
|
| 2645 |
-
"total_bits":
|
| 2646 |
"q_proj": {
|
| 2647 |
"group_size": {
|
| 2648 |
"2": 64
|
|
@@ -2681,10 +2681,10 @@
|
|
| 2681 |
},
|
| 2682 |
"o_proj": {
|
| 2683 |
"group_size": {
|
| 2684 |
-
"
|
| 2685 |
},
|
| 2686 |
"bits": [
|
| 2687 |
-
|
| 2688 |
],
|
| 2689 |
"bits_prop": [
|
| 2690 |
1
|
|
@@ -2729,8 +2729,8 @@
|
|
| 2729 |
}
|
| 2730 |
},
|
| 2731 |
"model.layers.31": {
|
| 2732 |
-
"accuracy": 0.
|
| 2733 |
-
"total_bits":
|
| 2734 |
"q_proj": {
|
| 2735 |
"group_size": {
|
| 2736 |
"2": 64
|
|
@@ -2745,10 +2745,10 @@
|
|
| 2745 |
},
|
| 2746 |
"k_proj": {
|
| 2747 |
"group_size": {
|
| 2748 |
-
"
|
| 2749 |
},
|
| 2750 |
"bits": [
|
| 2751 |
-
|
| 2752 |
],
|
| 2753 |
"bits_prop": [
|
| 2754 |
1
|
|
@@ -2769,10 +2769,10 @@
|
|
| 2769 |
},
|
| 2770 |
"o_proj": {
|
| 2771 |
"group_size": {
|
| 2772 |
-
"
|
| 2773 |
},
|
| 2774 |
"bits": [
|
| 2775 |
-
|
| 2776 |
],
|
| 2777 |
"bits_prop": [
|
| 2778 |
1
|
|
@@ -2781,10 +2781,10 @@
|
|
| 2781 |
},
|
| 2782 |
"up_proj": {
|
| 2783 |
"group_size": {
|
| 2784 |
-
"
|
| 2785 |
},
|
| 2786 |
"bits": [
|
| 2787 |
-
|
| 2788 |
],
|
| 2789 |
"bits_prop": [
|
| 2790 |
1
|
|
@@ -2793,10 +2793,10 @@
|
|
| 2793 |
},
|
| 2794 |
"gate_proj": {
|
| 2795 |
"group_size": {
|
| 2796 |
-
"
|
| 2797 |
},
|
| 2798 |
"bits": [
|
| 2799 |
-
|
| 2800 |
],
|
| 2801 |
"bits_prop": [
|
| 2802 |
1
|
|
|
|
| 1 |
{
|
| 2 |
"measurement": {
|
| 3 |
"model.layers.0": {
|
| 4 |
+
"accuracy": 0.7869586944580078,
|
| 5 |
+
"total_bits": 466380288,
|
| 6 |
"q_proj": {
|
| 7 |
"group_size": {
|
| 8 |
"2": 64
|
|
|
|
| 17 |
},
|
| 18 |
"k_proj": {
|
| 19 |
"group_size": {
|
| 20 |
+
"4": 128
|
| 21 |
},
|
| 22 |
"bits": [
|
| 23 |
+
4
|
| 24 |
],
|
| 25 |
"bits_prop": [
|
| 26 |
1
|
|
|
|
| 41 |
},
|
| 42 |
"o_proj": {
|
| 43 |
"group_size": {
|
| 44 |
+
"2": 64
|
| 45 |
},
|
| 46 |
"bits": [
|
| 47 |
+
2
|
| 48 |
],
|
| 49 |
"bits_prop": [
|
| 50 |
1
|
|
|
|
| 77 |
},
|
| 78 |
"down_proj": {
|
| 79 |
"group_size": {
|
| 80 |
+
"2": 64
|
| 81 |
},
|
| 82 |
"bits": [
|
| 83 |
+
2
|
| 84 |
],
|
| 85 |
"bits_prop": [
|
| 86 |
1
|
|
|
|
| 89 |
}
|
| 90 |
},
|
| 91 |
"model.layers.1": {
|
| 92 |
+
"accuracy": 0.647979736328125,
|
| 93 |
+
"total_bits": 458124288,
|
| 94 |
"q_proj": {
|
| 95 |
"group_size": {
|
| 96 |
"2": 64
|
|
|
|
| 165 |
},
|
| 166 |
"down_proj": {
|
| 167 |
"group_size": {
|
| 168 |
+
"2": 64
|
| 169 |
},
|
| 170 |
"bits": [
|
| 171 |
+
