Add files using upload-large-folder tool
Browse files- quant_log.csv +112 -112
quant_log.csv
CHANGED
|
@@ -1,113 +1,113 @@
|
|
| 1 |
layer,module,loss,samples,damp,time
|
| 2 |
-
0,self_attn.k_proj,0.00286047,0.01000,1.
|
| 3 |
-
0,self_attn.v_proj,0.00008587,0.01000,0.
|
| 4 |
-
0,self_attn.q_proj,0.00604117,0.01000,0.
|
| 5 |
-
0,self_attn.o_proj,0.00000667,0.01000,0.
|
| 6 |
-
0,mlp.up_proj,0.00457087,0.01000,0.
|
| 7 |
-
0,mlp.gate_proj,0.00589667,0.01000,0.
|
| 8 |
-
0,mlp.down_proj,0.00007713,0.01000,3.
|
| 9 |
-
1,self_attn.k_proj,0.00531300,0.01000,0.
|
| 10 |
-
1,self_attn.v_proj,0.00031778,0.01000,0.
|
| 11 |
-
1,self_attn.q_proj,0.01118748,0.01000,0.
|
| 12 |
-
1,self_attn.o_proj,0.00001939,0.01000,0.
|
| 13 |
-
1,mlp.up_proj,0.00727253,0.01000,0.
|
| 14 |
-
1,mlp.gate_proj,0.00990856,0.01000,0.
|
| 15 |
-
1,mlp.down_proj,0.01561193,0.01000,3.
|
| 16 |
-
2,self_attn.k_proj,0.01063392,0.01000,0.
|
| 17 |
-
2,self_attn.v_proj,0.00081396,0.01000,0.
|
| 18 |
-
2,self_attn.q_proj,0.02113565,0.01000,0.
|
| 19 |
-
2,self_attn.o_proj,0.00002964,0.01000,0.
|
| 20 |
-
2,mlp.up_proj,0.00948823,0.01000,0.
|
| 21 |
-
2,mlp.gate_proj,0.01516710,0.01000,0.
|
| 22 |
-
2,mlp.down_proj,0.00012158,0.01000,3.
|
| 23 |
-
3,self_attn.k_proj,0.00612808,0.01000,0.
|
| 24 |
-
3,self_attn.v_proj,0.00091983,0.01000,0.
|
| 25 |
-
3,self_attn.q_proj,0.01371052,0.01000,0.
|
| 26 |
-
3,self_attn.o_proj,0.00006568,0.01000,0.
|
| 27 |
-
3,mlp.up_proj,0.01164620,0.01000,0.
|
| 28 |
-
3,mlp.gate_proj,0.02318169,0.01000,0.
|
| 29 |
-
3,mlp.down_proj,0.00017193,0.01000,3.
|
| 30 |
-
4,self_attn.k_proj,0.00695425,0.01000,0.
|
| 31 |
-
4,self_attn.v_proj,0.00087837,0.01000,0.
|
| 32 |
-
4,self_attn.q_proj,0.01412656,0.01000,0.
|
| 33 |
-
4,self_attn.o_proj,0.00007624,0.01000,0.
|
| 34 |
-
4,mlp.up_proj,0.01137217,0.01000,0.
|
| 35 |
-
4,mlp.gate_proj,0.02435286,0.01000,0.
|
| 36 |
-
4,mlp.down_proj,0.00018256,0.01000,3.
|
| 37 |
-
5,self_attn.k_proj,0.00952968,0.01000,0.
|
| 38 |
-
5,self_attn.v_proj,0.00077736,0.01000,0.
|
| 39 |
-
5,self_attn.q_proj,0.01787762,0.01000,0.
|
| 40 |
-
5,self_attn.o_proj,0.00006262,0.01000,0.
|
| 41 |
-
5,mlp.up_proj,0.01121267,0.01000,0.
|
| 42 |
-
5,mlp.gate_proj,0.02058079,0.01000,0.
|
| 43 |
-
5,mlp.down_proj,0.00019598,0.01000,3.
|
| 44 |
-
6,self_attn.k_proj,0.00714239,0.01000,0.
|
| 45 |
-
6,self_attn.v_proj,0.00087018,0.01000,0.
