SFconvertbot
commited on
Commit
•
2b9d42f
1
Parent(s):
28fc475
Adding `safetensors` variant of this model
Browse filesThis is an automated PR created with https://huggingface.co/spaces/safetensors/convert
This new file is equivalent to `pytorch_model.bin` but safe in the sense that
no arbitrary code can be put into it.
These files also happen to load much faster than their pytorch counterpart:
https://colab.research.google.com/github/huggingface/notebooks/blob/main/safetensors_doc/en/speed.ipynb
The widgets on your model page will run using this model even if this is not merged
making sure the file actually works.
If you find any issues: please report here: https://huggingface.co/spaces/safetensors/convert/discussions
Feel free to ignore this PR.
- model-00001-of-00007.safetensors +3 -0
- model-00002-of-00007.safetensors +3 -0
- model-00003-of-00007.safetensors +3 -0
- model-00004-of-00007.safetensors +3 -0
- model-00005-of-00007.safetensors +3 -0
- model-00006-of-00007.safetensors +3 -0
- model-00007-of-00007.safetensors +3 -0
- model.safetensors.index.json +297 -0
model-00001-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0a2c7ce460662c152dabe00b129e72321100ea064c2d6ff3249a76aa0c769b53
|
3 |
+
size 9766147384
|
model-00002-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:800d8ddba9bd35fdcc7353f44b18d09b3a494da08f0836a9e130faa05656ae7f
|
3 |
+
size 9865237992
|
model-00003-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4c0e334b2da9e08ede31ed1d554208e56e536d791764ba48e19ed57d8d64ec5b
|
3 |
+
size 9865238008
|
model-00004-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b58e716226a1dee2ebc94937f8992852c723fc7d147cf313c11ddb8ad79dbba4
|
3 |
+
size 9865238008
|
model-00005-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4577438656143d714ebdd2d382c4f70a79e3b9b10640f22da7a303ab0c825b2d
|
3 |
+
size 9865238008
|
model-00006-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f9b7d7015d6b7f4fe6a6dac83e835e70b272d18b23c8869606abb7b8dd7af747
|
3 |
+
size 9865238008
|
model-00007-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3c047a132e3a196744744b4ac75568e5b2f165da52e7f1a0df1acc5c0f749544
|
3 |
+
size 822098288
|
model.safetensors.index.json
ADDED
@@ -0,0 +1,297 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"metadata": {
|
3 |
+
"total_size": 59914401792
|
4 |
+
},
|
5 |
+
"weight_map": {
|
6 |
+
"transformer.blocks.0.attn.Wqkv.weight": "model-00001-of-00007.safetensors",
|
7 |
+
"transformer.blocks.0.attn.out_proj.weight": "model-00001-of-00007.safetensors",
|
8 |
+
"transformer.blocks.0.ffn.down_proj.weight": "model-00001-of-00007.safetensors",
|
9 |
+
"transformer.blocks.0.ffn.up_proj.weight": "model-00001-of-00007.safetensors",
|
10 |
+
"transformer.blocks.0.norm_1.weight": "model-00001-of-00007.safetensors",
|
11 |
+
"transformer.blocks.0.norm_2.weight": "model-00001-of-00007.safetensors",
|
12 |
+
"transformer.blocks.1.attn.Wqkv.weight": "model-00001-of-00007.safetensors",
|
13 |
+
"transformer.blocks.1.attn.out_proj.weight": "model-00001-of-00007.safetensors",
|
14 |
+
"transformer.blocks.1.ffn.down_proj.weight": "model-00001-of-00007.safetensors",
|
15 |
+
"transformer.blocks.1.ffn.up_proj.weight": "model-00001-of-00007.safetensors",
|
16 |
+
"transformer.blocks.1.norm_1.weight": "model-00001-of-00007.safetensors",
|
17 |
+
"transformer.blocks.1.norm_2.weight": "model-00001-of-00007.safetensors",
|
18 |
+
"transformer.blocks.10.attn.Wqkv.weight": "model-00002-of-00007.safetensors",
|
19 |
