Hugging Face's logo

keras
/
parseq

Text Generation
KerasHub
Model card Files
xet
Community
prasadsachin commited on
Commit
0b888cb
·
verified ·
1 Parent(s): 9965064

Upload folder using huggingface_hub

Browse files
Files changed (9) hide show
  1. README.md +22 -0
  2. assets/tokenizer/vocabulary.txt +94 -0
  3. config.json +62 -0
  4. image_converter.json +28 -0
  5. metadata.json +9 -0
  6. model.weights.h5 +3 -0
  7. preprocessor.json +67 -0
  8. task.json +142 -0
  9. tokenizer.json +18 -0
README.md ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: keras-hub
3
+ pipeline_tag: text-generation
4
+ ---
5
+ This is a [`PARSeq` model](https://keras.io/api/keras_hub/models/par_seq) uploaded using the KerasHub library and can be used with JAX, TensorFlow, and PyTorch backends.
6
+ This model is related to a `CausalLM` task.
7
+
8
+ Model config:
9
+ * **name:** par_seq_backbone
10
+ * **trainable:** True
11
+ * **dtype:** {'module': 'keras', 'class_name': 'DTypePolicy', 'config': {'name': 'float32'}, 'registered_name': None}
12
+ * **image_encoder:** {'module': 'keras_hub.src.models.vit.vit_backbone', 'class_name': 'ViTBackbone', 'config': {'name': 'encoder', 'trainable': True, 'dtype': {'module': 'keras', 'class_name': 'DTypePolicy', 'config': {'name': 'float32'}, 'registered_name': None}, 'image_shape': [32, 128, 3], 'patch_size': [4, 8], 'num_layers': 12, 'num_heads': 6, 'hidden_dim': 384, 'mlp_dim': 1536, 'dropout_rate': 0.0, 'attention_dropout': 0.0, 'layer_norm_epsilon': 1e-06, 'use_mha_bias': True, 'use_mlp_bias': True, 'use_class_token': False, 'use_patch_bias': True}, 'registered_name': 'keras_hub>ViTBackbone'}
13
+ * **vocabulary_size:** 97
14
+ * **max_label_length:** 25
15
+ * **decoder_hidden_dim:** 384
16
+ * **num_decoder_layers:** 1
17
+ * **num_decoder_heads:** 12
18
+ * **decoder_mlp_dim:** 1536
19
+ * **dropout_rate:** 0.1
20
+ * **attention_dropout:** 0.1
21
+
22
+ This model card has been generated automatically and should be completed by the model author. See [Model Cards documentation](https://huggingface.co/docs/hub/model-cards) for more information.
assets/tokenizer/vocabulary.txt ADDED
@@ -0,0 +1,94 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 0
2
+ 1
3
+ 2
4
+ 3
5
+ 4
6
+ 5
7
+ 6
8
+ 7
9
+ 8
10
+ 9
11
+ a
12
+ b
13
+ c
14
+ d
15
+ e
16
+ f
17
+ g
18
+ h
19
+ i
20
+ j
21
+ k
22
+ l
23
+ m
24
+ n
25
+ o
26
+ p
27
+ q
28
+ r
29
+ s
30
+ t
31
+ u
32
+ v
33
+ w
34
+ x
35
+ y
36
+ z
37
+ A
38
+ B
39
+ C
40
+ D
41
+ E
42
+ F
43
+ G
44
+ H
45
+ I
46
+ J
47
+ K
48
+ L
49
+ M
50
+ N
51
+ O
52
+ P
53
+ Q
54
+ R
55
+ S
56
+ T
57
+ U
58
+ V
59
+ W
60
+ X
61
+ Y
62
+ Z
63
+ !
64
+ "
65
+ #
66
+ $
67
+ %
68
+ &
69
+ '
70
+ (
71
+ )
72
+ *
73
+ +
74
+ ,
75
+ -
76
+ .
