update

.gitattributes

@@ -1,35 +0,0 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:32a6de181787291a0ce29130d04e6a9ca0ad2d6d01f522d34d81e0e22b1a83c8
+oid sha256:946260478233ff53a84d891fe84462672d015bd6bde920c8e6f9f08f9fc42c4e
 size 3437638320

quant_strategy.json

@@ -1,14 +1,14 @@
 {
     "measurement": {
         "model.layers.0": {
-            "accuracy": 0.9328658580780029,
-            "total_bits": 648037632,
+            "accuracy": 0.8820657730102539,
+            "total_bits": 573724416,
             "q_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -17,10 +17,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -41,10 +41,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -89,14 +89,14 @@
             }
         },
         "model.layers.1": {
-            "accuracy": 0.9108891487121582,
-            "total_bits": 648037632,
+            "accuracy": 0.8954000473022461,
+            "total_bits": 573724416,
             "q_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -105,10 +105,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -129,10 +129,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -177,7 +177,7 @@
             }
         },
         "model.layers.2": {
-            "accuracy": 0.9520481824874878,
+            "accuracy": 0.9489546418190002,
             "total_bits": 458124288,
             "q_proj": {
                 "group_size": {
@@ -265,7 +265,7 @@
             }
         },
         "model.layers.3": {
-            "accuracy": 0.9584293961524963,
+            "accuracy": 0.9551604390144348,
             "total_bits": 466380288,
             "q_proj": {
                 "group_size": {
@@ -353,8 +353,8 @@
             }
         },
         "model.layers.4": {
-            "accuracy": 0.9519124627113342,
-            "total_bits": 466380288,
+            "accuracy": 0.948026180267334,
+            "total_bits": 458124288,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -369,10 +369,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -441,7 +441,7 @@
             }
         },
         "model.layers.5": {
-            "accuracy": 0.9403245449066162,
+            "accuracy": 0.938827395439148,
             "total_bits": 458124288,
             "q_proj": {
                 "group_size": {
@@ -529,8 +529,8 @@
             }
         },
         "model.layers.6": {
-            "accuracy": 0.9401131868362427,
-            "total_bits": 491152896,
+            "accuracy": 0.9337625503540039,
+            "total_bits": 458124288,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -569,10 +569,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -617,8 +617,8 @@
             }
         },
         "model.layers.7": {
-            "accuracy": 0.9464203119277954,
-            "total_bits": 606753024,
+            "accuracy": 0.9263930320739746,
+            "total_bits": 466380288,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -633,10 +633,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -657,10 +657,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -693,10 +693,10 @@
             },
             "down_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -705,14 +705,14 @@
             }
         },
         "model.layers.8": {
-            "accuracy": 0.9558004140853882,
-            "total_bits": 639781632,
+            "accuracy": 0.9275798797607422,
+            "total_bits": 466380288,
             "q_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -721,10 +721,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -745,10 +745,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -781,10 +781,10 @@
             },
             "down_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -793,14 +793,14 @@
             }
         },
         "model.layers.9": {
-            "accuracy": 0.9580546617507935,
-            "total_bits": 648037632,
+            "accuracy": 0.9196293354034424,
+            "total_bits": 458124288,
             "q_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -809,10 +809,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -833,10 +833,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -869,10 +869,10 @@
             },
             "down_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -881,14 +881,14 @@
             }
         },
         "model.layers.10": {
-            "accuracy": 0.9503617286682129,
-            "total_bits": 639781632,
+            "accuracy": 0.9175989627838135,
+            "total_bits": 466380288,
             "q_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -897,10 +897,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -921,10 +921,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -957,10 +957,10 @@
             },
             "down_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -969,8 +969,8 @@
             }
         },
         "model.layers.11": {
-            "accuracy": 0.9408208131790161,
-            "total_bits": 606753024,
+            "accuracy": 0.9152019023895264,
+            "total_bits": 466380288,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -985,10 +985,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -1009,10 +1009,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1045,10 +1045,10 @@
             },
             "down_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1057,14 +1057,14 @@
             }
         },
         "model.layers.12": {
-            "accuracy": 0.9455924034118652,
-            "total_bits": 639781632,
+            "accuracy": 0.9095911979675293,
+            "total_bits": 466380288,
             "q_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1073,10 +1073,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -1097,10 +1097,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1133,10 +1133,10 @@
             },
             "down_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1145,14 +1145,14 @@
             }
         },
         "model.layers.13": {
-            "accuracy": 0.9480124711990356,
-            "total_bits": 648037632,
+            "accuracy": 0.9048597812652588,
+            "total_bits": 458124288,
             "q_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1161,10 +1161,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1185,10 +1185,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1221,10 +1221,10 @@
             },
             "down_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1233,8 +1233,8 @@
             }
         },
         "model.layers.14": {
-            "accuracy": 0.9390950202941895,
-            "total_bits": 615009024,
+            "accuracy": 0.9164364337921143,
+            "total_bits": 573724416,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -1249,10 +1249,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1273,10 +1273,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1321,14 +1321,14 @@
             }
         },
         "model.layers.15": {
-            "accuracy": 0.9407241344451904,
-            "total_bits": 648037632,
+            "accuracy": 0.9038996696472168,
+            "total_bits": 573724416,
             "q_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1337,10 +1337,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1361,10 +1361,10 @@
             },
             "o_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1409,14 +1409,14 @@
             }