+
"transformer.blocks.10.attn.out_proj.weight": "model-00002-of-00007.safetensors",
|
20 |
+
"transformer.blocks.10.ffn.down_proj.weight": "model-00002-of-00007.safetensors",
|
21 |
+
"transformer.blocks.10.ffn.up_proj.weight": "model-00002-of-00007.safetensors",
|
22 |
+
"transformer.blocks.10.norm_1.weight": "model-00002-of-00007.safetensors",
|
23 |
+
"transformer.blocks.10.norm_2.weight": "model-00002-of-00007.safetensors",
|
24 |
+
"transformer.blocks.11.attn.Wqkv.weight": "model-00002-of-00007.safetensors",
|
25 |
+
"transformer.blocks.11.attn.out_proj.weight": "model-00002-of-00007.safetensors",
|
26 |
+
"transformer.blocks.11.ffn.down_proj.weight": "model-00002-of-00007.safetensors",
|
27 |
+
"transformer.blocks.11.ffn.up_proj.weight": "model-00002-of-00007.safetensors",
|
28 |
+
"transformer.blocks.11.norm_1.weight": "model-00002-of-00007.safetensors",
|
29 |
+
"transformer.blocks.11.norm_2.weight": "model-00002-of-00007.safetensors",
|
30 |
+
"transformer.blocks.12.attn.Wqkv.weight": "model-00002-of-00007.safetensors",
|
31 |
+
"transformer.blocks.12.attn.out_proj.weight": "model-00002-of-00007.safetensors",
|
32 |
+
"transformer.blocks.12.ffn.down_proj.weight": "model-00002-of-00007.safetensors",
|
33 |
+
"transformer.blocks.12.ffn.up_proj.weight": "model-00002-of-00007.safetensors",
|
34 |
+
"transformer.blocks.12.norm_1.weight": "model-00002-of-00007.safetensors",
|
35 |
+
"transformer.blocks.12.norm_2.weight": "model-00002-of-00007.safetensors",
|
36 |
+
"transformer.blocks.13.attn.Wqkv.weight": "model-00002-of-00007.safetensors",
|
37 |
+
"transformer.blocks.13.attn.out_proj.weight": "model-00002-of-00007.safetensors",
|
38 |
+
"transformer.blocks.13.ffn.down_proj.weight": "model-00002-of-00007.safetensors",
|
39 |
+
"transformer.blocks.13.ffn.up_proj.weight": "model-00002-of-00007.safetensors",
|
40 |
+
"transformer.blocks.13.norm_1.weight": "model-00002-of-00007.safetensors",
|
41 |
+
"transformer.blocks.13.norm_2.weight": "model-00002-of-00007.safetensors",
|
42 |
+
"transformer.blocks.14.attn.Wqkv.weight": "model-00002-of-00007.safetensors",
|
43 |
+
"transformer.blocks.14.attn.out_proj.weight": "model-00002-of-00007.safetensors",
|
44 |
+
"transformer.blocks.14.ffn.down_proj.weight": "model-00002-of-00007.safetensors",
|
45 |
+
"transformer.blocks.14.ffn.up_proj.weight": "model-00002-of-00007.safetensors",
|
46 |
+
"transformer.blocks.14.norm_1.weight": "model-00002-of-00007.safetensors",
|
47 |
+
"transformer.blocks.14.norm_2.weight": "model-00002-of-00007.safetensors",
|
48 |
+
"transformer.blocks.15.attn.Wqkv.weight": "model-00002-of-00007.safetensors",
|
49 |
+
"transformer.blocks.15.attn.out_proj.weight": "model-00002-of-00007.safetensors",
|
50 |
+
"transformer.blocks.15.ffn.down_proj.weight": "model-00003-of-00007.safetensors",
|
51 |
+
"transformer.blocks.15.ffn.up_proj.weight": "model-00003-of-00007.safetensors",
|
52 |
+
"transformer.blocks.15.norm_1.weight": "model-00002-of-00007.safetensors",
|
53 |
+
"transformer.blocks.15.norm_2.weight": "model-00002-of-00007.safetensors",
|
54 |
+
"transformer.blocks.16.attn.Wqkv.weight": "model-00003-of-00007.safetensors",
|
55 |
+
"transformer.blocks.16.attn.out_proj.weight": "model-00003-of-00007.safetensors",
|
56 |
+
"transformer.blocks.16.ffn.down_proj.weight": "model-00003-of-00007.safetensors",
|
57 |
+
"transformer.blocks.16.ffn.up_proj.weight": "model-00003-of-00007.safetensors",
|
58 |
+
"transformer.blocks.16.norm_1.weight": "model-00003-of-00007.safetensors",