77
+ /
78
+ :
79
+ ;
80
+ <
81
+ =
82
+ >
83
+ ?
84
+ @
85
+ [
86
+ \
87
+ ]
88
+ ^
89
+ _
90
+ `
91
+ {
92
+ |
93
+ }
94
+ ~
config.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.parseq.parseq_backbone",
3
+ "class_name": "PARSeqBackbone",
4
+ "config": {
5
+ "name": "par_seq_backbone",
6
+ "trainable": true,
7
+ "dtype": {
8
+ "module": "keras",
9
+ "class_name": "DTypePolicy",
10
+ "config": {
11
+ "name": "float32"
12
+ },
13
+ "registered_name": null
14
+ },
15
+ "image_encoder": {
16
+ "module": "keras_hub.src.models.vit.vit_backbone",
17
+ "class_name": "ViTBackbone",
18
+ "config": {
19
+ "name": "encoder",
20
+ "trainable": true,
21
+ "dtype": {
22
+ "module": "keras",
23
+ "class_name": "DTypePolicy",
24
+ "config": {
25
+ "name": "float32"
26
+ },
27
+ "registered_name": null
28
+ },
29
+ "image_shape": [
30
+ 32,
31
+ 128,
32
+ 3
33
+ ],
34
+ "patch_size": [
35
+ 4,
36
+ 8
37
+ ],
38
+ "num_layers": 12,
39
+ "num_heads": 6,
40
+ "hidden_dim": 384,
41
+ "mlp_dim": 1536,
42
+ "dropout_rate": 0.0,
43
+ "attention_dropout": 0.0,
44
+ "layer_norm_epsilon": 1e-06,
45
+ "use_mha_bias": true,
46
+ "use_mlp_bias": true,
47
+ "use_class_token": false,
48
+ "use_patch_bias": true
49
+ },
50
+ "registered_name": "keras_hub>ViTBackbone"
51
+ },
52
+ "vocabulary_size": 97,
53
+ "max_label_length": 25,
54
+ "decoder_hidden_dim": 384,
55
+ "num_decoder_layers": 1,
56
+ "num_decoder_heads": 12,
57
+ "decoder_mlp_dim": 1536,
58
+ "dropout_rate": 0.1,
59
+ "attention_dropout": 0.1
60
+ },
61
+ "registered_name": "keras_hub>PARSeqBackbone"
62
+ }
image_converter.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.parseq.parseq_image_converter",
3
+ "class_name": "PARSeqImageConverter",
4
+ "config": {
5
+ "name": "par_seq_image_converter",
6
+ "trainable": true,
7
+ "dtype": {
8
+ "module": "keras",
9
+ "class_name": "DTypePolicy",
10
+ "config": {
11
+ "name": "float32"
12
+ },
13
+ "registered_name": null
14
+ },
15
+ "image_size": [
16
+ 32,
17
+ 128
18
+ ],
19
+ "scale": 0.00784313725490196,
20
+ "offset": -1.0,
21
+ "interpolation": "bicubic",
22
+ "antialias": false,
23
+ "crop_to_aspect_ratio": true,
24
+ "pad_to_aspect_ratio": false,
25
+ "bounding_box_format": "yxyx"
26
+ },
27
+ "registered_name": "keras_hub>PARSeqImageConverter"
28
+ }
metadata.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "keras_version": "3.10.0",
3
+ "keras_hub_version": "0.23.0.dev0",
4
+ "parameter_count": 23832671,
5
+ "date_saved": "2025-10-16@21:30:21",
6
+ "tasks": [
7
+ "CausalLM"
8
+ ]
9
+ }
model.weights.h5 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7442e2a2d9e3250f38cd44433f0e786334bc72eaaef686676d69dd14a0a899f7
3
+ size 95903800
preprocessor.json ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.parseq.parseq_causal_lm_preprocessor",
3
+ "class_name": "PARSeqCausalLMPreprocessor",
4
+ "config": {
