Add files using upload-large-folder tool
Browse files- model.safetensors +1 -1
- quant_log.csv +196 -196
model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 2255767328
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c39cf80182cd44a4517c067951857427c1013ff5da1029b810f755bf0c19a530
|
3 |
size 2255767328
|
quant_log.csv
CHANGED
@@ -1,197 +1,197 @@
|
|
1 |
layer,module,loss,samples,damp,time
|
2 |
-
0,self_attn.k_proj,0.
|
3 |
-
0,self_attn.v_proj,0.
|
4 |
-
0,self_attn.q_proj,
|
5 |
-
0,self_attn.o_proj,0.
|
6 |
-
0,mlp.up_proj,0.
|
7 |
-
0,mlp.gate_proj,0.
|
8 |
-
0,mlp.down_proj,0.
|
9 |
-
1,self_attn.k_proj,0.
|
10 |
-
1,self_attn.v_proj,0.
|
11 |
-
1,self_attn.q_proj,
|
12 |
-
1,self_attn.o_proj,0.
|
13 |
-
1,mlp.up_proj,0.
|
14 |
-
1,mlp.gate_proj,
|
15 |
-
1,mlp.down_proj,0.
|
16 |
-
2,self_attn.k_proj,
|
17 |
-
2,self_attn.v_proj,0.
|
18 |
-
2,self_attn.q_proj,
|
19 |
-
2,self_attn.o_proj,0.
|
20 |
-
2,mlp.up_proj,
|
21 |
-
2,mlp.gate_proj,
|
22 |
-
2,mlp.down_proj,0.
|
23 |
-
3,self_attn.k_proj,
|
24 |
-
3,self_attn.v_proj,0.
|
25 |
-
3,self_attn.q_proj,
|
26 |
-
3,self_attn.o_proj,0.
|
27 |
-
3,mlp.up_proj,
|
28 |
-
3,mlp.gate_proj,
|
29 |
-
3,mlp.down_proj,0.
|
30 |
-
4,self_attn.k_proj,
|
31 |
-
4,self_attn.v_proj,0.
|
32 |
-
4,self_attn.q_proj,
|
33 |
-
4,self_attn.o_proj,0.
|
34 |
-
4,mlp.up_proj,
|
35 |
-
4,mlp.gate_proj,
|
36 |
-
4,mlp.down_proj,0.
|
37 |
-
5,self_attn.k_proj,
|
38 |
-
5,self_attn.v_proj,0.
|
39 |
-
5,self_attn.q_proj,
|
40 |
-
5,self_attn.o_proj,0.
|
41 |
-
5,mlp.up_proj,
|
42 |
-
5,mlp.gate_proj,
|
43 |
-
5,mlp.down_proj,0.
|
44 |
-
6,self_attn.k_proj,
|
45 |
-
6,self_attn.v_proj,0.
|
46 |
-
6,self_attn.q_proj,
|
47 |
-
6,self_attn.o_proj,0.
|
48 |
-
6,mlp.up_proj,
|
49 |
-
6,mlp.gate_proj,
|
50 |
-
6,mlp.down_proj,0.
|
51 |
-
7,self_attn.k_proj,
|
52 |
-
7,self_attn.v_proj,0.
|
53 |
-
7,self_attn.q_proj,
|
54 |
-
7,self_attn.o_proj,0.
|
55 |
-
7,mlp.up_proj,
|
56 |
-
7,mlp.gate_proj,
|
57 |
-
7,mlp.down_proj,0.
|
58 |
-
8,self_attn.k_proj,
|
59 |
-
8,self_attn.v_proj,0.
|
60 |
-
8,self_attn.q_proj,
|
61 |
-
8,self_attn.o_proj,0.
|
62 |
-
8,mlp.up_proj,
|
63 |
-
8,mlp.gate_proj,
|
64 |
-
8,mlp.down_proj,0.
|
65 |
-
9,self_attn.k_proj,
|
66 |
-
9,self_attn.v_proj,0.
|
67 |
-
9,self_attn.q_proj,
|
68 |
-
9,self_attn.o_proj,0.
