Spaces:
Sleeping
Sleeping
Rasmus Lellep
commited on
Commit
·
2370590
1
Parent(s):
a4a8a33
add demo
Browse files- .DS_Store +0 -0
- README.md +7 -7
- app.py +330 -0
- kuidastaltsutadalaamat +1 -0
- requirements.txt +8 -0
.DS_Store
ADDED
Binary file (6.15 kB). View file
|
|
README.md
CHANGED
@@ -1,13 +1,13 @@
|
|
1 |
---
|
2 |
-
title:
|
3 |
emoji: 🐢
|
4 |
colorFrom: indigo
|
5 |
-
colorTo:
|
6 |
sdk: gradio
|
7 |
-
sdk_version: 5.
|
|
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
|
|
1 |
---
|
2 |
+
title: smugri4-1808-ep3
|
3 |
emoji: 🐢
|
4 |
colorFrom: indigo
|
5 |
+
colorTo: blue
|
6 |
sdk: gradio
|
7 |
+
sdk_version: 5.41.0
|
8 |
+
python_version: 3.11
|
9 |
app_file: app.py
|
10 |
pinned: false
|
11 |
+
license: other
|
12 |
+
short_description: Smugri language detection and translation
|
13 |
+
---
|
|
app.py
ADDED
@@ -0,0 +1,330 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import torch
|
3 |
+
from kuidastaltsutadalaamat.trainllm import load_model, load_tokenizer
|
4 |
+
from kuidastaltsutadalaamat.inference import llm_generate
|
5 |
+
from kuidastaltsutadalaamat.data import LazyTokenizingInferenceDataset
|
6 |
+
from kuidastaltsutadalaamat.promptops import *
|
7 |
+
|
8 |
+
accel = None
|
9 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
10 |
+
model = load_model("mphi/smugri4-1808-ep3-tmptest", device, accelerator=accel, attention="eager") #eager for cpu
|
11 |
+
model.eval()
|
12 |
+
tokenizer = load_tokenizer("mphi/smugri4-1808-ep3-tmptest", accelerator=accel)
|
13 |
+
languages = ["English",
|
14 |
+
"English, bible",
|
15 |
+
"English, dictionary",
|
16 |
+
"English, speech",
|
17 |
+
"Erzya",
|
18 |
+
"Erzya, bible",
|
19 |
+
"Erzya, dictionary",
|
20 |
+
"Estonian",
|
21 |
+
"Estonian, Amb, dictionary",
|
22 |
+
"Estonian, Ann, dictionary",
|
23 |
+
"Estonian, Aud, dictionary",
|
24 |
+
"Estonian, Hag, dictionary",
|
25 |
+
"Estonian, Han, dictionary",
|
26 |
+
"Estonian, Hii, emm, dictionary",
|
27 |
+
"Estonian, Hii, käi, dictionary",
|
28 |
+
"Estonian, Hii, phl, dictionary",
|
29 |
+
"Estonian, Hii, rei, dictionary",
|
30 |
+
"Estonian, Hiiu, dictionary",
|
31 |
+
"Estonian, Hjn, dictionary",
|
32 |
+
"Estonian, Hljr, dictionary",
|
33 |
+
"Estonian, Hmd, dictionary",
|
34 |
+
"Estonian, Hää, dictionary",
|
35 |
+
"Estonian, Iisi, dictionary",
|
36 |
+
"Estonian, Iisk, dictionary",
|
37 |
+
"Estonian, Iisr, dictionary",
|
38 |
+
"Estonian, Jjn, dictionary",
|
39 |
+
"Estonian, Jmd, dictionary",
|
40 |
+
"Estonian, Juu, dictionary",
|
41 |
+
"Estonian, Jõer, dictionary",
|
42 |
+
"Estonian, Jõh, dictionary",
