Upload model
Browse files- config.json +299 -0
- generation_config.json +9 -0
- model.safetensors +3 -0
config.json
ADDED
@@ -0,0 +1,299 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "text_recognizer_final/checkpoint-2000",
|
3 |
+
"architectures": [
|
4 |
+
"LangVisionEncoderDecoderModel"
|
5 |
+
],
|
6 |
+
"decoder": {
|
7 |
+
"_name_or_path": "",
|
8 |
+
"activation_dropout": 0.0,
|
9 |
+
"activation_function": "gelu",
|
10 |
+
"add_cross_attention": true,
|
11 |
+
"add_final_layer_norm": true,
|
12 |
+
"architectures": [
|
13 |
+
"MBartForCausalLM"
|
14 |
+
],
|
15 |
+
"attention_dropout": 0.0,
|
16 |
+
"bad_words_ids": null,
|
17 |
+
"begin_suppress_tokens": null,
|
18 |
+
"bos_token_id": 0,
|
19 |
+
"chunk_size_feed_forward": 0,
|
20 |
+
"classifier_dropout": 0.0,
|
21 |
+
"cross_attention_hidden_size": null,
|
22 |
+
"d_expert": 1024,
|
23 |
+
"d_model": 1024,
|
24 |
+
"decoder_attention_heads": 16,
|
25 |
+
"decoder_ffn_dim": 4096,
|
26 |
+
"decoder_layerdrop": 0.0,
|
27 |
+
"decoder_layers": 7,
|
28 |
+
"decoder_start_token_id": null,
|
29 |
+
"diversity_penalty": 0.0,
|
30 |
+
"do_sample": false,
|
31 |
+
"dropout": 0.1,
|
32 |
+
"early_stopping": false,
|
33 |
+
"encoder_attention_heads": 16,
|
34 |
+
"encoder_ffn_dim": 4096,
|
35 |
+
"encoder_layerdrop": 0.0,
|
36 |
+
"encoder_layers": 12,
|
37 |
+
"encoder_no_repeat_ngram_size": 0,
|
38 |
+
"eos_token_id": 2,
|
39 |
+
"exponential_decay_length_penalty": null,
|
40 |
+
"finetuning_task": null,
|
41 |
+
"forced_bos_token_id": null,
|
42 |
+
"forced_eos_token_id": 2,
|
43 |
+
"id2label": {
|
44 |
+
"0": "LABEL_0",
|
45 |
+
"1": "LABEL_1"
|
46 |
+
},
|
47 |
+
"init_std": 0.02,
|
48 |
+
"is_decoder": true,
|
49 |
+
"is_encoder_decoder": false,
|
50 |
+
"kv_heads": 4,
|
51 |
+
"label2id": {
|
52 |
+
"LABEL_0": 0,
|
53 |
+
"LABEL_1": 1
|
54 |
+
},
|
55 |
+
"langs": {
|
56 |
+
"af": 65539,
|
57 |
+
"am": 65540,
|
58 |
+
"ar": 65541,
|
59 |
+
"as": 65542,
|
60 |
+
"az": 65543,
|
61 |
+
"be": 65544,
|
62 |
+
"bg": 65545,
|
63 |
+
"bn": 65546,
|
64 |
+
"br": 65547,
|
65 |
+
"bs": 65548,
|
66 |
+
"ca": 65549,
|
67 |
+
"cs": 65550,
|
68 |
+
"cy": 65551,
|
69 |
+
"da": 65552,
|
70 |
+
"de": 65553,
|
71 |
+
"el": 65554,
|
72 |
+
"en": 65555,
|
73 |
+
"eo": 65556,
|
74 |
+
"es": 65557,
|
75 |
+
"et": 65558,
|
76 |
+
"eu": 65559,
|
77 |
+
"fa": 65560,
|
78 |
+
"fi": 65561,
|
79 |
+
"fr": 65562,
|
80 |
+
"fy": 65563,
|
81 |
+
"ga": 65564,
|
82 |
+
"gd": 65565,
|
83 |
+
"gl": 65566,
|
84 |
+
"gu": 65567,
|
85 |
+
"ha": 65568,
|
86 |
+
"he": 65569,
|
87 |
+
"hi": 65570,
|
88 |
+
"hr": 65571,
|
89 |
+
"hu": 65572,
|
90 |
+
"hy": 65573,
|
91 |
+
"id": 65574,
|
92 |
+
"is": 65575,
|
93 |
+
"it": 65576,
|
94 |
+
"ja": 65577,
|
95 |
+
"jv": 65578,
|
96 |
+
"ka": 65579,
|
97 |
+
"kk": 65580,
|
98 |
+
"km": 65581,
|
99 |
+
"kn": 65582,
|
100 |
+
"ko": 65583,
|
101 |
+
"ku": 65584,
|
102 |
+
"ky": 65585,
|
103 |
+
"la": 65586,
|
104 |
+
"lo": 65587,
|
105 |
+
"lt": 65588,
|
106 |
+
"lv": 65589,
|
107 |
+
