wangrongsheng
commited on
Commit
•
70bcfba
1
Parent(s):
3db8ae4
Update from root
Browse files- 1200/mp_rank_00_model_states.pt +3 -0
- 1500/mp_rank_00_model_states.pt +3 -0
- 1800/mp_rank_00_model_states.pt +3 -0
- 2100/mp_rank_00_model_states.pt +3 -0
- 2400/mp_rank_00_model_states.pt +3 -0
- 2700/mp_rank_00_model_states.pt +3 -0
- 300/mp_rank_00_model_states.pt +3 -0
- 3000/mp_rank_00_model_states.pt +3 -0
- 600/mp_rank_00_model_states.pt +3 -0
- 900/mp_rank_00_model_states.pt +3 -0
- latest +1 -0
- model_config.json +73 -0
1200/mp_rank_00_model_states.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1b2362a0a3673850c2d09507f00e3aedc5aff422bf28d30d9a18d9977e51101c
|
3 |
+
size 15622949761
|
1500/mp_rank_00_model_states.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8db099d997774154cbbb241c0bfb4b6beec66131d5dd4c592dd6122542e60abf
|
3 |
+
size 15622949761
|
1800/mp_rank_00_model_states.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0a8de2b8e603f00c6307561ccdecdd1534755d139bccb45d2c44d7faafd0a067
|
3 |
+
size 15622949761
|
2100/mp_rank_00_model_states.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5c117649d1fd572a8ed37db9b7430fd3574c5dd35db81ee16542b661d18c02c4
|
3 |
+
size 15622949761
|
2400/mp_rank_00_model_states.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:10120a9d2c68a4698c95728d8a5aa8ce7c9a7ce70c0c052547ee6c2ad5c07b1c
|
3 |
+
size 15622949761
|
2700/mp_rank_00_model_states.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1a5bc51ee3e77a330f9e6ea3453e022a13e015da2ade646cac588367206e0342
|
3 |
+
size 15622949761
|
300/mp_rank_00_model_states.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:431f5d57c09b5572466b9caa5c7f351169b6e7391d7b6219df4176f0881e36f8
|
3 |
+
size 15622949761
|
3000/mp_rank_00_model_states.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:32323f0213ebd6c0ba5238f65d482df19863c06172bc3969fdda599f2f91b0a4
|
3 |
+
size 15622949761
|
600/mp_rank_00_model_states.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5ac45fec305e7b769fa4125f898c5296c4da99f2a600b0e370726fdf3197c8e5
|
3 |
+
size 15622949761
|
900/mp_rank_00_model_states.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:209e70cad1af589beafbc886d1f78bad4e5afd2ccbbb0ac93cac624fb63f55a4
|
3 |
+
size 15622949761
|
latest
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
3000
|
model_config.json
ADDED
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_class": "FineTuneVisualGLMModel",
|
3 |
+
"tokenizer_type": "THUDM/chatglm-6b",
|
4 |
+
"num_layers": 28,
|
5 |
+
"hidden_size": 4096,
|
6 |
+
"num_attention_heads": 32,
|
7 |
+
"vocab_size": 130528,
|
8 |
+
"layernorm_order": "post",
|
9 |
+
"model_parallel_size": 1,
|
10 |
+
"max_sequence_length": 2048,
|
11 |
+
"pre_seq_len": 128,
|
12 |
+
"lora_rank": 10,
|
13 |
+
"use_ptuning": false,
|
14 |
+
"use_lora": true,
|
15 |
+
"image_length": 32,
|
16 |
+
"eva_args": {
|
17 |
+
"num_layers": 39,
|
18 |
+
"hidden_size": 1408,
|
19 |
+
"num_attention_heads": 16,
|
20 |
+
"vocab_size": 1,
|
21 |
+
"layernorm_order": "pre",
|
22 |
+
"model_parallel_size": 1,
|
23 |
+
"max_sequence_length": 257,
|
24 |
+
"inner_hidden_size": 6144,
|
25 |
+
"use_final_layernorm": false,
|
26 |
+
"layernorm_epsilon": 1e-06,
|
27 |
+
"image_size": [
|
28 |
+
224,
|
29 |
+
224
|
30 |
+
],
|
31 |
+
"pre_len": 1,
|
32 |
+
"post_len": 0,
|
33 |
+
"in_channels": 3,
|
34 |
+
"num_classes": 0,
|
35 |
+
"patch_size": 14
|
36 |
+
},
|
37 |
+
"qformer_args": {
|
38 |
+
"num_layers": 12,
|
39 |
+
"hidden_size": 768,
|
40 |
+
"num_attention_heads": 12,
|
41 |
+
"vocab_size": 32,
|
42 |
+
"layernorm_order": "post",
|
43 |
+
"model_parallel_size": 1,
|
44 |
+
"max_sequence_length": 0,
|
45 |
+
"is_decoder": [
|
46 |
+
true,
|
47 |
+
false,
|
48 |
+
true,
|
49 |
+
false,
|
50 |
+
true,
|
51 |
+
false,
|
52 |
+
true,
|
53 |
+
false,
|
54 |
+
true,
|
55 |
+
false,
|
56 |
+
true,
|
57 |
+
false
|
58 |
+
],
|
59 |
+
"cross_attn_hidden_size": 1408,
|
60 |
+
"layernorm_epsilon": 1e-12
|
61 |
+
},
|
62 |
+
"bos_token_id": 130004,
|
63 |
+
"mask_token_id": 130000,
|
64 |
+
"gmask_token_id": 130001,
|
65 |
+
"image_size": [
|
66 |
+
224,
|
67 |
+
224
|
68 |
+
],
|
69 |
+
"pre_len": 1,
|
70 |
+
"post_len": 0,
|
71 |
+
"in_channels": 3,
|
72 |
+
"patch_size": 14
|
73 |
+
}
|