|
59 |
+
"transformer.blocks.16.norm_2.weight": "model-00003-of-00007.safetensors",
|
60 |
+
"transformer.blocks.17.attn.Wqkv.weight": "model-00003-of-00007.safetensors",
|
61 |
+
"transformer.blocks.17.attn.out_proj.weight": "model-00003-of-00007.safetensors",
|
62 |
+
"transformer.blocks.17.ffn.down_proj.weight": "model-00003-of-00007.safetensors",
|
63 |
+
"transformer.blocks.17.ffn.up_proj.weight": "model-00003-of-00007.safetensors",
|
64 |
+
"transformer.blocks.17.norm_1.weight": "model-00003-of-00007.safetensors",
|
65 |
+
"transformer.blocks.17.norm_2.weight": "model-00003-of-00007.safetensors",
|
66 |
+
"transformer.blocks.18.attn.Wqkv.weight": "model-00003-of-00007.safetensors",
|
67 |
+
"transformer.blocks.18.attn.out_proj.weight": "model-00003-of-00007.safetensors",
|
68 |
+
"transformer.blocks.18.ffn.down_proj.weight": "model-00003-of-00007.safetensors",
|
69 |
+
"transformer.blocks.18.ffn.up_proj.weight": "model-00003-of-00007.safetensors",
|
70 |
+
"transformer.blocks.18.norm_1.weight": "model-00003-of-00007.safetensors",
|
71 |
+
"transformer.blocks.18.norm_2.weight": "model-00003-of-00007.safetensors",
|
72 |
+
"transformer.blocks.19.attn.Wqkv.weight": "model-00003-of-00007.safetensors",
|
73 |
+
"transformer.blocks.19.attn.out_proj.weight": "model-00003-of-00007.safetensors",
|
74 |
+
"transformer.blocks.19.ffn.down_proj.weight": "model-00003-of-00007.safetensors",
|
75 |
+
"transformer.blocks.19.ffn.up_proj.weight": "model-00003-of-00007.safetensors",
|
76 |
+
"transformer.blocks.19.norm_1.weight": "model-00003-of-00007.safetensors",
|
77 |
+
"transformer.blocks.19.norm_2.weight": "model-00003-of-00007.safetensors",
|
78 |
+
"transformer.blocks.2.attn.Wqkv.weight": "model-00001-of-00007.safetensors",
|
79 |
+
"transformer.blocks.2.attn.out_proj.weight": "model-00001-of-00007.safetensors",
|
80 |
+
"transformer.blocks.2.ffn.down_proj.weight": "model-00001-of-00007.safetensors",
|
81 |
+
"transformer.blocks.2.ffn.up_proj.weight": "model-00001-of-00007.safetensors",
|
82 |
+
"transformer.blocks.2.norm_1.weight": "model-00001-of-00007.safetensors",
|
83 |
+
"transformer.blocks.2.norm_2.weight": "model-00001-of-00007.safetensors",
|
84 |
+
"transformer.blocks.20.attn.Wqkv.weight": "model-00003-of-00007.safetensors",
|
85 |
+
"transformer.blocks.20.attn.out_proj.weight": "model-00003-of-00007.safetensors",
|
86 |
+
"transformer.blocks.20.ffn.down_proj.weight": "model-00003-of-00007.safetensors",
|
87 |
+
"transformer.blocks.20.ffn.up_proj.weight": "model-00003-of-00007.safetensors",
|
88 |
+
"transformer.blocks.20.norm_1.weight": "model-00003-of-00007.safetensors",
|
89 |
+
"transformer.blocks.20.norm_2.weight": "model-00003-of-00007.safetensors",
|
90 |
+
"transformer.blocks.21.attn.Wqkv.weight": "model-00003-of-00007.safetensors",
|
91 |
+
"transformer.blocks.21.attn.out_proj.weight": "model-00003-of-00007.safetensors",
|
92 |
+
"transformer.blocks.21.ffn.down_proj.weight": "model-00003-of-00007.safetensors",
|
93 |
+
"transformer.blocks.21.ffn.up_proj.weight": "model-00003-of-00007.safetensors",
|
94 |
+
"transformer.blocks.21.norm_1.weight": "model-00003-of-00007.safetensors",
|
95 |
+
"transformer.blocks.21.norm_2.weight": "model-00003-of-00007.safetensors",
|
96 |
+
"transformer.blocks.22.attn.Wqkv.weight": "model-00003-of-00007.safetensors",
|
97 |
+
"transformer.blocks.22.attn.out_proj.weight": "model-00003-of-00007.safetensors",
|
98 |
+
"transformer.blocks.22.ffn.down_proj.weight": "model-00003-of-00007.safetensors",