"mg": 65590,
|
108 |
+
"mk": 65591,
|
109 |
+
"ml": 65592,
|
110 |
+
"mn": 65593,
|
111 |
+
"mr": 65594,
|
112 |
+
"ms": 65595,
|
113 |
+
"my": 65596,
|
114 |
+
"ne": 65597,
|
115 |
+
"nl": 65598,
|
116 |
+
"no": 65599,
|
117 |
+
"om": 65600,
|
118 |
+
"or": 65601,
|
119 |
+
"pa": 65602,
|
120 |
+
"pl": 65603,
|
121 |
+
"ps": 65604,
|
122 |
+
"pt": 65605,
|
123 |
+
"ro": 65606,
|
124 |
+
"ru": 65607,
|
125 |
+
"sa": 65608,
|
126 |
+
"sd": 65609,
|
127 |
+
"si": 65610,
|
128 |
+
"sk": 65611,
|
129 |
+
"sl": 65612,
|
130 |
+
"so": 65613,
|
131 |
+
"sq": 65614,
|
132 |
+
"sr": 65615,
|
133 |
+
"su": 65616,
|
134 |
+
"sv": 65617,
|
135 |
+
"sw": 65618,
|
136 |
+
"ta": 65619,
|
137 |
+
"te": 65620,
|
138 |
+
"th": 65621,
|
139 |
+
"tl": 65622,
|
140 |
+
"tr": 65623,
|
141 |
+
"ug": 65624,
|
142 |
+
"uk": 65625,
|
143 |
+
"ur": 65626,
|
144 |
+
"uz": 65627,
|
145 |
+
"vi": 65628,
|
146 |
+
"xh": 65629,
|
147 |
+
"yi": 65630,
|
148 |
+
"zh": 65631
|
149 |
+
},
|
150 |
+
"length_penalty": 1.0,
|
151 |
+
"max_length": 256,
|
152 |
+
"max_position_embeddings": 1536,
|
153 |
+
"min_length": 0,
|
154 |
+
"model_type": "mbart",
|
155 |
+
"moe_layers": [
|
156 |
+
3
|
157 |
+
],
|
158 |
+
"no_repeat_ngram_size": 0,
|
159 |
+
"num_beam_groups": 1,
|
160 |
+
"num_beams": 1,
|
161 |
+
"num_decoder_layers": 6,
|
162 |
+
"num_hidden_layers": 12,
|
163 |
+
"num_return_sequences": 1,
|
164 |
+
"output_attentions": false,
|
165 |
+
"output_hidden_states": false,
|
166 |
+
"output_scores": false,
|
167 |
+
"pad_token_id": 1,
|
168 |
+
"prefix": null,
|
169 |
+
"problem_type": null,
|
170 |
+
"pruned_heads": {},
|
171 |
+
"remove_invalid_values": false,
|
172 |
+
"repetition_penalty": 1.0,
|
173 |
+
"return_dict": true,
|
174 |
+
"return_dict_in_generate": false,
|
175 |
+
"scale_embedding": true,
|
176 |
+
"sep_token_id": null,
|
177 |
+
"suppress_tokens": null,
|
178 |
+
"task_specific_params": null,
|
179 |
+
"temperature": 1.0,
|
180 |
+
"tf_legacy_loss": false,
|
181 |
+
"tie_encoder_decoder": false,
|
182 |
+
"tie_word_embeddings": true,
|
183 |
+
"tokenizer_class": null,
|
184 |
+
"top_k": 50,
|
185 |
+
"top_p": 1.0,
|
186 |
+
"torch_dtype": "float32",
|
187 |
+
"torchscript": false,
|
188 |
+
"typical_p": 1.0,
|
189 |
+
"use_bfloat16": false,
|
190 |
+
"use_cache": true,
|
191 |
+
"use_moe": true,
|
192 |
+
"vocab_size": 65792
|
193 |
+
},
|
194 |
+
"decoder_start_token_id": 1,
|
195 |
+
"encoder": {
|
196 |
+
"_name_or_path": "",
|
197 |
+
"add_cross_attention": false,
|
198 |
+
"architectures": [
|
199 |
+
"DonutSwinModel"
|
200 |
+
],
|
201 |
+
"attention_probs_dropout_prob": 0.0,
|
202 |
+
"bad_words_ids": null,
|
203 |
+
"begin_suppress_tokens": null,
|
204 |
+
"bos_token_id": null,
|
205 |
+
"chunk_size_feed_forward": 0,
|
206 |
+
"cross_attention_hidden_size": null,
|
207 |
+
"decoder_start_token_id": null,
|
208 |
+
"depths": [
|
209 |
+
2,
|
210 |
+
2,
|
211 |
+
14,
|
212 |
+
2
|
213 |
+
],
|
214 |
+
"diversity_penalty": 0.0,
|
215 |
+
"do_sample": false,
|