|
69 |
-
9,mlp.up_proj,
|
70 |
-
9,mlp.gate_proj,
|
71 |
-
9,mlp.down_proj,0.
|
72 |
-
10,self_attn.k_proj,
|
73 |
-
10,self_attn.v_proj,0.
|
74 |
-
10,self_attn.q_proj,
|
75 |
-
10,self_attn.o_proj,0.
|
76 |
-
10,mlp.up_proj,
|
77 |
-
10,mlp.gate_proj,
|
78 |
-
10,mlp.down_proj,0.
|
79 |
-
11,self_attn.k_proj,
|
80 |
-
11,self_attn.v_proj,0.
|
81 |
-
11,self_attn.q_proj,
|
82 |
-
11,self_attn.o_proj,0.
|
83 |
-
11,mlp.up_proj,
|
84 |
-
11,mlp.gate_proj,
|
85 |
-
11,mlp.down_proj,0.
|
86 |
-
12,self_attn.k_proj,
|
87 |
-
12,self_attn.v_proj,0.
|
88 |
-
12,self_attn.q_proj,
|
89 |
-
12,self_attn.o_proj,0.
|
90 |
-
12,mlp.up_proj,
|
91 |
-
12,mlp.gate_proj,
|
92 |
-
12,mlp.down_proj,0.
|
93 |
-
13,self_attn.k_proj,
|
94 |
-
13,self_attn.v_proj,0.
|
95 |
-
13,self_attn.q_proj,
|
96 |
-
13,self_attn.o_proj,0.
|
97 |
-
13,mlp.up_proj,
|
98 |
-
13,mlp.gate_proj,
|
99 |
-
13,mlp.down_proj,0.
|
100 |
-
14,self_attn.k_proj,
|
101 |
-
14,self_attn.v_proj,0.
|
102 |
-
14,self_attn.q_proj,
|
103 |
-
14,self_attn.o_proj,0.
|
104 |
-
14,mlp.up_proj,
|
105 |
-
14,mlp.gate_proj,
|
106 |
-
14,mlp.down_proj,0.
|
107 |
-
15,self_attn.k_proj,
|
108 |
-
15,self_attn.v_proj,0.
|
109 |
-
15,self_attn.q_proj,
|
110 |
-
15,self_attn.o_proj,0.
|
111 |
-
15,mlp.up_proj,
|
112 |
-
15,mlp.gate_proj,
|
113 |
-
15,mlp.down_proj,0.
|
114 |
-
16,self_attn.k_proj,
|
115 |
-
16,self_attn.v_proj,0.
|
116 |
-
16,self_attn.q_proj,
|
117 |
-
16,self_attn.o_proj,0.
|
118 |
-
16,mlp.up_proj,
|
119 |
-
16,mlp.gate_proj,
|
120 |
-
16,mlp.down_proj,0.
|
121 |
-
17,self_attn.k_proj,
|
122 |
-
17,self_attn.v_proj,0.
|
123 |
-
17,self_attn.q_proj,
|
124 |
-
17,self_attn.o_proj,0.
|
125 |
-
17,mlp.up_proj,
|
126 |
-
17,mlp.gate_proj,
|
127 |
-
17,mlp.down_proj,0.
|
128 |
-
18,self_attn.k_proj,
|
129 |
-
18,self_attn.v_proj,0.
|
130 |
-
18,self_attn.q_proj,
|
131 |
-
18,self_attn.o_proj,0.
|
132 |
-
18,mlp.up_proj,
|
133 |
-
18,mlp.gate_proj,
|
134 |
-
18,mlp.down_proj,0.
|
135 |
-
19,self_attn.k_proj,
|
136 |
-
19,self_attn.v_proj,
|
137 |
-
19,self_attn.q_proj,
|
138 |
-
19,self_attn.o_proj,0.
|
139 |
-
19,mlp.up_proj,
|
140 |
-
19,mlp.gate_proj,
|
141 |
-
19,mlp.down_proj,0.