|
99 |
+
"transformer.blocks.22.ffn.up_proj.weight": "model-00003-of-00007.safetensors",
|
100 |
+
"transformer.blocks.22.norm_1.weight": "model-00003-of-00007.safetensors",
|
101 |
+
"transformer.blocks.22.norm_2.weight": "model-00003-of-00007.safetensors",
|
102 |
+
"transformer.blocks.23.attn.Wqkv.weight": "model-00003-of-00007.safetensors",
|
103 |
+
"transformer.blocks.23.attn.out_proj.weight": "model-00003-of-00007.safetensors",
|
104 |
+
"transformer.blocks.23.ffn.down_proj.weight": "model-00004-of-00007.safetensors",
|
105 |
+
"transformer.blocks.23.ffn.up_proj.weight": "model-00004-of-00007.safetensors",
|
106 |
+
"transformer.blocks.23.norm_1.weight": "model-00003-of-00007.safetensors",
|
107 |
+
"transformer.blocks.23.norm_2.weight": "model-00003-of-00007.safetensors",
|
108 |
+
"transformer.blocks.24.attn.Wqkv.weight": "model-00004-of-00007.safetensors",
|
109 |
+
"transformer.blocks.24.attn.out_proj.weight": "model-00004-of-00007.safetensors",
|
110 |
+
"transformer.blocks.24.ffn.down_proj.weight": "model-00004-of-00007.safetensors",
|
111 |
+
"transformer.blocks.24.ffn.up_proj.weight": "model-00004-of-00007.safetensors",
|
112 |
+
"transformer.blocks.24.norm_1.weight": "model-00004-of-00007.safetensors",
|
113 |
+
"transformer.blocks.24.norm_2.weight": "model-00004-of-00007.safetensors",
|
114 |
+
"transformer.blocks.25.attn.Wqkv.weight": "model-00004-of-00007.safetensors",
|
115 |
+
"transformer.blocks.25.attn.out_proj.weight": "model-00004-of-00007.safetensors",
|
116 |
+
"transformer.blocks.25.ffn.down_proj.weight": "model-00004-of-00007.safetensors",
|
117 |
+
"transformer.blocks.25.ffn.up_proj.weight": "model-00004-of-00007.safetensors",
|
118 |
+
"transformer.blocks.25.norm_1.weight": "model-00004-of-00007.safetensors",
|
119 |
+
"transformer.blocks.25.norm_2.weight": "model-00004-of-00007.safetensors",
|
120 |
+
"transformer.blocks.26.attn.Wqkv.weight": "model-00004-of-00007.safetensors",
|
121 |
+
"transformer.blocks.26.attn.out_proj.weight": "model-00004-of-00007.safetensors",
|
122 |
+
"transformer.blocks.26.ffn.down_proj.weight": "model-00004-of-00007.safetensors",
|
123 |
+
"transformer.blocks.26.ffn.up_proj.weight": "model-00004-of-00007.safetensors",
|
124 |
+
"transformer.blocks.26.norm_1.weight": "model-00004-of-00007.safetensors",
|
125 |
+
"transformer.blocks.26.norm_2.weight": "model-00004-of-00007.safetensors",
|
126 |
+
"transformer.blocks.27.attn.Wqkv.weight": "model-00004-of-00007.safetensors",
|
127 |
+
"transformer.blocks.27.attn.out_proj.weight": "model-00004-of-00007.safetensors",
|
128 |
+
"transformer.blocks.27.ffn.down_proj.weight": "model-00004-of-00007.safetensors",
|
129 |
+
"transformer.blocks.27.ffn.up_proj.weight": "model-00004-of-00007.safetensors",
|
130 |
+
"transformer.blocks.27.norm_1.weight": "model-00004-of-00007.safetensors",
|
131 |
+
"transformer.blocks.27.norm_2.weight": "model-00004-of-00007.safetensors",
|
132 |
+
"transformer.blocks.28.attn.Wqkv.weight": "model-00004-of-00007.safetensors",
|
133 |
+
"transformer.blocks.28.attn.out_proj.weight": "model-00004-of-00007.safetensors",
|
134 |
+
"transformer.blocks.28.ffn.down_proj.weight": "model-00004-of-00007.safetensors",
|
135 |
+
"transformer.blocks.28.ffn.up_proj.weight": "model-00004-of-00007.safetensors",
|
136 |
+
"transformer.blocks.28.norm_1.weight": "model-00004-of-00007.safetensors",
|
137 |
+
"transformer.blocks.28.norm_2.weight": "model-00004-of-00007.safetensors",
|
138 |
