Upload 8 files
Browse files- config.json +3 -0
- generation_config.json +3 -0
- model.safetensors +3 -0
- quant_log.csv +97 -0
- quantize_config.json +3 -0
- special_tokens_map.json +3 -0
- tokenizer.json +3 -0
- tokenizer_config.json +3 -0
config.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5f976c53db8a0ead8feb5d942fa0d04dbead6c7392d47f65be4136192fdd9f5
|
| 3 |
+
size 1357
|
generation_config.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9b5e58d98752a3eedb39e914b06d32739cfead94176fe549a1458c4b23ba958e
|
| 3 |
+
size 157
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b942ee3279376377bc2f845ccad7bd5ddf24e4c3d14ea66cee70836ff52845d
|
| 3 |
+
size 672098142
|
quant_log.csv
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
layer,module,loss,samples,damp,time
|
| 2 |
+
0,self_attention.query_key_value,0.0000381449,0.05000,0.577
|
| 3 |
+
0,self_attention.dense,0.0000001857,0.05000,0.267
|
| 4 |
+
0,mlp.dense_h_to_4h,0.0000368609,0.05000,0.214
|
| 5 |
+
0,mlp.dense_4h_to_h,0.0000006347,0.05000,0.887
|
| 6 |
+
1,self_attention.query_key_value,0.0000500630,0.05000,0.227
|
| 7 |
+
1,self_attention.dense,0.0000001825,0.05000,0.224
|
| 8 |
+
1,mlp.dense_h_to_4h,0.0000487011,0.05000,0.235
|
| 9 |
+
1,mlp.dense_4h_to_h,0.0000008219,0.05000,1.018
|
| 10 |
+
2,self_attention.query_key_value,0.0000561378,0.05000,0.502
|
| 11 |
+
2,self_attention.dense,0.0000001609,0.05000,0.506
|
| 12 |
+
2,mlp.dense_h_to_4h,0.0000726370,0.05000,0.561
|
| 13 |
+
2,mlp.dense_4h_to_h,0.0000010389,0.05000,2.045
|
| 14 |
+
3,self_attention.query_key_value,0.0000580876,0.05000,0.572
|
| 15 |
+
3,self_attention.dense,0.0000002400,0.05000,0.565
|
| 16 |
+
3,mlp.dense_h_to_4h,0.0000801581,0.05000,0.576
|
| 17 |
+
3,mlp.dense_4h_to_h,0.0000013766,0.05000,2.265
|
| 18 |
+
4,self_attention.query_key_value,0.0000581960,0.05000,0.554
|
| 19 |
+
4,self_attention.dense,0.0000003266,0.05000,0.386
|
| 20 |
+
4,mlp.dense_h_to_4h,0.0000759618,0.05000,0.584
|
| 21 |
+
4,mlp.dense_4h_to_h,0.0000015705,0.05000,2.402
|
| 22 |
+
5,self_attention.query_key_value,0.0000669416,0.05000,0.624
|
| 23 |
+
5,self_attention.dense,0.0000004256,0.05000,0.633
|
| 24 |
+
5,mlp.dense_h_to_4h,0.0000628368,0.05000,0.610
|
| 25 |
+
5,mlp.dense_4h_to_h,0.0000012865,0.05000,2.523
|
| 26 |
+
6,self_attention.query_key_value,0.0000840233,0.05000,0.642
|
| 27 |
+
6,self_attention.dense,0.0000005805,0.05000,0.608
|
| 28 |
+
6,mlp.dense_h_to_4h,0.0000768948,0.05000,0.663
|
| 29 |
+
6,mlp.dense_4h_to_h,0.0000013620,0.05000,2.504
|
| 30 |
+
7,self_attention.query_key_value,0.0000712257,0.05000,0.658
|
| 31 |
+
7,self_attention.dense,0.0000005356,0.05000,0.657
|
| 32 |
+
7,mlp.dense_h_to_4h,0.0000416796,0.05000,0.678
|
| 33 |
+
7,mlp.dense_4h_to_h,0.0000009094,0.05000,2.609
|
| 34 |
+
8,self_attention.query_key_value,0.0001037150,0.05000,0.657
|
| 35 |
+
8,self_attention.dense,0.0000006358,0.05000,0.647
|
| 36 |
+
8,mlp.dense_h_to_4h,0.0000305985,0.05000,0.682
|
| 37 |
+
8,mlp.dense_4h_to_h,0.0000007244,0.05000,2.597
|
| 38 |
+
9,self_attention.query_key_value,0.0001000133,0.05000,0.672
|
| 39 |
+
9,self_attention.dense,0.0000005241,0.05000,0.685
|
| 40 |
+
9,mlp.dense_h_to_4h,0.0000235288,0.05000,0.719
|
| 41 |
+
9,mlp.dense_4h_to_h,0.0000006183,0.05000,2.701
|
| 42 |
+
10,self_attention.query_key_value,0.0001578843,0.05000,0.646
|
| 43 |
+
10,self_attention.dense,0.0000007396,0.05000,0.697
|
| 44 |
+
10,mlp.dense_h_to_4h,0.0000708152,0.05000,0.690
|
| 45 |
+
10,mlp.dense_4h_to_h,0.0000014019,0.05000,2.684
|
| 46 |
+
11,self_attention.query_key_value,0.0001074080,0.05000,0.675
|
| 47 |
+
11,self_attention.dense,0.0000006555,0.05000,0.695
|
| 48 |
+
11,mlp.dense_h_to_4h,0.0000729902,0.05000,0.668