|
142 |
-
20,self_attn.k_proj,
|
143 |
-
20,self_attn.v_proj,
|
144 |
-
20,self_attn.q_proj,
|
145 |
-
20,self_attn.o_proj,0.
|
146 |
-
20,mlp.up_proj,
|
147 |
-
20,mlp.gate_proj,
|
148 |
-
20,mlp.down_proj,0.
|
149 |
-
21,self_attn.k_proj,
|
150 |
-
21,self_attn.v_proj,
|
151 |
-
21,self_attn.q_proj,
|
152 |
-
21,self_attn.o_proj,0.
|
153 |
-
21,mlp.up_proj,
|
154 |
-
21,mlp.gate_proj,
|
155 |
-
21,mlp.down_proj,0.
|
156 |
-
22,self_attn.k_proj,
|
157 |
-
22,self_attn.v_proj,
|
158 |
-
22,self_attn.q_proj,
|
159 |
-
22,self_attn.o_proj,0.
|
160 |
-
22,mlp.up_proj,
|
161 |
-
22,mlp.gate_proj,
|
162 |
-
22,mlp.down_proj,0.
|
163 |
-
23,self_attn.k_proj,
|
164 |
-
23,self_attn.v_proj,
|
165 |
-
23,self_attn.q_proj,
|
166 |
-
23,self_attn.o_proj,0.
|
167 |
-
23,mlp.up_proj,
|
168 |
-
23,mlp.gate_proj,
|
169 |
-
23,mlp.down_proj,0.
|
170 |
-
24,self_attn.k_proj,
|
171 |
-
24,self_attn.v_proj,
|
172 |
-
24,self_attn.q_proj,
|
173 |
-
24,self_attn.o_proj,0.
|
174 |
-
24,mlp.up_proj,
|
175 |
-
24,mlp.gate_proj,
|
176 |
-
24,mlp.down_proj,0.
|
177 |
-
25,self_attn.k_proj,
|
178 |
-
25,self_attn.v_proj,1.
|
179 |
-
25,self_attn.q_proj,
|
180 |
-
25,self_attn.o_proj,0.
|
181 |
-
25,mlp.up_proj,
|
182 |
-
25,mlp.gate_proj,
|
183 |
-
25,mlp.down_proj,0.
|
184 |
-
26,self_attn.k_proj,
|
185 |
-
26,self_attn.v_proj,
|
186 |
-
26,self_attn.q_proj,
|
187 |
-
26,self_attn.o_proj,0.
|
188 |
-
26,mlp.up_proj,
|
189 |
-
26,mlp.gate_proj,
|
190 |
-
26,mlp.down_proj,
|
191 |
-
27,self_attn.k_proj,
|
192 |
-
27,self_attn.v_proj,
|
193 |
-
27,self_attn.q_proj,
|
194 |
-
27,self_attn.o_proj,0.
|
195 |
-
27,mlp.up_proj,
|
196 |
-
27,mlp.gate_proj,
|
197 |
-
27,mlp.down_proj,
|
|
|
1 |
layer,module,loss,samples,damp,time
|
2 |
+
0,self_attn.k_proj,0.57348979,0.01000,1.101
|
3 |
+
0,self_attn.v_proj,0.02992631,0.01000,1.083
|
4 |
+
0,self_attn.q_proj,1.12482524,0.01000,1.095
|
5 |
+
0,self_attn.o_proj,0.00123981,0.01000,1.088
|
6 |
+
0,mlp.up_proj,0.71145487,0.01000,1.110
|
7 |
+
0,mlp.gate_proj,0.81329715,0.01000,1.129
|
8 |
+
0,mlp.down_proj,0.01093513,0.01000,3.119
|
9 |
+
1,self_attn.k_proj,0.74367076,0.01000,1.122
|
10 |
+
1,self_attn.v_proj,0.08594101,0.01000,1.106
|
11 |
+
1,self_attn.q_proj,1.27566791,0.01000,1.114
|
12 |
+
1,self_attn.o_proj,0.00476055,0.01000,1.105
|
13 |
+
1,mlp.up_proj,0.99940991,0.01000,1.137
|
14 |
+
1,mlp.gate_proj,1.13917911,0.01000,1.131
|
15 |
+