5
+ "name": "par_seq_causal_lm_preprocessor",
6
+ "trainable": true,
7
+ "dtype": {
8
+ "module": "keras",
9
+ "class_name": "DTypePolicy",
10
+ "config": {
11
+ "name": "float32"
12
+ },
13
+ "registered_name": null
14
+ },
15
+ "tokenizer": {
16
+ "module": "keras_hub.src.models.parseq.parseq_tokenizer",
17
+ "class_name": "PARSeqTokenizer",
18
+ "config": {
19
+ "name": "par_seq_tokenizer",
20
+ "trainable": true,
21
+ "dtype": {
22
+ "module": "keras",
23
+ "class_name": "DTypePolicy",
24
+ "config": {
25
+ "name": "int32"
26
+ },
27
+ "registered_name": null
28
+ },
29
+ "config_file": "tokenizer.json"
30
+ },
31
+ "registered_name": "keras_hub>PARSeqTokenizer"
32
+ },
33
+ "image_converter": {
34
+ "module": "keras_hub.src.models.parseq.parseq_image_converter",
35
+ "class_name": "PARSeqImageConverter",
36
+ "config": {
37
+ "name": "par_seq_image_converter",
38
+ "trainable": true,
39
+ "dtype": {
40
+ "module": "keras",
41
+ "class_name": "DTypePolicy",
42
+ "config": {
43
+ "name": "float32"
44
+ },
45
+ "registered_name": null
46
+ },
47
+ "image_size": [
48
+ 32,
49
+ 128
50
+ ],
51
+ "scale": 0.00784313725490196,
52
+ "offset": -1.0,
53
+ "interpolation": "bicubic",
54
+ "antialias": false,
55
+ "crop_to_aspect_ratio": true,
56
+ "pad_to_aspect_ratio": false,
57
+ "bounding_box_format": "yxyx"
58
+ },
59
+ "registered_name": "keras_hub>PARSeqImageConverter"
60
+ },
61
+ "config_file": "preprocessor.json",
62
+ "sequence_length": 25,
63
+ "add_start_token": true,
64
+ "add_end_token": true
65
+ },
66
+ "registered_name": "keras_hub>PARSeqCausalLMPreprocessor"
67
+ }
task.json ADDED
@@ -0,0 +1,142 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.parseq.parseq_causal_lm",
3
+ "class_name": "PARSeqCausalLM",
4
+ "config": {
5
+ "backbone": {
6
+ "module": "keras_hub.src.models.parseq.parseq_backbone",
7
+ "class_name": "PARSeqBackbone",
8
+ "config": {
9
+ "name": "par_seq_backbone",
10
+ "trainable": true,
11
+ "dtype": {
12
+ "module": "keras",
13
+ "class_name": "DTypePolicy",
14
+ "config": {
15
+ "name": "float32"
16
+ },
17
+ "registered_name": null
18
+ },
19
+ "image_encoder": {
20
+ "module": "keras_hub.src.models.vit.vit_backbone",
21
+ "class_name": "ViTBackbone",
22
+ "config": {
23
+ "name": "encoder",
24
+ "trainable": true,
25
+ "dtype": {
26
+ "module": "keras",
27
+ "class_name": "DTypePolicy",
28
+ "config": {
29
+ "name": "float32"
30
+ },
31
+ "registered_name": null
32
+ },
33
+ "image_shape": [
34
+ 32,
35
+ 128,
36
+ 3
37
+ ],
38
+ "patch_size": [
39
+ 4,
40
+ 8
41
+ ],
42
+ "num_layers": 12,
43
+ "num_heads": 6,
44
+ "hidden_dim": 384,
45
+ "mlp_dim": 1536,
46
+ "dropout_rate": 0.0,
47
+ "attention_dropout": 0.0,
48
+ "layer_norm_epsilon": 1e-06,
49
+ "use_mha_bias": true,
50
+ "use_mlp_bias": true,
51
+ "use_class_token": false,
52
+ "use_patch_bias": true
53
+ },
54