216 |
+
"drop_path_rate": 0.1,
|
217 |
+
"early_stopping": false,
|
218 |
+
"embed_dim": 128,
|
219 |
+
"encoder_no_repeat_ngram_size": 0,
|
220 |
+
"eos_token_id": null,
|
221 |
+
"exponential_decay_length_penalty": null,
|
222 |
+
"finetuning_task": null,
|
223 |
+
"forced_bos_token_id": null,
|
224 |
+
"forced_eos_token_id": null,
|
225 |
+
"hidden_act": "gelu",
|
226 |
+
"hidden_dropout_prob": 0.0,
|
227 |
+
"hidden_size": 1024,
|
228 |
+
"id2label": {
|
229 |
+
"0": "LABEL_0",
|
230 |
+
"1": "LABEL_1"
|
231 |
+
},
|
232 |
+
"image_size": [
|
233 |
+
196,
|
234 |
+
896
|
235 |
+
],
|
236 |
+
"initializer_range": 0.02,
|
237 |
+
"is_decoder": false,
|
238 |
+
"is_encoder_decoder": false,
|
239 |
+
"label2id": {
|
240 |
+
"LABEL_0": 0,
|
241 |
+
"LABEL_1": 1
|
242 |
+
},
|
243 |
+
"layer_norm_eps": 1e-05,
|
244 |
+
"length_penalty": 1.0,
|
245 |
+
"max_length": 20,
|
246 |
+
"min_length": 0,
|
247 |
+
"mlp_ratio": 4.0,
|
248 |
+
"model_type": "donut-swin",
|
249 |
+
"no_repeat_ngram_size": 0,
|
250 |
+
"num_beam_groups": 1,
|
251 |
+
"num_beams": 1,
|
252 |
+
"num_channels": 3,
|
253 |
+
"num_heads": [
|
254 |
+
4,
|
255 |
+
8,
|
256 |
+
16,
|
257 |
+
32
|
258 |
+
],
|
259 |
+
"num_layers": 4,
|
260 |
+
"num_return_sequences": 1,
|
261 |
+
"output_attentions": false,
|
262 |
+
"output_hidden_states": false,
|
263 |
+
"output_scores": false,
|
264 |
+
"pad_token_id": null,
|
265 |
+
"patch_size": 4,
|
266 |
+
"path_norm": true,
|
267 |
+
"prefix": null,
|
268 |
+
"problem_type": null,
|
269 |
+
"pruned_heads": {},
|
270 |
+
"qkv_bias": true,
|
271 |
+
"remove_invalid_values": false,
|
272 |
+
"repetition_penalty": 1.0,
|
273 |
+
"return_dict": true,
|
274 |
+
"return_dict_in_generate": false,
|
275 |
+
"sep_token_id": null,
|
276 |
+
"suppress_tokens": null,
|
277 |
+
"task_specific_params": null,
|
278 |
+
"temperature": 1.0,
|
279 |
+
"tf_legacy_loss": false,
|
280 |
+
"tie_encoder_decoder": false,
|
281 |
+
"tie_word_embeddings": true,
|
282 |
+
"tokenizer_class": null,
|
283 |
+
"top_k": 50,
|
284 |
+
"top_p": 1.0,
|
285 |
+
"torch_dtype": "float32",
|
286 |
+
"torchscript": false,
|
287 |
+
"typical_p": 1.0,
|
288 |
+
"use_2d_embeddings": false,
|
289 |
+
"use_absolute_embeddings": true,
|
290 |
+
"use_bfloat16": false,
|
291 |
+
"window_size": 7
|
292 |
+
},
|
293 |
+
"is_encoder_decoder": true,
|
294 |
+
"model_type": "vision-encoder-decoder",
|
295 |
+
"pad_token_id": 0,
|
296 |
+
"tie_word_embeddings": false,
|
297 |
+
"torch_dtype": "float16",
|
298 |
+
"transformers_version": "4.36.2"
|
299 |
+
}
|
generation_config.json
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_from_model_config": true,
|
3 |
+
"bos_token_id": 0,
|
4 |
+
"eos_token_id": 2,
|
5 |
+
"forced_eos_token_id": 2,
|
6 |
+
"max_length": 512,
|
7 |
+
"pad_token_id": 1,
|
8 |
+
"transformers_version": "4.36.2"
|
9 |
+
}
|
model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:922fd2283c7a9fafcc171b4840cf19aab4a70a890ec350adc832a8be9846b7ac
|
3 |
+
size 1048833584
|