1,mlp.down_proj,0.12524097,0.01000,3.134
|
16 |
+
2,self_attn.k_proj,3.61536407,0.01000,1.099
|
17 |
+
2,self_attn.v_proj,0.36897749,0.01000,1.096
|
18 |
+
2,self_attn.q_proj,6.09369278,0.01000,1.125
|
19 |
+
2,self_attn.o_proj,0.00595129,0.01000,1.123
|
20 |
+
2,mlp.up_proj,1.57945013,0.01000,1.193
|
21 |
+
2,mlp.gate_proj,1.86483383,0.01000,1.140
|
22 |
+
2,mlp.down_proj,0.03576290,0.01000,3.141
|
23 |
+
3,self_attn.k_proj,2.34764647,0.01000,1.187
|
24 |
+
3,self_attn.v_proj,0.41367924,0.01000,1.129
|
25 |
+
3,self_attn.q_proj,4.28345490,0.01000,1.130
|
26 |
+
3,self_attn.o_proj,0.01283259,0.01000,1.112
|
27 |
+
3,mlp.up_proj,2.06658602,0.01000,1.146
|
28 |
+
3,mlp.gate_proj,2.75131416,0.01000,1.137
|
29 |
+
3,mlp.down_proj,0.05543867,0.01000,3.116
|
30 |
+
4,self_attn.k_proj,2.20512915,0.01000,1.152
|
31 |
+
4,self_attn.v_proj,0.42244333,0.01000,1.107
|
32 |
+
4,self_attn.q_proj,4.19914150,0.01000,1.129
|
33 |
+
4,self_attn.o_proj,0.01911609,0.01000,1.133
|
34 |
+
4,mlp.up_proj,2.42648292,0.01000,1.136
|
35 |
+
4,mlp.gate_proj,3.61736584,0.01000,1.187
|
36 |
+
4,mlp.down_proj,0.07793631,0.01000,3.147
|
37 |
+
5,self_attn.k_proj,3.54735780,0.01000,1.106
|
38 |
+
5,self_attn.v_proj,0.42594641,0.01000,1.106
|
39 |
+
5,self_attn.q_proj,5.93417883,0.01000,1.122
|
40 |
+
5,self_attn.o_proj,0.02330331,0.01000,1.122
|
41 |
+
5,mlp.up_proj,2.81386232,0.01000,1.139
|
42 |
+
5,mlp.gate_proj,3.96837521,0.01000,1.151
|
43 |
+
5,mlp.down_proj,0.10042880,0.01000,3.100
|
44 |
+
6,self_attn.k_proj,2.70604372,0.01000,1.122
|
45 |
+
6,self_attn.v_proj,0.41878402,0.01000,1.100
|
46 |
+
6,self_attn.q_proj,5.07468605,0.01000,1.164
|
47 |
+
6,self_attn.o_proj,0.03428555,0.01000,1.130
|
48 |
+
6,mlp.up_proj,2.92948699,0.01000,1.128
|
49 |
+
6,mlp.gate_proj,4.09053612,0.01000,1.139
|
50 |
+
6,mlp.down_proj,0.11020730,0.01000,3.142
|
51 |
+
7,self_attn.k_proj,2.56192708,0.01000,1.105
|
52 |
+
7,self_attn.v_proj,0.37280792,0.01000,1.113
|
53 |
+
7,self_attn.q_proj,4.33243322,0.01000,1.117
|
54 |
+
7,self_attn.o_proj,0.04311103,0.01000,1.124
|
55 |
+
7,mlp.up_proj,3.09119558,0.01000,1.154
|
56 |
+
7,mlp.gate_proj,3.96394897,0.01000,1.116
|
57 |
+
7,mlp.down_proj,0.12291946,0.01000,3.111
|
58 |
+
8,self_attn.k_proj,3.28836107,0.01000,1.114
|
59 |
+
8,self_attn.v_proj,0.47108710,0.01000,1.112
|
60 |
+
8,self_attn.q_proj,5.49438858,0.01000,1.126
|
61 |
+
8,self_attn.o_proj,0.06026545,0.01000,1.109
|