+
"transformer.blocks.29.attn.Wqkv.weight": "model-00004-of-00007.safetensors",
|
139 |
+
"transformer.blocks.29.attn.out_proj.weight": "model-00004-of-00007.safetensors",
|
140 |
+
"transformer.blocks.29.ffn.down_proj.weight": "model-00004-of-00007.safetensors",
|
141 |
+
"transformer.blocks.29.ffn.up_proj.weight": "model-00004-of-00007.safetensors",
|
142 |
+
"transformer.blocks.29.norm_1.weight": "model-00004-of-00007.safetensors",
|
143 |
+
"transformer.blocks.29.norm_2.weight": "model-00004-of-00007.safetensors",
|
144 |
+
"transformer.blocks.3.attn.Wqkv.weight": "model-00001-of-00007.safetensors",
|
145 |
+
"transformer.blocks.3.attn.out_proj.weight": "model-00001-of-00007.safetensors",
|
146 |
+
"transformer.blocks.3.ffn.down_proj.weight": "model-00001-of-00007.safetensors",
|
147 |
+
"transformer.blocks.3.ffn.up_proj.weight": "model-00001-of-00007.safetensors",
|
148 |
+
"transformer.blocks.3.norm_1.weight": "model-00001-of-00007.safetensors",
|
149 |
+
"transformer.blocks.3.norm_2.weight": "model-00001-of-00007.safetensors",
|
150 |
+
"transformer.blocks.30.attn.Wqkv.weight": "model-00004-of-00007.safetensors",
|
151 |
+
"transformer.blocks.30.attn.out_proj.weight": "model-00004-of-00007.safetensors",
|
152 |
+
"transformer.blocks.30.ffn.down_proj.weight": "model-00004-of-00007.safetensors",
|
153 |
+
"transformer.blocks.30.ffn.up_proj.weight": "model-00004-of-00007.safetensors",
|
154 |
+
"transformer.blocks.30.norm_1.weight": "model-00004-of-00007.safetensors",
|
155 |
+
"transformer.blocks.30.norm_2.weight": "model-00004-of-00007.safetensors",
|
156 |
+
"transformer.blocks.31.attn.Wqkv.weight": "model-00004-of-00007.safetensors",
|
157 |
+
"transformer.blocks.31.attn.out_proj.weight": "model-00004-of-00007.safetensors",
|
158 |
+
"transformer.blocks.31.ffn.down_proj.weight": "model-00005-of-00007.safetensors",
|
159 |
+
"transformer.blocks.31.ffn.up_proj.weight": "model-00005-of-00007.safetensors",
|
160 |
+
"transformer.blocks.31.norm_1.weight": "model-00004-of-00007.safetensors",
|
161 |
+
"transformer.blocks.31.norm_2.weight": "model-00004-of-00007.safetensors",
|
162 |
+
"transformer.blocks.32.attn.Wqkv.weight": "model-00005-of-00007.safetensors",
|
163 |
+
"transformer.blocks.32.attn.out_proj.weight": "model-00005-of-00007.safetensors",
|
164 |
+
"transformer.blocks.32.ffn.down_proj.weight": "model-00005-of-00007.safetensors",
|
165 |
+
"transformer.blocks.32.ffn.up_proj.weight": "model-00005-of-00007.safetensors",
|
166 |
+
"transformer.blocks.32.norm_1.weight": "model-00005-of-00007.safetensors",
|
167 |
+
"transformer.blocks.32.norm_2.weight": "model-00005-of-00007.safetensors",
|
168 |
+
"transformer.blocks.33.attn.Wqkv.weight": "model-00005-of-00007.safetensors",
|
169 |
+
"transformer.blocks.33.attn.out_proj.weight": "model-00005-of-00007.safetensors",
|
170 |
+
"transformer.blocks.33.ffn.down_proj.weight": "model-00005-of-00007.safetensors",
|
171 |
+
"transformer.blocks.33.ffn.up_proj.weight": "model-00005-of-00007.safetensors",
|
172 |
+
"transformer.blocks.33.norm_1.weight": "model-00005-of-00007.safetensors",
|
173 |
+
"transformer.blocks.33.norm_2.weight": "model-00005-of-00007.safetensors",
|
174 |
+
"transformer.blocks.34.attn.Wqkv.weight": "model-00005-of-00007.safetensors",
|
175 |
+
"transformer.blocks.34.attn.out_proj.weight": "model-00005-of-00007.safetensors",
|
176 |
+
"transformer.blocks.34.ffn.down_proj.weight": "model-00005-of-00007.safetensors",
|
177 |
+