|
| 46 |
-
6,self_attn.q_proj,0.01150759,0.01000,0.
|
| 47 |
-
6,self_attn.o_proj,0.00010737,0.01000,0.
|
| 48 |
-
6,mlp.up_proj,0.01072427,0.01000,0.
|
| 49 |
-
6,mlp.gate_proj,0.01942263,0.01000,0.
|
| 50 |
-
6,mlp.down_proj,0.00018506,0.01000,3.
|
| 51 |
-
7,self_attn.k_proj,0.00642806,0.01000,0.
|
| 52 |
-
7,self_attn.v_proj,0.00086181,0.01000,0.
|
| 53 |
-
7,self_attn.q_proj,0.01206517,0.01000,0.
|
| 54 |
-
7,self_attn.o_proj,0.00006596,0.01000,0.
|
| 55 |
-
7,mlp.up_proj,0.01061943,0.01000,0.
|
| 56 |
-
7,mlp.gate_proj,0.01755598,0.01000,0.
|
| 57 |
-
7,mlp.down_proj,0.00019826,0.01000,3.
|
| 58 |
-
8,self_attn.k_proj,0.00819136,0.01000,0.
|
| 59 |
-
8,self_attn.v_proj,0.00080792,0.01000,0.
|
| 60 |
-
8,self_attn.q_proj,0.01346879,0.01000,0.
|
| 61 |
-
8,self_attn.o_proj,0.00010528,0.01000,0.
|
| 62 |
-
8,mlp.up_proj,0.01235651,0.01000,0.
|
| 63 |
-
8,mlp.gate_proj,0.02037454,0.01000,0.
|
| 64 |
-
8,mlp.down_proj,0.00030105,0.01000,3.
|
| 65 |
-
9,self_attn.k_proj,0.00627975,0.01000,0.
|
| 66 |
-
9,self_attn.v_proj,0.00099315,0.01000,0.
|
| 67 |
-
9,self_attn.q_proj,0.01541151,0.01000,0.
|
| 68 |
-
9,self_attn.o_proj,0.00013672,0.01000,0.
|
| 69 |
-
9,mlp.up_proj,0.01343576,0.01000,
|
| 70 |
-
9,mlp.gate_proj,0.02248907,0.01000,0.
|
| 71 |
-
9,mlp.down_proj,0.00037896,0.01000,3.
|
| 72 |
-
10,self_attn.k_proj,0.00739591,0.01000,0.
|
| 73 |
-
10,self_attn.v_proj,0.00134160,0.01000,0.
|
| 74 |
-
10,self_attn.q_proj,0.01688930,0.01000,0.
|
| 75 |
-
10,self_attn.o_proj,0.00008455,0.01000,0.
|
| 76 |
-
10,mlp.up_proj,0.01514964,0.01000,0.
|
| 77 |
-
10,mlp.gate_proj,0.02416436,0.01000,0.
|
| 78 |
-
10,mlp.down_proj,0.00045021,0.01000,3.
|
| 79 |
-
11,self_attn.k_proj,0.00875334,0.01000,0.
|
| 80 |
-
11,self_attn.v_proj,0.00117116,0.01000,0.
|
| 81 |
-
11,self_attn.q_proj,0.01637944,0.01000,0.
|
| 82 |
-
11,self_attn.o_proj,0.00007648,0.01000,0.
|
| 83 |
-
11,mlp.up_proj,0.01644409,0.01000,0.
|
| 84 |
-
11,mlp.gate_proj,0.02566493,0.01000,0.
|
| 85 |
-
11,mlp.down_proj,0.00046611,0.01000,3.
|
| 86 |
-
12,self_attn.k_proj,0.00815642,0.01000,0.
|
| 87 |
-
12,self_attn.v_proj,0.00120938,0.01000,0.
|
| 88 |
-
12,self_attn.q_proj,0.01570916,0.01000,0.
|
| 89 |
-
12,self_attn.o_proj,0.00008020,0.01000,0.
|
| 90 |
-
12,mlp.up_proj,0.01633884,0.01000,0.
|
| 91 |
-
12,mlp.gate_proj,0.02448233,0.01000,0.