62 |
+
8,mlp.up_proj,3.16844177,0.01000,1.147
|
63 |
+
8,mlp.gate_proj,4.13681126,0.01000,1.133
|
64 |
+
8,mlp.down_proj,0.12639077,0.01000,3.141
|
65 |
+
9,self_attn.k_proj,3.11001682,0.01000,1.147
|
66 |
+
9,self_attn.v_proj,0.57155442,0.01000,1.130
|
67 |
+
9,self_attn.q_proj,5.25478554,0.01000,1.176
|
68 |
+
9,self_attn.o_proj,0.05964671,0.01000,1.164
|
69 |
+
9,mlp.up_proj,3.15676737,0.01000,1.162
|
70 |
+
9,mlp.gate_proj,3.97617245,0.01000,1.133
|
71 |
+
9,mlp.down_proj,0.12478939,0.01000,3.119
|
72 |
+
10,self_attn.k_proj,3.26966572,0.01000,1.131
|
73 |
+
10,self_attn.v_proj,0.45004451,0.01000,1.112
|
74 |
+
10,self_attn.q_proj,5.30321121,0.01000,1.142
|
75 |
+
10,self_attn.o_proj,0.05341929,0.01000,1.132
|
76 |
+
10,mlp.up_proj,3.35345793,0.01000,1.196
|
77 |
+
10,mlp.gate_proj,3.94216251,0.01000,1.174
|
78 |
+
10,mlp.down_proj,0.13452166,0.01000,3.080
|
79 |
+
11,self_attn.k_proj,2.52076197,0.01000,1.152
|
80 |
+
11,self_attn.v_proj,0.53631318,0.01000,1.116
|
81 |
+
11,self_attn.q_proj,4.49594688,0.01000,1.112
|
82 |
+
11,self_attn.o_proj,0.06898431,0.01000,1.122
|
83 |
+
11,mlp.up_proj,3.61847162,0.01000,1.165
|
84 |
+
11,mlp.gate_proj,4.10153961,0.01000,1.116
|
85 |
+
11,mlp.down_proj,0.15326369,0.01000,3.086
|
86 |
+
12,self_attn.k_proj,3.76040173,0.01000,1.128
|
87 |
+
12,self_attn.v_proj,0.57905239,0.01000,1.176
|
88 |
+
12,self_attn.q_proj,6.26392555,0.01000,1.107
|
89 |
+
12,self_attn.o_proj,0.08288961,0.01000,1.122
|
90 |
+
12,mlp.up_proj,3.79364228,0.01000,1.191
|
91 |
+
12,mlp.gate_proj,4.26222324,0.01000,1.158
|
92 |
+
12,mlp.down_proj,0.17245661,0.01000,3.106
|
93 |
+
13,self_attn.k_proj,4.17454338,0.01000,1.124
|
94 |
+
13,self_attn.v_proj,0.65613842,0.01000,1.129
|
95 |
+
13,self_attn.q_proj,6.48309612,0.01000,1.112
|
96 |
+
13,self_attn.o_proj,0.08975884,0.01000,1.105
|
97 |
+
13,mlp.up_proj,4.12620115,0.01000,1.149
|
98 |
+
13,mlp.gate_proj,4.92243528,0.01000,1.127
|
99 |
+
13,mlp.down_proj,0.21988256,0.01000,3.108
|
100 |
+
14,self_attn.k_proj,3.32866430,0.01000,1.136
|
101 |
+
14,self_attn.v_proj,0.75655323,0.01000,1.097
|
102 |
+
14,self_attn.q_proj,6.95757151,0.01000,1.134
|
103 |
+
14,self_attn.o_proj,0.11242675,0.01000,1.145
|
104 |
+
14,mlp.up_proj,4.40335560,0.01000,1.159
|
105 |
+
14,mlp.gate_proj,5.29281664,0.01000,1.150
|
106 |
+
14,mlp.down_proj,0.27417624,0.01000,3.150
|
107 |
+
15,self_attn.k_proj,3.84503889,0.01000,1.148
|