"transformer.blocks.34.ffn.up_proj.weight": "model-00005-of-00007.safetensors",
|
178 |
+
"transformer.blocks.34.norm_1.weight": "model-00005-of-00007.safetensors",
|
179 |
+
"transformer.blocks.34.norm_2.weight": "model-00005-of-00007.safetensors",
|
180 |
+
"transformer.blocks.35.attn.Wqkv.weight": "model-00005-of-00007.safetensors",
|
181 |
+
"transformer.blocks.35.attn.out_proj.weight": "model-00005-of-00007.safetensors",
|
182 |
+
"transformer.blocks.35.ffn.down_proj.weight": "model-00005-of-00007.safetensors",
|
183 |
+
"transformer.blocks.35.ffn.up_proj.weight": "model-00005-of-00007.safetensors",
|
184 |
+
"transformer.blocks.35.norm_1.weight": "model-00005-of-00007.safetensors",
|
185 |
+
"transformer.blocks.35.norm_2.weight": "model-00005-of-00007.safetensors",
|
186 |
+
"transformer.blocks.36.attn.Wqkv.weight": "model-00005-of-00007.safetensors",
|
187 |
+
"transformer.blocks.36.attn.out_proj.weight": "model-00005-of-00007.safetensors",
|
188 |
+
"transformer.blocks.36.ffn.down_proj.weight": "model-00005-of-00007.safetensors",
|
189 |
+
"transformer.blocks.36.ffn.up_proj.weight": "model-00005-of-00007.safetensors",
|
190 |
+
"transformer.blocks.36.norm_1.weight": "model-00005-of-00007.safetensors",
|
191 |
+
"transformer.blocks.36.norm_2.weight": "model-00005-of-00007.safetensors",
|
192 |
+
"transformer.blocks.37.attn.Wqkv.weight": "model-00005-of-00007.safetensors",
|
193 |
+
"transformer.blocks.37.attn.out_proj.weight": "model-00005-of-00007.safetensors",
|
194 |
+
"transformer.blocks.37.ffn.down_proj.weight": "model-00005-of-00007.safetensors",
|
195 |
+
"transformer.blocks.37.ffn.up_proj.weight": "model-00005-of-00007.safetensors",
|
196 |
+
"transformer.blocks.37.norm_1.weight": "model-00005-of-00007.safetensors",
|
197 |
+
"transformer.blocks.37.norm_2.weight": "model-00005-of-00007.safetensors",
|
198 |
+
"transformer.blocks.38.attn.Wqkv.weight": "model-00005-of-00007.safetensors",
|
199 |
+
"transformer.blocks.38.attn.out_proj.weight": "model-00005-of-00007.safetensors",
|
200 |
+
"transformer.blocks.38.ffn.down_proj.weight": "model-00005-of-00007.safetensors",
|
201 |
+
"transformer.blocks.38.ffn.up_proj.weight": "model-00005-of-00007.safetensors",
|
202 |
+
"transformer.blocks.38.norm_1.weight": "model-00005-of-00007.safetensors",
|
203 |
+
"transformer.blocks.38.norm_2.weight": "model-00005-of-00007.safetensors",
|
204 |
+
"transformer.blocks.39.attn.Wqkv.weight": "model-00005-of-00007.safetensors",
|
205 |
+
"transformer.blocks.39.attn.out_proj.weight": "model-00005-of-00007.safetensors",
|
206 |
+
"transformer.blocks.39.ffn.down_proj.weight": "model-00006-of-00007.safetensors",
|
207 |
+
"transformer.blocks.39.ffn.up_proj.weight": "model-00006-of-00007.safetensors",
|
208 |
+
"transformer.blocks.39.norm_1.weight": "model-00005-of-00007.safetensors",
|
209 |
+
"transformer.blocks.39.norm_2.weight": "model-00005-of-00007.safetensors",
|
210 |
+
"transformer.blocks.4.attn.Wqkv.weight": "model-00001-of-00007.safetensors",
|
211 |
+
"transformer.blocks.4.attn.out_proj.weight": "model-00001-of-00007.safetensors",
|
212 |
+
"transformer.blocks.4.ffn.down_proj.weight": "model-00001-of-00007.safetensors",
|
213 |
+
"transformer.blocks.4.ffn.up_proj.weight": "model-00001-of-00007.safetensors",
|
214 |
+
"transformer.blocks.4.norm_1.weight": "model-00001-of-00007.safetensors",
|
215 |
+
"transformer.blocks.4.norm_2.weight": "model-00001-of-00007.safetensors",
|
216 |
+
"transformer.blocks.40.attn.Wqkv.weight": "model-00006-of-00007.safetensors",