|
43 |
+
"Estonian, Jür, dictionary",
|
44 |
+
"Estonian, Kad, dictionary",
|
45 |
+
"Estonian, Kihnu, dictionary",
|
46 |
+
"Estonian, Kjn, dictionary",
|
47 |
+
"Estonian, Kodi, dictionary",
|
48 |
+
"Estonian, Kos, dictionary",
|
49 |
+
"Estonian, Kse, dictionary",
|
50 |
+
"Estonian, Kuuk, dictionary",
|
51 |
+
"Estonian, Kuur, dictionary",
|
52 |
+
"Estonian, Lng, dictionary",
|
53 |
+
"Estonian, Lüg, dictionary",
|
54 |
+
"Estonian, Mar, dictionary",
|
55 |
+
"Estonian, Mih, dictionary",
|
56 |
+
"Estonian, Muhu, dictionary",
|
57 |
+
"Estonian, Mul, hel, dictionary",
|
58 |
+
"Estonian, Mul, hls, dictionary",
|
59 |
+
"Estonian, Mul, krk, dictionary",
|
60 |
+
"Estonian, Mul, trv, dictionary",
|
61 |
+
"Estonian, Mulgi",
|
62 |
+
"Estonian, Mulgi, dictionary",
|
63 |
+
"Estonian, Mär, dictionary",
|
64 |
+
"Estonian, Pjg, dictionary",
|
65 |
+
"Estonian, Plt, dictionary",
|
66 |
+
"Estonian, Ranna, dictionary",
|
67 |
+
"Estonian, Rap, dictionary",
|
68 |
+
"Estonian, Rid, dictionary",
|
69 |
+
"Estonian, Saa, dictionary",
|
70 |
+
"Estonian, Saa, jäm, dictionary",
|
71 |
+
"Estonian, Saa, kaa, dictionary",
|
72 |
+
"Estonian, Saa, khk, dictionary",
|
73 |
+
"Estonian, Saa, kär, dictionary",
|
74 |
+
"Estonian, Saa, mus, dictionary",
|
75 |
+
"Estonian, Saa, pha, dictionary",
|
76 |
+
"Estonian, Saa, pöi, dictionary",
|
77 |
+
"Estonian, Seto",
|
78 |
+
"Estonian, Seto, dictionary",
|
79 |
+
"Estonian, Trm, dictionary",
|
80 |
+
"Estonian, Trt, nõo, dictionary",
|
81 |
+
"Estonian, Trt, ote, dictionary",
|
82 |
+
"Estonian, Trt, puh, dictionary",
|
83 |
+
"Estonian, Trt, ran, dictionary",
|
84 |
+
"Estonian, Tõs, dictionary",
|
85 |
+
"Estonian, Tür, dictionary",
|
86 |
+
"Estonian, Vair, dictionary",
|
87 |
+
"Estonian, Var, dictionary",
|
88 |
+
"Estonian, Vig, dictionary",
|
89 |
+
"Estonian, Vjg, dictionary",
|
90 |
+
"Estonian, Vng, dictionary",
|
91 |
+
"Estonian, Vän, dictionary",
|
92 |
+
"Estonian, bible",
|
93 |
+
"Estonian, dictionary",
|
94 |
+
"Estonian, Äks, dictionary",
|
95 |
+
"Finnish",
|
96 |
+
"Finnish, bible",
|
97 |
+
"Finnish, dictionary",
|
98 |
+
"French",
|
99 |
+
"German",
|
100 |
+
"Hill Mari",
|
101 |
+
"Hill Mari, bible",
|
102 |
+
"Hungarian",
|
103 |
+
"Hungarian, bible",
|
104 |
+
"Inari Sami",
|
105 |
+
"Inari Sami, dictionary",
|
106 |
+
"Izhorian",
|
107 |
+
"Izhorian, Alamaluuga, speech",
|
108 |
+
"Izhorian, Mehmet",
|
109 |
+
"Izhorian, Soikkola",
|
110 |
+
"Izhorian, Soikkola, speech",
|
111 |
+
"Izhorian, speech",
|
112 |
+
"Kazym Khanty",
|
113 |
+
"Kazym Khanty, 2000",
|
114 |
+
"Kazym Khanty, 2013",