2
|
| 172 |
],
|
| 173 |
"bits_prop": [
|
| 174 |
1
|
|
|
|
| 177 |
}
|
| 178 |
},
|
| 179 |
"model.layers.2": {
|
| 180 |
+
"accuracy": 0.9489546418190002,
|
| 181 |
"total_bits": 458124288,
|
| 182 |
"q_proj": {
|
| 183 |
"group_size": {
|
|
|
|
| 265 |
}
|
| 266 |
},
|
| 267 |
"model.layers.3": {
|
| 268 |
+
"accuracy": 0.9539257287979126,
|
| 269 |
+
"total_bits": 458124288,
|
| 270 |
"q_proj": {
|
| 271 |
"group_size": {
|
| 272 |
"2": 64
|
|
|
|
| 281 |
},
|
| 282 |
"k_proj": {
|
| 283 |
"group_size": {
|
| 284 |
+
"2": 64
|
| 285 |
},
|
| 286 |
"bits": [
|
| 287 |
+
2
|
| 288 |
],
|
| 289 |
"bits_prop": [
|
| 290 |
1
|
|
|
|
| 353 |
}
|
| 354 |
},
|
| 355 |
"model.layers.4": {
|
| 356 |
+
"accuracy": 0.948026180267334,
|
| 357 |
"total_bits": 458124288,
|
| 358 |
"q_proj": {
|
| 359 |
"group_size": {
|
|
|
|
| 441 |
}
|
| 442 |
},
|
| 443 |
"model.layers.5": {
|
| 444 |
+
"accuracy": 0.938827395439148,
|
| 445 |
"total_bits": 458124288,
|
| 446 |
"q_proj": {
|
| 447 |
"group_size": {
|
|
|
|
| 529 |
}
|
| 530 |
},
|
| 531 |
"model.layers.6": {
|
| 532 |
+
"accuracy": 0.9337625503540039,
|
| 533 |
"total_bits": 458124288,
|
| 534 |
"q_proj": {
|
| 535 |
"group_size": {
|
|
|
|
| 617 |
}
|
| 618 |
},
|
| 619 |
"model.layers.7": {
|
| 620 |
+
"accuracy": 0.9189379215240479,
|
| 621 |
"total_bits": 458124288,
|
| 622 |
"q_proj": {
|
| 623 |
"group_size": {
|
|
|
|
| 705 |
}
|
| 706 |
},
|
| 707 |
"model.layers.8": {
|
| 708 |
+
"accuracy": 0.9234421253204346,
|
| 709 |
+
"total_bits": 458124288,
|
| 710 |
"q_proj": {
|
| 711 |
"group_size": {
|
| 712 |
"2": 64
|
|
|
|
| 745 |
},
|
| 746 |
"o_proj": {
|
| 747 |
"group_size": {
|
| 748 |
+
"2": 64
|
| 749 |
},
|
| 750 |
"bits": [
|
| 751 |
+
2
|
| 752 |
],
|
| 753 |
"bits_prop": [
|
| 754 |
1
|
|
|
|
| 793 |
}
|
| 794 |
},
|
| 795 |
"model.layers.9": {
|
| 796 |
+
"accuracy": 0.9196293354034424,
|
| 797 |
"total_bits": 458124288,
|
| 798 |
"q_proj": {
|
| 799 |
"group_size": {
|
|
|
|
| 881 |
}
|
| 882 |
},
|
| 883 |
"model.layers.10": {
|
| 884 |
+
"accuracy": 0.9117538928985596,
|
| 885 |
"total_bits": 458124288,
|
| 886 |
"q_proj": {
|
| 887 |
"group_size": {
|
|
|
|
| 969 |
}
|
| 970 |
},
|
| 971 |
"model.layers.11": {
|
| 972 |
+
"accuracy": 0.9103360176086426,
|
| 973 |
"total_bits": 458124288,
|
| 974 |
"q_proj": {
|
| 975 |
"group_size": {
|
|
|
|
| 1057 |
}
|
| 1058 |
},
|
| 1059 |
"model.layers.12": {
|
| 1060 |
+