|
| 92 |
-
12,mlp.down_proj,0.00052249,0.01000,3.
|
| 93 |
-
13,self_attn.k_proj,0.00808590,0.01000,0.
|
| 94 |
-
13,self_attn.v_proj,0.00208899,0.01000,0.
|
| 95 |
-
13,self_attn.q_proj,0.01784795,0.01000,0.
|
| 96 |
-
13,self_attn.o_proj,0.00009968,0.01000,0.
|
| 97 |
-
13,mlp.up_proj,0.01809692,0.01000,0.
|
| 98 |
-
13,mlp.gate_proj,0.02517116,0.01000,0.
|
| 99 |
-
13,mlp.down_proj,0.00071865,0.01000,3.
|
| 100 |
-
14,self_attn.k_proj,0.00787623,0.01000,0.
|
| 101 |
-
14,self_attn.v_proj,0.00425377,0.01000,0.
|
| 102 |
-
14,self_attn.q_proj,0.01617077,0.01000,0.
|
| 103 |
-
14,self_attn.o_proj,0.00016092,0.01000,0.
|
| 104 |
-
14,mlp.up_proj,0.01908463,0.01000,0.
|
| 105 |
-
14,mlp.gate_proj,0.02866445,0.01000,
|
| 106 |
-
14,mlp.down_proj,0.00098983,0.01000,3.
|
| 107 |
-
15,self_attn.k_proj,0.00855702,0.01000,0.
|
| 108 |
-
15,self_attn.v_proj,0.00422558,0.01000,0.
|
| 109 |
-
15,self_attn.q_proj,0.01472759,0.01000,0.
|
| 110 |
-
15,self_attn.o_proj,0.00137697,0.01000,0.
|
| 111 |
-
15,mlp.up_proj,0.02298289,0.01000,0.
|
| 112 |
-
15,mlp.gate_proj,0.03041823,0.01000,0.
|
| 113 |
-
15,mlp.down_proj,0.00414630,0.01000,3.
|
|
|
|
| 1 |
layer,module,loss,samples,damp,time
|
| 2 |
+
0,self_attn.k_proj,0.00286047,0.01000,1.670
|
| 3 |
+
0,self_attn.v_proj,0.00008587,0.01000,0.935
|
| 4 |
+
0,self_attn.q_proj,0.00604117,0.01000,0.934
|
| 5 |
+
0,self_attn.o_proj,0.00000667,0.01000,0.928
|
| 6 |
+
0,mlp.up_proj,0.00457087,0.01000,0.938
|
| 7 |
+
0,mlp.gate_proj,0.00589667,0.01000,0.938
|
| 8 |
+
0,mlp.down_proj,0.00007713,0.01000,3.753
|
| 9 |
+
1,self_attn.k_proj,0.00531300,0.01000,0.932
|
| 10 |
+
1,self_attn.v_proj,0.00031778,0.01000,0.926
|
| 11 |
+
1,self_attn.q_proj,0.01118748,0.01000,0.925
|
| 12 |
+
1,self_attn.o_proj,0.00001939,0.01000,0.934
|
| 13 |
+
1,mlp.up_proj,0.00727253,0.01000,0.965
|
| 14 |
+
1,mlp.gate_proj,0.00990856,0.01000,0.979
|
| 15 |
+
1,mlp.down_proj,0.01561193,0.01000,3.898
|
| 16 |
+
2,self_attn.k_proj,0.01063392,0.01000,0.935
|
| 17 |
+
2,self_attn.v_proj,0.00081396,0.01000,0.937
|
| 18 |
+
2,self_attn.q_proj,0.02113565,0.01000,0.929
|
| 19 |
+
2,self_attn.o_proj,0.00002964,0.01000,0.937
|
| 20 |
+
2,mlp.up_proj,0.00948823,0.01000,0.955
|
| 21 |
+
2,mlp.gate_proj,0.01516710,0.01000,0.958
|
| 22 |
+
2,mlp.down_proj,0.00012158,0.01000,3.900
|
| 23 |
+
3,self_attn.k_proj,0.00612808,0.01000,0.937
|
| 24 |
+
3,self_attn.v_proj,0.00091983,0.01000,0.930