|
115 |
+
"Kildin Sami",
|
116 |
+
"Kildin Sami, Orth1",
|
117 |
+
"Kildin Sami, Orth1, dictionary",
|
118 |
+
"Kildin Sami, Orth2",
|
119 |
+
"Kildin Sami, Orth2, dictionary",
|
120 |
+
"Komi-Permyak",
|
121 |
+
"Komi-Permyak, bible",
|
122 |
+
"Komi-Zyrian",
|
123 |
+
"Komi-Zyrian, bible",
|
124 |
+
"Komi-Zyrian, dictionary",
|
125 |
+
"Kven",
|
126 |
+
"Kven, dictionary",
|
127 |
+
"Latvian",
|
128 |
+
"Latvian, bible",
|
129 |
+
"Latvian, dictionary",
|
130 |
+
"Livonian, Idaliivi, ft",
|
131 |
+
"Livonian, Ira, ft",
|
132 |
+
"Livonian, Lääneliivi, ft",
|
133 |
+
"Livonian, Standard",
|
134 |
+
"Livonian, Standard, dictionary",
|
135 |
+
"Livvi",
|
136 |
+
"Livvi, Impilahti",
|
137 |
+
"Livvi, Kondushi",
|
138 |
+
"Livvi, Kotkozero",
|
139 |
+
"Livvi, Nekkula",
|
140 |
+
"Livvi, Newwritten",
|
141 |
+
"Livvi, Oldwritten",
|
142 |
+
"Livvi, Rypushkalitsa",
|
143 |
+
"Livvi, Salmi",
|
144 |
+
"Livvi, Syamozero",
|
145 |
+
"Livvi, Tulmozero",
|
146 |
+
"Livvi, Vedlozero",
|
147 |
+
"Livvi, Vidlitsa",
|
148 |
+
"Livvi, bible",
|
149 |
+
"Ludian",
|
150 |
+
"Ludian, Central",
|
151 |
+
"Ludian, Miikul",
|
152 |
+
"Ludian, Miikul, dictionary",
|
153 |
+
"Ludian, Mikhailovskoye",
|
154 |
+
"Ludian, Newwritten",
|
155 |
+
"Ludian, Northern",
|
156 |
+
"Ludian, Southern",
|
157 |
+
"Lule Sami",
|
158 |
+
"Lule Sami, dictionary",
|
159 |
+
"Mansi, Obs",
|
160 |
+
"Mansi, Sosv",
|
161 |
+
"Mansi, Unk",
|
162 |
+
"Mansi, Unk, bible",
|
163 |
+
"Mansi, Verh",
|
164 |
+
"Meadow Mari",
|
165 |
+
"Meadow Mari, bible",
|
166 |
+
"Meadow Mari, dictionary",
|
167 |
+
"Meänkieli",
|
168 |
+
"Moksha",
|
169 |
+
"Moksha, bible",
|
170 |
+
"Northern Sami",
|
171 |
+
"Northern Sami, dictionary",
|
172 |
+
"Norwegian",
|
173 |
+
"Norwegian, bible",
|
174 |
+
"Norwegian, dictionary",
|
175 |
+
"Pite Sami",
|
176 |
+
"Priur Khanty",
|
177 |
+
"Proper Karelian",
|
178 |
+
"Proper Karelian, Dyorzha",
|
179 |
+
"Proper Karelian, Ilomantsi",
|
180 |
+
"Proper Karelian, Keret",
|
181 |
+
"Proper Karelian, Kestenga",
|
182 |
+
"Proper Karelian, Kontokki",
|
183 |
+
"Proper Karelian, Korbiselga",
|
184 |
+
"Proper Karelian, Myandyselga",
|
185 |
+
"Proper Karelian, Newwritten",
|
186 |
+
"Proper Karelian, Newwrittentver",
|
187 |
+
"Proper Karelian, Oldwritten",
|
188 |
+
"Proper Karelian, Oldwrittentver",
|
189 |
+
"Proper Karelian, Oulanga",
|
190 |
+
"Proper Karelian, Padany",
|
191 |
+
"Proper Karelian, Panozero",
|
192 |
+
"Proper Karelian, Poduzhemye",
|
193 |
+
"Proper Karelian, Porosozero",
|
194 |
+
"Proper Karelian, Reboly",
|
195 |
+
"Proper Karelian, Rugozero",
