config.json
4.6 KB · 178 lines · json Raw
1 {
2 "_commit_hash": null,
3 "architectures": [
4 "CLIPSegForImageSegmentation"
5 ],
6 "conditional_layer": 0,
7 "decoder_attention_dropout": 0.0,
8 "decoder_hidden_act": "quick_gelu",
9 "decoder_intermediate_size": 2048,
10 "decoder_num_attention_heads": 4,
11 "extract_layers": [
12 3,
13 6,
14 9
15 ],
16 "initializer_factor": 1.0,
17 "logit_scale_init_value": 2.6592,
18 "model_type": "clipseg",
19 "projection_dim": 512,
20 "reduce_dim": 64,
21 "text_config": {
22 "_name_or_path": "",
23 "add_cross_attention": false,
24 "architectures": null,
25 "attention_dropout": 0.0,
26 "bad_words_ids": null,
27 "begin_suppress_tokens": null,
28 "bos_token_id": 0,
29 "chunk_size_feed_forward": 0,
30 "cross_attention_hidden_size": null,
31 "decoder_start_token_id": null,
32 "diversity_penalty": 0.0,
33 "do_sample": false,
34 "dropout": 0.0,
35 "early_stopping": false,
36 "encoder_no_repeat_ngram_size": 0,
37 "eos_token_id": 2,
38 "exponential_decay_length_penalty": null,
39 "finetuning_task": null,
40 "forced_bos_token_id": null,
41 "forced_eos_token_id": null,
42 "hidden_act": "quick_gelu",
43 "hidden_size": 512,
44 "id2label": {
45 "0": "LABEL_0",
46 "1": "LABEL_1"
47 },
48 "initializer_factor": 1.0,
49 "initializer_range": 0.02,
50 "intermediate_size": 2048,
51 "is_decoder": false,
52 "is_encoder_decoder": false,
53 "label2id": {
54 "LABEL_0": 0,
55 "LABEL_1": 1
56 },
57 "layer_norm_eps": 1e-05,
58 "length_penalty": 1.0,
59 "max_length": 20,
60 "max_position_embeddings": 77,
61 "min_length": 0,
62 "model_type": "clipseg_text_model",
63 "no_repeat_ngram_size": 0,
64 "num_attention_heads": 8,
65 "num_beam_groups": 1,
66 "num_beams": 1,
67 "num_hidden_layers": 12,
68 "num_return_sequences": 1,
69 "output_attentions": false,
70 "output_hidden_states": false,
71 "output_scores": false,
72 "pad_token_id": 1,
73 "prefix": null,
74 "problem_type": null,
75 "pruned_heads": {},
76 "remove_invalid_values": false,
77 "repetition_penalty": 1.0,
78 "return_dict": true,
79 "return_dict_in_generate": false,
80 "sep_token_id": null,
81 "suppress_tokens": null,
82 "task_specific_params": null,
83 "temperature": 1.0,
84 "tf_legacy_loss": false,
85 "tie_encoder_decoder": false,
86 "tie_word_embeddings": true,
87 "tokenizer_class": null,
88 "top_k": 50,
89 "top_p": 1.0,
90 "torch_dtype": null,
91 "torchscript": false,
92 "transformers_version": "4.25.0.dev0",
93 "typical_p": 1.0,
94 "use_bfloat16": false,
95 "vocab_size": 49408
96 },
97 "torch_dtype": "float32",
98 "transformers_version": null,
99 "use_complex_transposed_convolution": true,
100 "vision_config": {
101 "_name_or_path": "",
102 "add_cross_attention": false,
103 "architectures": null,
104 "attention_dropout": 0.0,
105 "bad_words_ids": null,
106 "begin_suppress_tokens": null,
107 "bos_token_id": null,
108 "chunk_size_feed_forward": 0,
109 "cross_attention_hidden_size": null,
110 "decoder_start_token_id": null,
111 "diversity_penalty": 0.0,
112 "do_sample": false,
113 "dropout": 0.0,
114 "early_stopping": false,
115 "encoder_no_repeat_ngram_size": 0,
116 "eos_token_id": null,
117 "exponential_decay_length_penalty": null,
118 "finetuning_task": null,
119 "forced_bos_token_id": null,
120 "forced_eos_token_id": null,
121 "hidden_act": "quick_gelu",
122 "hidden_size": 768,
123 "id2label": {
124 "0": "LABEL_0",
125 "1": "LABEL_1"
126 },
127 "image_size": 224,
128 "initializer_factor": 1.0,
129 "initializer_range": 0.02,
130 "intermediate_size": 3072,
131 "is_decoder": false,
132 "is_encoder_decoder": false,
133 "label2id": {
134 "LABEL_0": 0,
135 "LABEL_1": 1
136 },
137 "layer_norm_eps": 1e-05,
138 "length_penalty": 1.0,
139 "max_length": 20,
140 "min_length": 0,
141 "model_type": "clipseg_vision_model",
142 "no_repeat_ngram_size": 0,
143 "num_attention_heads": 12,
144 "num_beam_groups": 1,
145 "num_beams": 1,
146 "num_channels": 3,
147 "num_hidden_layers": 12,
148 "num_return_sequences": 1,
149 "output_attentions": false,
150 "output_hidden_states": false,
151 "output_scores": false,
152 "pad_token_id": null,
153 "patch_size": 16,
154 "prefix": null,
155 "problem_type": null,
156 "pruned_heads": {},
157 "remove_invalid_values": false,
158 "repetition_penalty": 1.0,
159 "return_dict": true,
160 "return_dict_in_generate": false,
161 "sep_token_id": null,
162 "suppress_tokens": null,
163 "task_specific_params": null,
164 "temperature": 1.0,
165 "tf_legacy_loss": false,
166 "tie_encoder_decoder": false,
167 "tie_word_embeddings": true,
168 "tokenizer_class": null,
169 "top_k": 50,
170 "top_p": 1.0,
171 "torch_dtype": null,
172 "torchscript": false,
173 "transformers_version": "4.25.0.dev0",
174 "typical_p": 1.0,
175 "use_bfloat16": false
176 }
177 }
178