         },
         "model.layers.16": {
-            "accuracy": 0.9330849647521973,
-            "total_bits": 648037632,
+            "accuracy": 0.9254007339477539,
+            "total_bits": 615009024,
             "q_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1497,14 +1497,14 @@
             }
         },
         "model.layers.17": {
-            "accuracy": 0.9421899318695068,
-            "total_bits": 722356992,
+            "accuracy": 0.9273276329040527,
+            "total_bits": 648037632,
             "q_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -1513,10 +1513,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -1549,10 +1549,10 @@
             },
             "up_proj": {
                 "group_size": {
-                    "4": 128
+                    "2": 64
                 },
                 "bits": [
-                    4
+                    2
                 ],
                 "bits_prop": [
                     1
@@ -1585,7 +1585,7 @@
             }
         },
         "model.layers.18": {
-            "accuracy": 0.9388406276702881,
+            "accuracy": 0.9346868991851807,
             "total_bits": 722356992,
             "q_proj": {
                 "group_size": {
@@ -1673,14 +1673,14 @@
             }
         },
         "model.layers.19": {
-            "accuracy": 0.9384818077087402,
-            "total_bits": 722356992,
+            "accuracy": 0.9401049613952637,
+            "total_bits": 755385600,
             "q_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -1761,8 +1761,8 @@
             }
         },
         "model.layers.20": {
-            "accuracy": 0.9433858394622803,
-            "total_bits": 722356992,
+            "accuracy": 0.9692505598068237,
+            "total_bits": 846216960,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -1777,10 +1777,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -1825,10 +1825,10 @@
             },
             "gate_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -1849,8 +1849,8 @@
             }
         },
         "model.layers.21": {
-            "accuracy": 0.9452800750732422,
-            "total_bits": 722356992,
+            "accuracy": 0.9710770845413208,
+            "total_bits": 846216960,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -1865,10 +1865,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -1913,10 +1913,10 @@
             },
             "gate_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -1937,8 +1937,8 @@
             }
         },
         "model.layers.22": {
-            "accuracy": 0.9475953578948975,
-            "total_bits": 722356992,
+            "accuracy": 0.9737789630889893,
+            "total_bits": 846216960,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -1953,10 +1953,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2001,10 +2001,10 @@
             },
             "gate_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2025,8 +2025,8 @@
             }
         },
         "model.layers.23": {
-            "accuracy": 0.9454758167266846,
-            "total_bits": 722356992,
+            "accuracy": 0.9729764461517334,
+            "total_bits": 846216960,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -2041,10 +2041,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2089,10 +2089,10 @@
             },
             "gate_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2113,8 +2113,8 @@
             }
         },
         "model.layers.24": {
-            "accuracy": 0.9459505081176758,
-            "total_bits": 722356992,
+            "accuracy": 0.9737536907196045,
+            "total_bits": 846216960,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -2129,10 +2129,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2177,10 +2177,10 @@
             },
             "gate_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2201,8 +2201,8 @@
             }
         },
         "model.layers.25": {
-            "accuracy": 0.9440786838531494,
-            "total_bits": 722356992,
+            "accuracy": 0.9738945960998535,
+            "total_bits": 846216960,
             "q_proj": {
                 "group_size": {
                     "2": 64
@@ -2217,10 +2217,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2265,10 +2265,10 @@
             },
             "gate_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2289,14 +2289,14 @@
             }
         },
         "model.layers.26": {
-            "accuracy": 0.9433612823486328,
-            "total_bits": 722356992,
+            "accuracy": 0.9777768850326538,
+            "total_bits": 879245568,
             "q_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2305,10 +2305,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2353,10 +2353,10 @@
             },
             "gate_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2377,14 +2377,14 @@
             }
         },
         "model.layers.27": {
-            "accuracy": 0.9458332061767578,
-            "total_bits": 722356992,
+            "accuracy": 0.9767287969589233,
+            "total_bits": 879245568,
             "q_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2393,10 +2393,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2429,10 +2429,10 @@
             },
             "up_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2465,14 +2465,14 @@
             }
         },
         "model.layers.28": {
-            "accuracy": 0.9405984878540039,
-            "total_bits": 722356992,
+            "accuracy": 0.9730557203292847,
+            "total_bits": 879245568,
             "q_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2481,10 +2481,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2529,10 +2529,10 @@
             },
             "gate_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2553,14 +2553,14 @@
             }
         },
         "model.layers.29": {
-            "accuracy": 0.9377729892730713,
-            "total_bits": 722356992,
+            "accuracy": 0.9716944694519043,
+            "total_bits": 879245568,
             "q_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2569,10 +2569,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2605,10 +2605,10 @@
             },
             "up_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2641,14 +2641,14 @@
             }
         },
         "model.layers.30": {
-            "accuracy": 0.934866189956665,
-            "total_bits": 722356992,
+            "accuracy": 0.9692039489746094,
+            "total_bits": 879245568,
             "q_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2657,10 +2657,10 @@
             },
             "k_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2693,10 +2693,10 @@
             },
             "up_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1
@@ -2729,14 +2729,14 @@
             }
         },
         "model.layers.31": {
-            "accuracy": 0.9650318622589111,
-            "total_bits": 846216960,
+            "accuracy": 0.9576601982116699,
+            "total_bits": 879245568,
             "q_proj": {
                 "group_size": {
-                    "2": 64
+                    "4": 128
                 },
                 "bits": [
-                    2
+                    4
                 ],
                 "bits_prop": [
                     1