"accuracy": 0.9095911979675293,
|
| 1061 |
+
"total_bits": 466380288,
|
| 1062 |
"q_proj": {
|
| 1063 |
"group_size": {
|
| 1064 |
+
"2": 64
|
| 1065 |
},
|
| 1066 |
"bits": [
|
| 1067 |
+
2
|
| 1068 |
],
|
| 1069 |
"bits_prop": [
|
| 1070 |
1
|
|
|
|
| 1073 |
},
|
| 1074 |
"k_proj": {
|
| 1075 |
"group_size": {
|
| 1076 |
+
"4": 128
|
| 1077 |
},
|
| 1078 |
"bits": [
|
| 1079 |
+
4
|
| 1080 |
],
|
| 1081 |
"bits_prop": [
|
| 1082 |
1
|
|
|
|
| 1145 |
}
|
| 1146 |
},
|
| 1147 |
"model.layers.13": {
|
| 1148 |
+
"accuracy": 0.9048597812652588,
|
| 1149 |
"total_bits": 458124288,
|
| 1150 |
"q_proj": {
|
| 1151 |
"group_size": {
|
|
|
|
| 1233 |
}
|
| 1234 |
},
|
| 1235 |
"model.layers.14": {
|
| 1236 |
+
"accuracy": 0.8939895629882812,
|
| 1237 |
"total_bits": 458124288,
|
| 1238 |
"q_proj": {
|
| 1239 |
"group_size": {
|
|
|
|
| 1321 |
}
|
| 1322 |
},
|
| 1323 |
"model.layers.15": {
|
| 1324 |
+
"accuracy": 0.88232421875,
|
| 1325 |
"total_bits": 458124288,
|
| 1326 |
"q_proj": {
|
| 1327 |
"group_size": {
|
|
|
|
| 1409 |
}
|
| 1410 |
},
|
| 1411 |
"model.layers.16": {
|
| 1412 |
+
"accuracy": 0.8802495002746582,
|
| 1413 |
"total_bits": 458124288,
|
| 1414 |
"q_proj": {
|
| 1415 |
"group_size": {
|
|
|
|
| 1497 |
}
|
| 1498 |
},
|
| 1499 |
"model.layers.17": {
|
| 1500 |
+
"accuracy": 0.8810954093933105,
|
| 1501 |
+
"total_bits": 466380288,
|
| 1502 |
"q_proj": {
|
| 1503 |
"group_size": {
|
| 1504 |
"2": 64
|
|
|
|
| 1513 |
},
|
| 1514 |
"k_proj": {
|
| 1515 |
"group_size": {
|
| 1516 |
+
"4": 128
|
| 1517 |
},
|
| 1518 |
"bits": [
|
| 1519 |
+
4
|
| 1520 |
],
|
| 1521 |
"bits_prop": [
|
| 1522 |
1
|
|
|
|
| 1537 |
},
|
| 1538 |
"o_proj": {
|
| 1539 |
"group_size": {
|
| 1540 |
+
"2": 64
|
| 1541 |
},
|
| 1542 |
"bits": [
|
| 1543 |
+
2
|
| 1544 |
],
|
| 1545 |
"bits_prop": [
|
| 1546 |
1
|
|
|
|
| 1585 |
}
|
| 1586 |
},
|
| 1587 |
"model.layers.18": {
|
| 1588 |
+
"accuracy": 0.8681793212890625,
|
| 1589 |
+
"total_bits": 466380288,
|
| 1590 |
"q_proj": {
|
| 1591 |
"group_size": {
|
| 1592 |
"2": 64
|
|
|
|
| 1601 |
},
|
| 1602 |
"k_proj": {
|
| 1603 |
"group_size": {
|
| 1604 |
+
"4": 128
|
| 1605 |
},
|
| 1606 |
"bits": [
|
| 1607 |
+
4
|
| 1608 |
],
|
| 1609 |
"bits_prop": [
|
| 1610 |
1
|
|
|
|
| 1625 |
},
|
| 1626 |
"o_proj": {
|
| 1627 |
"group_size": {
|
| 1628 |
+
"2": 64
|
| 1629 |
},
|
| 1630 |
"bits": [
|
| 1631 |
+
2
|
| 1632 |
],
|
| 1633 |
"bits_prop": [
|
| 1634 |
1
|
|
|