+ "registered_name": "keras_hub>ViTBackbone"
55
+ },
56
+ "vocabulary_size": 97,
57
+ "max_label_length": 25,
58
+ "decoder_hidden_dim": 384,
59
+ "num_decoder_layers": 1,
60
+ "num_decoder_heads": 12,
61
+ "decoder_mlp_dim": 1536,
62
+ "dropout_rate": 0.1,
63
+ "attention_dropout": 0.1
64
+ },
65
+ "registered_name": "keras_hub>PARSeqBackbone"
66
+ },
67
+ "preprocessor": {
68
+ "module": "keras_hub.src.models.parseq.parseq_causal_lm_preprocessor",
69
+ "class_name": "PARSeqCausalLMPreprocessor",
70
+ "config": {
71
+ "name": "par_seq_causal_lm_preprocessor",
72
+ "trainable": true,
73
+ "dtype": {
74
+ "module": "keras",
75
+ "class_name": "DTypePolicy",
76
+ "config": {
77
+ "name": "float32"
78
+ },
79
+ "registered_name": null
80
+ },
81
+ "tokenizer": {
82
+ "module": "keras_hub.src.models.parseq.parseq_tokenizer",
83
+ "class_name": "PARSeqTokenizer",
84
+ "config": {
85
+ "name": "par_seq_tokenizer",
86
+ "trainable": true,
87
+ "dtype": {
88
+ "module": "keras",
89
+ "class_name": "DTypePolicy",
90
+ "config": {
91
+ "name": "int32"
92
+ },
93
+ "registered_name": null
94
+ },
95
+ "config_file": "tokenizer.json"
96
+ },
97
+ "registered_name": "keras_hub>PARSeqTokenizer"
98
+ },
99
+ "image_converter": {
100
+ "module": "keras_hub.src.models.parseq.parseq_image_converter",
101
+ "class_name": "PARSeqImageConverter",
102
+ "config": {
103
+ "name": "par_seq_image_converter",
104
+ "trainable": true,
105
+ "dtype": {
106
+ "module": "keras",
107
+ "class_name": "DTypePolicy",
108
+ "config": {
109
+ "name": "float32"
110
+ },
111
+ "registered_name": null
112
+ },
113
+ "image_size": [
114
+ 32,
115
+ 128
116
+ ],
117
+ "scale": 0.00784313725490196,
118
+ "offset": -1.0,
119
+ "interpolation": "bicubic",
120
+ "antialias": false,
121
+ "crop_to_aspect_ratio": true,
122
+ "pad_to_aspect_ratio": false,
123
+ "bounding_box_format": "yxyx"
124
+ },
125
+ "registered_name": "keras_hub>PARSeqImageConverter"
126
+ },
127
+ "config_file": "preprocessor.json",
128
+ "sequence_length": 25,
129
+ "add_start_token": true,
130
+ "add_end_token": true
131
+ },
132
+ "registered_name": "keras_hub>PARSeqCausalLMPreprocessor"
133
+ },
134
+ "name": "par_seq_causal_lm",
135
+ "num_perms": 6,
136
+ "add_forward_perms": true,
137
+ "add_mirrored_perms": true,
138
+ "seed": null,
139
+ "end_token_id": 0
140
+ },
141
+ "registered_name": "keras_hub>PARSeqCausalLM"
142
+ }
tokenizer.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.parseq.parseq_tokenizer",
3
+ "class_name": "PARSeqTokenizer",
4
+ "config": {
5
+ "name": "par_seq_tokenizer",
6
+ "trainable": true,
7
+ "dtype": {
8
+ "module": "keras",
9
+ "class_name": "DTypePolicy",
10
+ "config": {
11
+ "name": "int32"
12
+ },
13
+ "registered_name": null
14
+ },
15
+ "config_file": "tokenizer.json"
16
+ },
17
+ "registered_name": "keras_hub>PARSeqTokenizer"
18
+ }