|
196 |
+
"Proper Karelian, Suistamo",
|
197 |
+
"Proper Karelian, Suoyarvi",
|
198 |
+
"Proper Karelian, Suvi",
|
199 |
+
"Proper Karelian, Tikhtozero",
|
200 |
+
"Proper Karelian, Tikhvin",
|
201 |
+
"Proper Karelian, Tolmachi",
|
202 |
+
"Proper Karelian, Tunguda",
|
203 |
+
"Proper Karelian, Uhta",
|
204 |
+
"Proper Karelian, Valdai",
|
205 |
+
"Proper Karelian, Vesyegonsk",
|
206 |
+
"Proper Karelian, Viena",
|
207 |
+
"Proper Karelian, Voknavolok",
|
208 |
+
"Proper Karelian, Vychetaibola",
|
209 |
+
"Proper Karelian, Yushkozero",
|
210 |
+
"Proper Karelian, bible",
|
211 |
+
"Russian",
|
212 |
+
"Russian, bible",
|
213 |
+
"Russian, dictionary",
|
214 |
+
"Russian, speech",
|
215 |
+
"Shur Khanty",
|
216 |
+
"Shur Khanty, 2013",
|
217 |
+
"Shur Khanty, bible",
|
218 |
+
"Skolt Sami",
|
219 |
+
"Skolt Sami, dictionary",
|
220 |
+
"Southern Sami",
|
221 |
+
"Southern Sami, dictionary",
|
222 |
+
"Sred Khanty",
|
223 |
+
"Surgut Khanty",
|
224 |
+
"Surgut Khanty, 2000",
|
225 |
+
"Surgut Khanty, 2013",
|
226 |
+
"Swedish",
|
227 |
+
"Udmurt",
|
228 |
+
"Udmurt, bible",
|
229 |
+
"Udmurt, dictionary",
|
230 |
+
"Ume Sami",
|
231 |
+
"Unk Khanty",
|
232 |
+
"Vakh Khanty",
|
233 |
+
"Vakh Khanty, 2013",
|
234 |
+
"Veps",
|
235 |
+
"Veps, Centraleastern",
|
236 |
+
"Veps, Centralwestern",
|
237 |
+
"Veps, Newwritten",
|
238 |
+
"Veps, Northern",
|
239 |
+
"Veps, Southern",
|
240 |
+
"Veps, bible",
|
241 |
+
"Votic, I",
|
242 |
+
"Votic, Idavadja, ft",
|
243 |
+
"Votic, J",
|
244 |
+
"Votic, Ja",
|
245 |
+
"Votic, K",
|
246 |
+
"Votic, Ke",
|
247 |
+
"Votic, Kõ",
|
248 |
+
"Votic, L",
|
249 |
+
"Votic, Li",
|
250 |
+
"Votic, Lu",
|
251 |
+
"Votic, Läänevadja, ft",
|
252 |
+
"Votic, M",
|
253 |
+
"Votic, P",
|
254 |
+
"Votic, Po",
|
255 |
+
"Votic, R",
|
256 |
+
"Votic, Ra",
|
257 |
+
"Votic, S",
|
258 |
+
"Votic, Standard",
|
259 |
+
"Votic, U",
|
260 |
+
"Votic, Unk",
|
261 |
+
"Votic, Unk, dictionary",
|
262 |
+
"Votic, Unk, speech",
|
263 |
+
"Votic, V",
|
264 |
+
"Votic, dictionary",
|
265 |
+
"Võro, Har, dictionary",
|
266 |
+
"Võro, Kan, dictionary",
|
267 |
+
"Võro, Krl, dictionary",
|
268 |
+
"Võro, Lei, dictionary",
|
269 |
+
"Võro, Lut, dictionary",
|
270 |
+
"Võro, Plv, dictionary",
|
271 |
+
"Võro, Räp, dictionary",
|
272 |
+
"Võro, Rõu, dictionary",
|
273 |
+
"Võro, Setom",
|
274 |
+
"Võro, Sõnaq",
|
275 |
+
"Võro, Sõnaq, dictionary",
|
276 |
+
"Võro, Uma",
|
277 |
+
"Võro, Urv, dictionary",
|
278 |
+
"Võro, Vas, dictionary",
|
279 |
+
"Võro, X"
|
280 |
+
]
|
281 |
+
|
282 |
+
def run_inference(text, from_lang, to_lang, mode):
|
283 |