|
| 25 |
+
3,self_attn.q_proj,0.01371052,0.01000,0.953
|
| 26 |
+
3,self_attn.o_proj,0.00006568,0.01000,0.946
|
| 27 |
+
3,mlp.up_proj,0.01164620,0.01000,0.950
|
| 28 |
+
3,mlp.gate_proj,0.02318169,0.01000,0.957
|
| 29 |
+
3,mlp.down_proj,0.00017193,0.01000,3.859
|
| 30 |
+
4,self_attn.k_proj,0.00695425,0.01000,0.928
|
| 31 |
+
4,self_attn.v_proj,0.00087837,0.01000,0.919
|
| 32 |
+
4,self_attn.q_proj,0.01412656,0.01000,0.935
|
| 33 |
+
4,self_attn.o_proj,0.00007624,0.01000,0.927
|
| 34 |
+
4,mlp.up_proj,0.01137217,0.01000,0.953
|
| 35 |
+
4,mlp.gate_proj,0.02435286,0.01000,0.948
|
| 36 |
+
4,mlp.down_proj,0.00018256,0.01000,3.840
|
| 37 |
+
5,self_attn.k_proj,0.00952968,0.01000,0.954
|
| 38 |
+
5,self_attn.v_proj,0.00077736,0.01000,0.955
|
| 39 |
+
5,self_attn.q_proj,0.01787762,0.01000,0.948
|
| 40 |
+
5,self_attn.o_proj,0.00006262,0.01000,0.955
|
| 41 |
+
5,mlp.up_proj,0.01121267,0.01000,0.959
|
| 42 |
+
5,mlp.gate_proj,0.02058079,0.01000,0.958
|
| 43 |
+
5,mlp.down_proj,0.00019598,0.01000,3.870
|
| 44 |
+
6,self_attn.k_proj,0.00714239,0.01000,0.926
|
| 45 |
+
6,self_attn.v_proj,0.00087018,0.01000,0.918
|
| 46 |
+
6,self_attn.q_proj,0.01150759,0.01000,0.936
|
| 47 |
+
6,self_attn.o_proj,0.00010737,0.01000,0.934
|
| 48 |
+
6,mlp.up_proj,0.01072427,0.01000,0.959
|
| 49 |
+
6,mlp.gate_proj,0.01942263,0.01000,0.961
|
| 50 |
+
6,mlp.down_proj,0.00018506,0.01000,3.867
|
| 51 |
+
7,self_attn.k_proj,0.00642806,0.01000,0.931
|
| 52 |
+
7,self_attn.v_proj,0.00086181,0.01000,0.936
|
| 53 |
+
7,self_attn.q_proj,0.01206517,0.01000,0.948
|
| 54 |
+
7,self_attn.o_proj,0.00006596,0.01000,0.933
|
| 55 |
+
7,mlp.up_proj,0.01061943,0.01000,0.951
|
| 56 |
+
7,mlp.gate_proj,0.01755598,0.01000,0.962
|
| 57 |
+
7,mlp.down_proj,0.00019826,0.01000,3.905
|
| 58 |
+
8,self_attn.k_proj,0.00819136,0.01000,0.928
|
| 59 |
+
8,self_attn.v_proj,0.00080792,0.01000,0.931
|
| 60 |
+
8,self_attn.q_proj,0.01346879,0.01000,0.946
|
| 61 |
+
8,self_attn.o_proj,0.00010528,0.01000,0.961
|
| 62 |
+
8,mlp.up_proj,0.01235651,0.01000,0.946
|
| 63 |
+
8,mlp.gate_proj,0.02037454,0.01000,0.958
|
| 64 |
+
8,mlp.down_proj,0.00030105,0.01000,3.859
|
| 65 |
+
9,self_attn.k_proj,0.00627975,0.01000,0.941
|
| 66 |
+
9,self_attn.v_proj,0.00099315,0.01000,0.934
|
| 67 |
+
9,self_attn.q_proj,0.01541151,0.01000,0.951
|
| 68 |
+
9,self_attn.o_proj,0.00013672,0.01000,0.945
|
| 69 |
+
9,mlp.up_proj,0.01343576,0.01000,0.932