|
| 1673 |
}
|
| 1674 |
},
|
| 1675 |
"model.layers.19": {
|
| 1676 |
+
"accuracy": 0.865323543548584,
|
| 1677 |
+
"total_bits": 458124288,
|
| 1678 |
"q_proj": {
|
| 1679 |
"group_size": {
|
| 1680 |
"2": 64
|
|
|
|
| 1713 |
},
|
| 1714 |
"o_proj": {
|
| 1715 |
"group_size": {
|
| 1716 |
+
"2": 64
|
| 1717 |
},
|
| 1718 |
"bits": [
|
| 1719 |
+
2
|
| 1720 |
],
|
| 1721 |
"bits_prop": [
|
| 1722 |
1
|
|
|
|
| 1761 |
}
|
| 1762 |
},
|
| 1763 |
"model.layers.20": {
|
| 1764 |
+
"accuracy": 0.8790159225463867,
|
| 1765 |
+
"total_bits": 466380288,
|
| 1766 |
"q_proj": {
|
| 1767 |
"group_size": {
|
| 1768 |
"2": 64
|
|
|
|
| 1777 |
},
|
| 1778 |
"k_proj": {
|
| 1779 |
"group_size": {
|
| 1780 |
+
"4": 128
|
| 1781 |
},
|
| 1782 |
"bits": [
|
| 1783 |
+
4
|
| 1784 |
],
|
| 1785 |
"bits_prop": [
|
| 1786 |
1
|
|
|
|
| 1849 |
}
|
| 1850 |
},
|
| 1851 |
"model.layers.21": {
|
| 1852 |
+
"accuracy": 0.8824334144592285,
|
| 1853 |
"total_bits": 458124288,
|
| 1854 |
"q_proj": {
|
| 1855 |
"group_size": {
|
|
|
|
| 1937 |
}
|
| 1938 |
},
|
| 1939 |
"model.layers.22": {
|
| 1940 |
+
"accuracy": 0.8895416259765625,
|
| 1941 |
"total_bits": 458124288,
|
| 1942 |
"q_proj": {
|
| 1943 |
"group_size": {
|
|
|
|
| 2025 |
}
|
| 2026 |
},
|
| 2027 |
"model.layers.23": {
|
| 2028 |
+
"accuracy": 0.889040470123291,
|
| 2029 |
+
"total_bits": 458124288,
|
| 2030 |
"q_proj": {
|
| 2031 |
"group_size": {
|
| 2032 |
"2": 64
|
|
|
|
| 2041 |
},
|
| 2042 |
"k_proj": {
|
| 2043 |
"group_size": {
|
| 2044 |
+
"2": 64
|
| 2045 |
},
|
| 2046 |
"bits": [
|
| 2047 |
+
2
|
| 2048 |
],
|
| 2049 |
"bits_prop": [
|
| 2050 |
1
|
|
|
|
| 2113 |
}
|
| 2114 |
},
|
| 2115 |
"model.layers.24": {
|
| 2116 |
+
"accuracy": 0.8891377449035645,
|
| 2117 |
"total_bits": 458124288,
|
| 2118 |
"q_proj": {
|
| 2119 |
"group_size": {
|
|
|
|
| 2201 |
}
|
| 2202 |
},
|
| 2203 |
"model.layers.25": {
|
| 2204 |
+
"accuracy": 0.888709545135498,
|
| 2205 |
"total_bits": 458124288,
|
| 2206 |
"q_proj": {
|
| 2207 |
"group_size": {
|
|
|
|
| 2289 |
}
|
| 2290 |
},
|
| 2291 |
"model.layers.26": {
|
| 2292 |
+
"accuracy": 0.8839888572692871,
|
| 2293 |
+
"total_bits": 458124288,
|
| 2294 |
"q_proj": {
|
| 2295 |
"group_size": {
|
| 2296 |
"2": 64
|
|
|
|
| 2305 |
},
|
| 2306 |
"k_proj": {
|
| 2307 |
"group_size": {
|
| 2308 |
+
"2": 64
|
| 2309 |
},
|
| 2310 |
"bits": [
|
| 2311 |
+
2
|
| 2312 |
],
|
| 2313 |
"bits_prop": [
|
| 2314 |
1
|
|
|
|