+
entry = {"src_segm": text, "task": mode}
|
284 |
+
if mode == "translate":
|
285 |
+
entry.update({"src_lang": from_lang, "tgt_lang": to_lang})
|
286 |
+
prompt_format = PF_SMUGRI_MT
|
287 |
+
else:
|
288 |
+
prompt_format = PF_SMUGRI_LID
|
289 |
+
|
290 |
+
ds = LazyTokenizingInferenceDataset([entry], tokenizer, prompt_format)
|
291 |
+
tok = ds[0]
|
292 |
+
output = llm_generate(model, tokenizer, tok, debug=False, max_len=512)
|
293 |
+
return output[0]
|
294 |
+
|
295 |
+
with gr.Blocks() as demo:
|
296 |
+
text_input = gr.Textbox(label="Text", lines=6, placeholder="Enter text...")
|
297 |
+
identify_btn = gr.Button("Identify language")
|
298 |
+
with gr.Row():
|
299 |
+
from_dropdown = gr.Dropdown(choices=languages, label="From", value=None)
|
300 |
+
to_dropdown = gr.Dropdown(choices=languages, label="To", value=None)
|
301 |
+
translate_btn = gr.Button("Translate", interactive=False)
|
302 |
+
output = gr.Textbox(label="Output", lines=6)
|
303 |
+
|
304 |
+
def toggle_identify(text):
|
305 |
+
return gr.update(interactive=bool(text.strip()))
|
306 |
+
text_input.change(toggle_identify, [text_input], [identify_btn])
|
307 |
+
|
308 |
+
def toggle_translate(text, f, t):
|
309 |
+
return gr.update(interactive=bool(text.strip() and f and t))
|
310 |
+
text_input.change(toggle_translate, [text_input, from_dropdown, to_dropdown], [translate_btn])
|
311 |
+
from_dropdown.change(toggle_translate, [text_input, from_dropdown, to_dropdown], [translate_btn])
|
312 |
+
to_dropdown.change(toggle_translate, [text_input, from_dropdown, to_dropdown], [translate_btn])
|
313 |
+
|
314 |
+
identify_btn.click(
|
315 |
+
fn=lambda text: run_inference(text, None, None, mode="identify"),
|
316 |
+
inputs=[text_input],
|
317 |
+
outputs=[from_dropdown],
|
318 |
+
).then(
|
319 |
+
lambda *args: gr.update(interactive=bool(text_input.value.strip() and from_dropdown.value and to_dropdown.value)),
|
320 |
+
[], [translate_btn]
|
321 |
+
)
|
322 |
+
|
323 |
+
translate_btn.click(
|
324 |
+
fn=lambda text, f, t: run_inference(text, f, t, mode="translate"),
|
325 |
+
inputs=[text_input, from_dropdown, to_dropdown],
|
326 |
+
outputs=[output]
|
327 |
+
)
|
328 |
+
|
329 |
+
if __name__ == "__main__":
|
330 |
+
demo.launch()
|
kuidastaltsutadalaamat
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
Subproject commit efc69e6fd102f04dbf87f0245ec8be99d78d463f
|
requirements.txt
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
gradio==5.41.0
|
2 |
+
torch>=2.1
|
3 |
+
accelerate
|
4 |
+
transformers
|
5 |
+
evaluate
|
6 |
+
packaging
|
7 |
+
ninja
|
8 |
+
#flash_attn
|