|
| 70 |
+
9,mlp.gate_proj,0.02248907,0.01000,0.959
|
| 71 |
+
9,mlp.down_proj,0.00037896,0.01000,3.871
|
| 72 |
+
10,self_attn.k_proj,0.00739591,0.01000,0.932
|
| 73 |
+
10,self_attn.v_proj,0.00134160,0.01000,0.929
|
| 74 |
+
10,self_attn.q_proj,0.01688930,0.01000,0.937
|
| 75 |
+
10,self_attn.o_proj,0.00008455,0.01000,0.942
|
| 76 |
+
10,mlp.up_proj,0.01514964,0.01000,0.976
|
| 77 |
+
10,mlp.gate_proj,0.02416436,0.01000,0.972
|
| 78 |
+
10,mlp.down_proj,0.00045021,0.01000,3.864
|
| 79 |
+
11,self_attn.k_proj,0.00875334,0.01000,0.923
|
| 80 |
+
11,self_attn.v_proj,0.00117116,0.01000,0.928
|
| 81 |
+
11,self_attn.q_proj,0.01637944,0.01000,0.928
|
| 82 |
+
11,self_attn.o_proj,0.00007648,0.01000,0.937
|
| 83 |
+
11,mlp.up_proj,0.01644409,0.01000,0.952
|
| 84 |
+
11,mlp.gate_proj,0.02566493,0.01000,0.943
|
| 85 |
+
11,mlp.down_proj,0.00046611,0.01000,3.861
|
| 86 |
+
12,self_attn.k_proj,0.00815642,0.01000,0.928
|
| 87 |
+
12,self_attn.v_proj,0.00120938,0.01000,0.953
|
| 88 |
+
12,self_attn.q_proj,0.01570916,0.01000,0.942
|
| 89 |
+
12,self_attn.o_proj,0.00008020,0.01000,0.930
|
| 90 |
+
12,mlp.up_proj,0.01633884,0.01000,0.937
|
| 91 |
+
12,mlp.gate_proj,0.02448233,0.01000,0.941
|
| 92 |
+
12,mlp.down_proj,0.00052249,0.01000,3.843
|
| 93 |
+
13,self_attn.k_proj,0.00808590,0.01000,0.942
|
| 94 |
+
13,self_attn.v_proj,0.00208899,0.01000,0.936
|
| 95 |
+
13,self_attn.q_proj,0.01784795,0.01000,0.940
|
| 96 |
+
13,self_attn.o_proj,0.00009968,0.01000,0.935
|
| 97 |
+
13,mlp.up_proj,0.01809692,0.01000,0.949
|
| 98 |
+
13,mlp.gate_proj,0.02517116,0.01000,0.955
|
| 99 |
+
13,mlp.down_proj,0.00071865,0.01000,3.883
|
| 100 |
+
14,self_attn.k_proj,0.00787623,0.01000,0.931
|
| 101 |
+
14,self_attn.v_proj,0.00425377,0.01000,0.939
|
| 102 |
+
14,self_attn.q_proj,0.01617077,0.01000,0.938
|
| 103 |
+
14,self_attn.o_proj,0.00016092,0.01000,0.941
|
| 104 |
+
14,mlp.up_proj,0.01908463,0.01000,0.943
|
| 105 |
+
14,mlp.gate_proj,0.02866445,0.01000,0.935
|
| 106 |
+
14,mlp.down_proj,0.00098983,0.01000,3.845
|
| 107 |
+
15,self_attn.k_proj,0.00855702,0.01000,0.940
|
| 108 |
+
15,self_attn.v_proj,0.00422558,0.01000,0.925
|
| 109 |
+
15,self_attn.q_proj,0.01472759,0.01000,0.967
|
| 110 |
+
15,self_attn.o_proj,0.00137697,0.01000,0.939
|
| 111 |
+
15,mlp.up_proj,0.02298289,0.01000,0.946
|
| 112 |
+
15,mlp.gate_proj,0.03041823,0.01000,0.952
|
| 113 |
+
15,mlp.down_proj,0.00414630,0.01000,3.863
|