| 2377 |
}
|
| 2378 |
},
|
| 2379 |
"model.layers.27": {
|
| 2380 |
+
"accuracy": 0.8789157867431641,
|
| 2381 |
+
"total_bits": 458124288,
|
| 2382 |
"q_proj": {
|
| 2383 |
"group_size": {
|
| 2384 |
"2": 64
|
|
|
|
| 2393 |
},
|
| 2394 |
"k_proj": {
|
| 2395 |
"group_size": {
|
| 2396 |
+
"2": 64
|
| 2397 |
},
|
| 2398 |
"bits": [
|
| 2399 |
+
2
|
| 2400 |
],
|
| 2401 |
"bits_prop": [
|
| 2402 |
1
|
|
|
|
| 2465 |
}
|
| 2466 |
},
|
| 2467 |
"model.layers.28": {
|
| 2468 |
+
"accuracy": 0.8596110343933105,
|
| 2469 |
"total_bits": 458124288,
|
| 2470 |
"q_proj": {
|
| 2471 |
"group_size": {
|
|
|
|
| 2553 |
}
|
| 2554 |
},
|
| 2555 |
"model.layers.29": {
|
| 2556 |
+
"accuracy": 0.8909573554992676,
|
| 2557 |
+
"total_bits": 573724416,
|
| 2558 |
"q_proj": {
|
| 2559 |
"group_size": {
|
| 2560 |
"2": 64
|
|
|
|
| 2629 |
},
|
| 2630 |
"down_proj": {
|
| 2631 |
"group_size": {
|
| 2632 |
+
"4": 128
|
| 2633 |
},
|
| 2634 |
"bits": [
|
| 2635 |
+
4
|
| 2636 |
],
|
| 2637 |
"bits_prop": [
|
| 2638 |
1
|
|
|
|
| 2641 |
}
|
| 2642 |
},
|
| 2643 |
"model.layers.30": {
|
| 2644 |
+
"accuracy": 0.8890519142150879,
|
| 2645 |
+
"total_bits": 606753024,
|
| 2646 |
"q_proj": {
|
| 2647 |
"group_size": {
|
| 2648 |
"2": 64
|
|
|
|
| 2681 |
},
|
| 2682 |
"o_proj": {
|
| 2683 |
"group_size": {
|
| 2684 |
+
"4": 128
|
| 2685 |
},
|
| 2686 |
"bits": [
|
| 2687 |
+
4
|
| 2688 |
],
|
| 2689 |
"bits_prop": [
|
| 2690 |
1
|
|
|
|
| 2729 |
}
|
| 2730 |
},
|
| 2731 |
"model.layers.31": {
|
| 2732 |
+
"accuracy": 0.9532780647277832,
|
| 2733 |
+
"total_bits": 846216960,
|
| 2734 |
"q_proj": {
|
| 2735 |
"group_size": {
|
| 2736 |
"2": 64
|
|
|
|
| 2745 |
},
|
| 2746 |
"k_proj": {
|
| 2747 |
"group_size": {
|
| 2748 |
+
"4": 128
|
| 2749 |
},
|
| 2750 |
"bits": [
|
| 2751 |
+
4
|
| 2752 |
],
|
| 2753 |
"bits_prop": [
|
| 2754 |
1
|
|
|
|
| 2769 |
},
|
| 2770 |
"o_proj": {
|
| 2771 |
"group_size": {
|
| 2772 |
+
"4": 128
|
| 2773 |
},
|
| 2774 |
"bits": [
|
| 2775 |
+
4
|
| 2776 |
],
|
| 2777 |
"bits_prop": [
|
| 2778 |
1
|
|
|
|
| 2781 |
},
|
| 2782 |
"up_proj": {
|
| 2783 |
"group_size": {
|
| 2784 |
+
"4": 128
|
| 2785 |
},
|
| 2786 |
"bits": [
|
| 2787 |
+
4
|
| 2788 |
],
|
| 2789 |
"bits_prop": [
|
| 2790 |
1
|
|
|
|
| 2793 |
},
|
| 2794 |
"gate_proj": {
|
| 2795 |
"group_size": {
|
| 2796 |
+
"4": 128
|
| 2797 |
},
|
| 2798 |
"bits": [
|
| 2799 |
+
4
|
| 2800 |
],
|
| 2801 |
"bits_prop": [
|
| 2802 |
1
|