|
217 |
+
"transformer.blocks.40.attn.out_proj.weight": "model-00006-of-00007.safetensors",
|
218 |
+
"transformer.blocks.40.ffn.down_proj.weight": "model-00006-of-00007.safetensors",
|
219 |
+
"transformer.blocks.40.ffn.up_proj.weight": "model-00006-of-00007.safetensors",
|
220 |
+
"transformer.blocks.40.norm_1.weight": "model-00006-of-00007.safetensors",
|
221 |
+
"transformer.blocks.40.norm_2.weight": "model-00006-of-00007.safetensors",
|
222 |
+
"transformer.blocks.41.attn.Wqkv.weight": "model-00006-of-00007.safetensors",
|
223 |
+
"transformer.blocks.41.attn.out_proj.weight": "model-00006-of-00007.safetensors",
|
224 |
+
"transformer.blocks.41.ffn.down_proj.weight": "model-00006-of-00007.safetensors",
|
225 |
+
"transformer.blocks.41.ffn.up_proj.weight": "model-00006-of-00007.safetensors",
|
226 |
+
"transformer.blocks.41.norm_1.weight": "model-00006-of-00007.safetensors",
|
227 |
+
"transformer.blocks.41.norm_2.weight": "model-00006-of-00007.safetensors",
|
228 |
+
"transformer.blocks.42.attn.Wqkv.weight": "model-00006-of-00007.safetensors",
|
229 |
+
"transformer.blocks.42.attn.out_proj.weight": "model-00006-of-00007.safetensors",
|
230 |
+
"transformer.blocks.42.ffn.down_proj.weight": "model-00006-of-00007.safetensors",
|
231 |
+
"transformer.blocks.42.ffn.up_proj.weight": "model-00006-of-00007.safetensors",
|
232 |
+
"transformer.blocks.42.norm_1.weight": "model-00006-of-00007.safetensors",
|
233 |
+
"transformer.blocks.42.norm_2.weight": "model-00006-of-00007.safetensors",
|
234 |
+
"transformer.blocks.43.attn.Wqkv.weight": "model-00006-of-00007.safetensors",
|
235 |
+
"transformer.blocks.43.attn.out_proj.weight": "model-00006-of-00007.safetensors",
|
236 |
+
"transformer.blocks.43.ffn.down_proj.weight": "model-00006-of-00007.safetensors",
|
237 |
+
"transformer.blocks.43.ffn.up_proj.weight": "model-00006-of-00007.safetensors",
|
238 |
+
"transformer.blocks.43.norm_1.weight": "model-00006-of-00007.safetensors",
|
239 |
+
"transformer.blocks.43.norm_2.weight": "model-00006-of-00007.safetensors",
|
240 |
+
"transformer.blocks.44.attn.Wqkv.weight": "model-00006-of-00007.safetensors",
|
241 |
+
"transformer.blocks.44.attn.out_proj.weight": "model-00006-of-00007.safetensors",
|
242 |
+
"transformer.blocks.44.ffn.down_proj.weight": "model-00006-of-00007.safetensors",
|
243 |
+
"transformer.blocks.44.ffn.up_proj.weight": "model-00006-of-00007.safetensors",
|
244 |
+
"transformer.blocks.44.norm_1.weight": "model-00006-of-00007.safetensors",
|
245 |
+
"transformer.blocks.44.norm_2.weight": "model-00006-of-00007.safetensors",
|
246 |
+
"transformer.blocks.45.attn.Wqkv.weight": "model-00006-of-00007.safetensors",
|
247 |
+
"transformer.blocks.45.attn.out_proj.weight": "model-00006-of-00007.safetensors",
|
248 |
+
"transformer.blocks.45.ffn.down_proj.weight": "model-00006-of-00007.safetensors",
|
249 |
+
"transformer.blocks.45.ffn.up_proj.weight": "model-00006-of-00007.safetensors",
|
250 |
+
"transformer.blocks.45.norm_1.weight": "model-00006-of-00007.safetensors",
|
251 |
+
"transformer.blocks.45.norm_2.weight": "model-00006-of-00007.safetensors",
|
252 |
+
"transformer.blocks.46.attn.Wqkv.weight": "model-00006-of-00007.safetensors",
|
253 |
+
"transformer.blocks.46.attn.out_proj.weight": "model-00006-of-00007.safetensors",
|
254 |
+
"transformer.blocks.46.ffn.down_proj.weight": "model-00006-of-00007.safetensors",
|
255 |
+
"transformer.blocks.46.ffn.up_proj.weight": "model-00006-of-00007.safetensors",
|
256 |
+