108 |
+
15,self_attn.v_proj,0.76655364,0.01000,1.103
|
109 |
+
15,self_attn.q_proj,7.31272936,0.01000,1.137
|
110 |
+
15,self_attn.o_proj,0.07379784,0.01000,1.101
|
111 |
+
15,mlp.up_proj,4.55024004,0.01000,1.200
|
112 |
+
15,mlp.gate_proj,5.88753128,0.01000,1.137
|
113 |
+
15,mlp.down_proj,0.29958892,0.01000,3.079
|
114 |
+
16,self_attn.k_proj,4.27809811,0.01000,1.117
|
115 |
+
16,self_attn.v_proj,0.83709764,0.01000,1.134
|
116 |
+
16,self_attn.q_proj,7.59177876,0.01000,1.126
|
117 |
+
16,self_attn.o_proj,0.05505002,0.01000,1.109
|
118 |
+
16,mlp.up_proj,4.66512632,0.01000,1.139
|
119 |
+
16,mlp.gate_proj,6.21557331,0.01000,1.145
|
120 |
+
16,mlp.down_proj,0.29199547,0.01000,3.142
|
121 |
+
17,self_attn.k_proj,3.99537516,0.01000,1.133
|
122 |
+
17,self_attn.v_proj,0.82619518,0.01000,1.118
|
123 |
+
17,self_attn.q_proj,7.35900211,0.01000,1.123
|
124 |
+
17,self_attn.o_proj,0.05116026,0.01000,1.113
|
125 |
+
17,mlp.up_proj,4.85350800,0.01000,1.139
|
126 |
+
17,mlp.gate_proj,6.57617331,0.01000,1.139
|
127 |
+
17,mlp.down_proj,0.31245530,0.01000,3.110
|
128 |
+
18,self_attn.k_proj,4.51366901,0.01000,1.136
|
129 |
+
18,self_attn.v_proj,0.99721968,0.01000,1.102
|
130 |
+
18,self_attn.q_proj,8.07002068,0.01000,1.129
|
131 |
+
18,self_attn.o_proj,0.05473011,0.01000,1.106
|
132 |
+
18,mlp.up_proj,5.22503281,0.01000,1.119
|
133 |
+
18,mlp.gate_proj,6.93419027,0.01000,1.149
|
134 |
+
18,mlp.down_proj,0.33353916,0.01000,3.098
|
135 |
+
19,self_attn.k_proj,4.22218132,0.01000,1.162
|
136 |
+
19,self_attn.v_proj,1.00496328,0.01000,1.149
|
137 |
+
19,self_attn.q_proj,7.31388569,0.01000,1.129
|
138 |
+
19,self_attn.o_proj,0.07915132,0.01000,1.183
|
139 |
+
19,mlp.up_proj,5.62336969,0.01000,1.137
|
140 |
+
19,mlp.gate_proj,7.33508205,0.01000,1.136
|
141 |
+
19,mlp.down_proj,0.41691202,0.01000,3.145
|
142 |
+
20,self_attn.k_proj,4.55664349,0.01000,1.113
|
143 |
+
20,self_attn.v_proj,1.19031966,0.01000,1.136
|
144 |
+
20,self_attn.q_proj,7.54975128,0.01000,1.128
|
145 |
+
20,self_attn.o_proj,0.07554041,0.01000,1.119
|
146 |
+
20,mlp.up_proj,5.75148487,0.01000,1.143
|
147 |
+
20,mlp.gate_proj,7.17578554,0.01000,1.166
|
148 |
+
20,mlp.down_proj,0.42174736,0.01000,3.105
|
149 |
+
21,self_attn.k_proj,4.42211723,0.01000,1.173
|
150 |
+
21,self_attn.v_proj,1.52764320,0.01000,1.114
|
151 |
+
21,self_attn.q_proj,7.39887810,0.01000,1.135
|
152 |
+
21,self_attn.o_proj,0.07596338,0.01000,1.121