|
| 49 |
+
11,mlp.dense_4h_to_h,0.0000014114,0.05000,2.677
|
| 50 |
+
12,self_attention.query_key_value,0.0001312283,0.05000,0.691
|
| 51 |
+
12,self_attention.dense,0.0000007142,0.05000,0.702
|
| 52 |
+
12,mlp.dense_h_to_4h,0.0000915113,0.05000,0.690
|
| 53 |
+
12,mlp.dense_4h_to_h,0.0000016252,0.05000,2.804
|
| 54 |
+
13,self_attention.query_key_value,0.0001273397,0.05000,0.693
|
| 55 |
+
13,self_attention.dense,0.0000007266,0.05000,0.673
|
| 56 |
+
13,mlp.dense_h_to_4h,0.0000935424,0.05000,0.680
|
| 57 |
+
13,mlp.dense_4h_to_h,0.0000015646,0.05000,2.833
|
| 58 |
+
14,self_attention.query_key_value,0.0001211264,0.05000,0.678
|
| 59 |
+
14,self_attention.dense,0.0000004718,0.05000,0.681
|
| 60 |
+
14,mlp.dense_h_to_4h,0.0000927910,0.05000,0.670
|
| 61 |
+
14,mlp.dense_4h_to_h,0.0000015917,0.05000,2.816
|
| 62 |
+
15,self_attention.query_key_value,0.0001268751,0.05000,0.705
|
| 63 |
+
15,self_attention.dense,0.0000005188,0.05000,0.688
|
| 64 |
+
15,mlp.dense_h_to_4h,0.0000943705,0.05000,0.730
|
| 65 |
+
15,mlp.dense_4h_to_h,0.0000016674,0.05000,2.914
|
| 66 |
+
16,self_attention.query_key_value,0.0001314014,0.05000,0.710
|
| 67 |
+
16,self_attention.dense,0.0000006596,0.05000,0.696
|
| 68 |
+
16,mlp.dense_h_to_4h,0.0000949185,0.05000,0.743
|
| 69 |
+
16,mlp.dense_4h_to_h,0.0000019402,0.05000,2.793
|
| 70 |
+
17,self_attention.query_key_value,0.0001188330,0.05000,0.688
|
| 71 |
+
17,self_attention.dense,0.0000006238,0.05000,0.684
|
| 72 |
+
17,mlp.dense_h_to_4h,0.0000951754,0.05000,0.707
|
| 73 |
+
17,mlp.dense_4h_to_h,0.0000023557,0.05000,2.790
|
| 74 |
+
18,self_attention.query_key_value,0.0001277351,0.05000,0.711
|
| 75 |
+
18,self_attention.dense,0.0000008486,0.05000,0.696
|
| 76 |
+
18,mlp.dense_h_to_4h,0.0000987540,0.05000,0.722
|
| 77 |
+
18,mlp.dense_4h_to_h,0.0000028473,0.05000,2.822
|
| 78 |
+
19,self_attention.query_key_value,0.0001311309,0.05000,0.717
|
| 79 |
+
19,self_attention.dense,0.0000011270,0.05000,0.718
|
| 80 |
+
19,mlp.dense_h_to_4h,0.0001040803,0.05000,0.720
|
| 81 |
+
19,mlp.dense_4h_to_h,0.0000035554,0.05000,2.850
|
| 82 |
+
20,self_attention.query_key_value,0.0001372522,0.05000,0.747
|
| 83 |
+
20,self_attention.dense,0.0000016063,0.05000,0.716
|
| 84 |
+
20,mlp.dense_h_to_4h,0.0001093306,0.05000,0.750
|
| 85 |
+
20,mlp.dense_4h_to_h,0.0000055216,0.05000,2.812
|
| 86 |
+
21,self_attention.query_key_value,0.0001562902,0.05000,0.747
|
| 87 |
+
21,self_attention.dense,0.0000033466,0.05000,0.724
|
| 88 |
+
21,mlp.dense_h_to_4h,0.0001095468,0.05000,0.742
|
| 89 |
+
21,mlp.dense_4h_to_h,0.0000084182,0.05000,2.795
|
| 90 |
+
22,self_attention.query_key_value,0.0001581889,0.05000,0.717
|
| 91 |
+
22,self_attention.dense,0.0000028682,0.05000,0.696
|
| 92 |
+
22,mlp.dense_h_to_4h,0.0001288263,0.05000,0.756
|
| 93 |
+
22,mlp.dense_4h_to_h,0.0000098412,0.05000,2.843
|
| 94 |
+
23,self_attention.query_key_value,0.0000716782,0.05000,0.750
|
| 95 |
+
23,self_attention.dense,0.0000005194,0.05000,0.707
|
| 96 |
+
23,mlp.dense_h_to_4h,0.0001178907,0.05000,0.723
|
| 97 |
+
23,mlp.dense_4h_to_h,0.0000037693,0.05000,2.867
|
quantize_config.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:046eac6257fd2cad445e86c9b152aa1d9d4939d75829561fd129ec1356ba26ee
|
| 3 |
+
size 542
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95206a0b849cbba9c3c0805024478799f3e734ed336ec57a7271247d85270867
|
| 3 |
+
size 438
|
tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d963066d6adae5034a1dc114c3ac444512de09928cf14ed4562ba94d9a440e66
|
| 3 |
+
size 21763085
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9a5b1ee7b43d7e8d3d2aedb08fa70a646594e101f8f937d51a0c76a7cf1cbb97
|
| 3 |
+
size 1058
|