"transformer.blocks.46.norm_1.weight": "model-00006-of-00007.safetensors",
|
257 |
+
"transformer.blocks.46.norm_2.weight": "model-00006-of-00007.safetensors",
|
258 |
+
"transformer.blocks.47.attn.Wqkv.weight": "model-00006-of-00007.safetensors",
|
259 |
+
"transformer.blocks.47.attn.out_proj.weight": "model-00006-of-00007.safetensors",
|
260 |
+
"transformer.blocks.47.ffn.down_proj.weight": "model-00007-of-00007.safetensors",
|
261 |
+
"transformer.blocks.47.ffn.up_proj.weight": "model-00007-of-00007.safetensors",
|
262 |
+
"transformer.blocks.47.norm_1.weight": "model-00006-of-00007.safetensors",
|
263 |
+
"transformer.blocks.47.norm_2.weight": "model-00006-of-00007.safetensors",
|
264 |
+
"transformer.blocks.5.attn.Wqkv.weight": "model-00001-of-00007.safetensors",
|
265 |
+
"transformer.blocks.5.attn.out_proj.weight": "model-00001-of-00007.safetensors",
|
266 |
+
"transformer.blocks.5.ffn.down_proj.weight": "model-00001-of-00007.safetensors",
|
267 |
+
"transformer.blocks.5.ffn.up_proj.weight": "model-00001-of-00007.safetensors",
|
268 |
+
"transformer.blocks.5.norm_1.weight": "model-00001-of-00007.safetensors",
|
269 |
+
"transformer.blocks.5.norm_2.weight": "model-00001-of-00007.safetensors",
|
270 |
+
"transformer.blocks.6.attn.Wqkv.weight": "model-00001-of-00007.safetensors",
|
271 |
+
"transformer.blocks.6.attn.out_proj.weight": "model-00001-of-00007.safetensors",
|
272 |
+
"transformer.blocks.6.ffn.down_proj.weight": "model-00001-of-00007.safetensors",
|
273 |
+
"transformer.blocks.6.ffn.up_proj.weight": "model-00001-of-00007.safetensors",
|
274 |
+
"transformer.blocks.6.norm_1.weight": "model-00001-of-00007.safetensors",
|
275 |
+
"transformer.blocks.6.norm_2.weight": "model-00001-of-00007.safetensors",
|
276 |
+
"transformer.blocks.7.attn.Wqkv.weight": "model-00001-of-00007.safetensors",
|
277 |
+
"transformer.blocks.7.attn.out_proj.weight": "model-00001-of-00007.safetensors",
|
278 |
+
"transformer.blocks.7.ffn.down_proj.weight": "model-00002-of-00007.safetensors",
|
279 |
+
"transformer.blocks.7.ffn.up_proj.weight": "model-00002-of-00007.safetensors",
|
280 |
+
"transformer.blocks.7.norm_1.weight": "model-00001-of-00007.safetensors",
|
281 |
+
"transformer.blocks.7.norm_2.weight": "model-00001-of-00007.safetensors",
|
282 |
+
"transformer.blocks.8.attn.Wqkv.weight": "model-00002-of-00007.safetensors",
|
283 |
+
"transformer.blocks.8.attn.out_proj.weight": "model-00002-of-00007.safetensors",
|
284 |
+
"transformer.blocks.8.ffn.down_proj.weight": "model-00002-of-00007.safetensors",
|
285 |
+
"transformer.blocks.8.ffn.up_proj.weight": "model-00002-of-00007.safetensors",
|
286 |
+
"transformer.blocks.8.norm_1.weight": "model-00002-of-00007.safetensors",
|
287 |
+
"transformer.blocks.8.norm_2.weight": "model-00002-of-00007.safetensors",
|
288 |
+
"transformer.blocks.9.attn.Wqkv.weight": "model-00002-of-00007.safetensors",
|
289 |
+
"transformer.blocks.9.attn.out_proj.weight": "model-00002-of-00007.safetensors",
|
290 |
+
"transformer.blocks.9.ffn.down_proj.weight": "model-00002-of-00007.safetensors",
|
291 |
+
"transformer.blocks.9.ffn.up_proj.weight": "model-00002-of-00007.safetensors",
|
292 |
+
"transformer.blocks.9.norm_1.weight": "model-00002-of-00007.safetensors",
|
293 |
+
"transformer.blocks.9.norm_2.weight": "model-00002-of-00007.safetensors",
|
294 |
+
"transformer.norm_f.weight": "model-00007-of-00007.safetensors",
|
295 |
+
"transformer.wte.weight": "model-00001-of-00007.safetensors"
|
296 |
+
}
|
297 |
+
}
|