|
153 |
+
21,mlp.up_proj,6.17190742,0.01000,1.145
|
154 |
+
21,mlp.gate_proj,7.76521111,0.01000,1.130
|
155 |
+
21,mlp.down_proj,0.44396168,0.01000,3.123
|
156 |
+
22,self_attn.k_proj,4.18366241,0.01000,1.106
|
157 |
+
22,self_attn.v_proj,1.55225110,0.01000,1.110
|
158 |
+
22,self_attn.q_proj,7.29905701,0.01000,1.152
|
159 |
+
22,self_attn.o_proj,0.06833672,0.01000,1.134
|
160 |
+
22,mlp.up_proj,6.65676212,0.01000,1.178
|
161 |
+
22,mlp.gate_proj,8.40142918,0.01000,1.131
|
162 |
+
22,mlp.down_proj,0.49725312,0.01000,3.171
|
163 |
+
23,self_attn.k_proj,4.44097424,0.01000,1.122
|
164 |
+
23,self_attn.v_proj,1.51754737,0.01000,1.107
|
165 |
+
23,self_attn.q_proj,7.22582531,0.01000,1.147
|
166 |
+
23,self_attn.o_proj,0.12661582,0.01000,1.114
|
167 |
+
23,mlp.up_proj,7.31189728,0.01000,1.131
|
168 |
+
23,mlp.gate_proj,9.66457844,0.01000,1.122
|
169 |
+
23,mlp.down_proj,0.59769487,0.01000,3.103
|
170 |
+
24,self_attn.k_proj,4.87307644,0.01000,1.105
|
171 |
+
24,self_attn.v_proj,2.15399027,0.01000,1.138
|
172 |
+
24,self_attn.q_proj,7.82933807,0.01000,1.128
|
173 |
+
24,self_attn.o_proj,0.14315630,0.01000,1.166
|
174 |
+
24,mlp.up_proj,8.24588013,0.01000,1.124
|
175 |
+
24,mlp.gate_proj,11.05454826,0.01000,1.127
|
176 |
+
24,mlp.down_proj,0.74648660,0.01000,3.122
|
177 |
+
25,self_attn.k_proj,4.00452709,0.01000,1.119
|
178 |
+
25,self_attn.v_proj,1.98942447,0.01000,1.128
|
179 |
+
25,self_attn.q_proj,7.78156376,0.01000,1.117
|
180 |
+
25,self_attn.o_proj,0.19304600,0.01000,1.117
|
181 |
+
25,mlp.up_proj,8.88384628,0.01000,1.165
|
182 |
+
25,mlp.gate_proj,11.88053036,0.01000,1.126
|
183 |
+
25,mlp.down_proj,0.99862719,0.01000,3.124
|
184 |
+
26,self_attn.k_proj,4.53559971,0.01000,1.138
|
185 |
+
26,self_attn.v_proj,2.68204308,0.01000,1.116
|
186 |
+
26,self_attn.q_proj,7.01802683,0.01000,1.171
|
187 |
+
26,self_attn.o_proj,0.28712690,0.01000,1.114
|
188 |
+
26,mlp.up_proj,9.31034470,0.01000,1.128
|
189 |
+
26,mlp.gate_proj,12.55875587,0.01000,1.154
|
190 |
+
26,mlp.down_proj,1.47142649,0.01000,3.096
|
191 |
+
27,self_attn.k_proj,3.15559483,0.01000,1.119
|
192 |
+
27,self_attn.v_proj,1.72210693,0.01000,1.119
|
193 |
+
27,self_attn.q_proj,5.58061886,0.01000,1.167
|
194 |
+
27,self_attn.o_proj,0.70267087,0.01000,1.129
|
195 |
+
27,mlp.up_proj,9.85498810,0.01000,1.158
|
196 |
+
27,mlp.gate_proj,11.81561852,0.01000,1.139
|
197 |
+
27,mlp.down_proj,3.57119989,0.01000,3.075
|