MuteJester commited on
Commit
1e943ad
Β·
verified Β·
1 Parent(s): 1aa5cdb

Upload AlignAIR 3.0 pretrained models (5 models)

Browse files
HUMAN_IGH_EXTENDED_576/README.md ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # AlignAIRR Model Bundle
2
+
3
+ Format Version: 3
4
+ Model Type: single_chain
5
+ Max Sequence Length: 576
6
+
7
+ ## Usage
8
+ ```python
9
+ from AlignAIR.Models import SingleChainAlignAIR
10
+ model = SingleChainAlignAIR.from_pretrained('PATH_TO_BUNDLE')
11
+ ```
12
+
13
+ ## Integrity
14
+ SHA256 fingerprint stored in fingerprint.txt. Do not modify bundle contents manually.
HUMAN_IGH_EXTENDED_576/VERSION ADDED
@@ -0,0 +1 @@
 
 
1
+ FORMAT_VERSION=3
HUMAN_IGH_EXTENDED_576/config.json ADDED
@@ -0,0 +1,434 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignairr_version": "2.0.2",
3
+ "chain_types": null,
4
+ "created_utc": "2026-03-03T15:07:27.032391+00:00",
5
+ "d_allele_count": 38,
6
+ "d_allele_latent_size": null,
7
+ "dataconfig_hash": "dfd7f5fe084ba7ded1c0d13c46ff25f8fe1a119c7290ceee819a87148e2f53bc",
8
+ "dataconfig_metadata": {
9
+ "allele_counts": {
10
+ "d": 37,
11
+ "j": 10,
12
+ "v": 342
13
+ },
14
+ "allele_names": {
15
+ "d": [
16
+ "IGHD1-5ABL*01",
17
+ "IGHD1-5ABL*02",
18
+ "IGHD1-5ABL*03",
19
+ "IGHD1-5ABL*04",
20
+ "IGHD1-5ABL*05",
21
+ "IGHD1-FLTI*01",
22
+ "IGHD2-7XRZ*02",
23
+ "IGHD2-7XRZ*03",
24
+ "IGHD2-7XRZ*04",
25
+ "IGHD2-AL4M*01",
26
+ "IGHD2-AL4M*02",
27
+ "IGHD2-IKOP*01",
28
+ "IGHD2-IKOP*02",
29
+ "IGHD2-XRP6*01",
30
+ "IGHD3-3HNK*01",
31
+ "IGHD3-3HNK*02",
32
+ "IGHD3-FMCQ*01",
33
+ "IGHD3-FMCQ*02",
34
+ "IGHD3-I3HM*01",
35
+ "IGHD3-JHII*01",
36
+ "IGHD3-LDJE*01",
37
+ "IGHD3-LDJE*02",
38
+ "IGHD4-ALRC*01",
39
+ "IGHD4-ALRC*02",
40
+ "IGHD4-F473*01",
41
+ "IGHD4-F473*02",
42
+ "IGHD5-36AY*01",
43
+ "IGHD5-36AY*02",
44
+ "IGHD5-36AY*03",
45
+ "IGHD5-CGIC*01",
46
+ "IGHD5-O3PH*01",
47
+ "IGHD6-5VO4*01",
48
+ "IGHD6-5VO4*02",
49
+ "IGHD6-5X42*01",
50
+ "IGHD6-G3XM*01",
51
+ "IGHD6-G3XM*02",
52
+ "IGHD7-3VE7*01"
53
+ ],
54
+ "j": [
55
+ "IGHJ1-VWSM*01",
56
+ "IGHJ2-AZGQ*01",
57
+ "IGHJ3-CBX5*01",
58
+ "IGHJ4-AG67*01",
59
+ "IGHJ4-AG67*02",
60
+ "IGHJ5-2TKS*01",
61
+ "IGHJ6-2SFB*01",
62
+ "IGHJ6-2SFB*02",
63
+ "IGHJ6-2SFB*03",
64
+ "IGHJ6-2SFB*04"
65
+ ],
66
+ "v": [
67
+ "IGHV1-27JB*01",
68
+ "IGHV1-27JB*02",
69
+ "IGHV1-27JB*03",
70
+ "IGHV1-27JB*04",
71
+ "IGHV1-27JB*05",
72
+ "IGHV1-27JB*06",
73
+ "IGHV1-27JB*07",
74
+ "IGHV1-27JB*08",
75
+ "IGHV1-27JB*09",
76
+ "IGHV1-2DHF*01",
77
+ "IGHV1-2DHF*02",
78
+ "IGHV1-2DHF*03",
79
+ "IGHV1-2DHF*04",
80
+ "IGHV1-2DHF*05",
81
+ "IGHV1-2DHF*06",
82
+ "IGHV1-2DHF*07",
83
+ "IGHV1-2DHF*08",
84
+ "IGHV1-2DHF*09",
85
+ "IGHV1-2DHF*10",
86
+ "IGHV1-2DHF*11",
87
+ "IGHV1-2DHF*12",
88
+ "IGHV1-2DHF*13",
89
+ "IGHV1-2DHF*14",
90
+ "IGHV1-2DHF*15",
91
+ "IGHV1-2DHF*16",
92
+ "IGHV1-2DHF*17",
93
+ "IGHV1-2DHF*18",
94
+ "IGHV1-2DHF*19",
95
+ "IGHV1-2DHF*20",
96
+ "IGHV1-2DHF*21",
97
+ "IGHV1-2DHF*22",
98
+ "IGHV1-2DHF*23",
99
+ "IGHV1-2DHF*24",
100
+ "IGHV1-2DHF*25",
101
+ "IGHV1-2DHF*26",
102
+ "IGHV1-2DHF*27",
103
+ "IGHV1-2DHF*28",
104
+ "IGHV1-2DHF*29",
105
+ "IGHV1-2DHF*30",
106
+ "IGHV1-2DHF*31",
107
+ "IGHV1-2DHF*32",
108
+ "IGHV1-5JOY*01",
109
+ "IGHV1-5JOY*02",
110
+ "IGHV1-5JOY*03",
111
+ "IGHV1-5JOY*04",
112
+ "IGHV1-5JOY*05",
113
+ "IGHV1-5LBE*01",
114
+ "IGHV1-5LBE*02",
115
+ "IGHV1-5LBE*03",
116
+ "IGHV1-5LBE*04",
117
+ "IGHV1-5LBE*05",
118
+ "IGHV1-ARUH*01",
119
+ "IGHV1-ARUH*02",
120
+ "IGHV1-ARUH*03",
121
+ "IGHV1-ARUH*04",
122
+ "IGHV1-ARUH*05",
123
+ "IGHV1-ARUH*06",
124
+ "IGHV1-AUTT*01",
125
+ "IGHV1-AUTT*02",
126
+ "IGHV1-AVZG*01",
127
+ "IGHV1-AVZG*02",
128
+ "IGHV1-AVZG*03",
129
+ "IGHV1-AVZG*04",
130
+ "IGHV1-AVZG*05",
131
+ "IGHV1-AVZG*06",
132
+ "IGHV1-F2NP*01",
133
+ "IGHV1-F2NP*02",
134
+ "IGHV1-F2NP*03",
135
+ "IGHV1-F2NP*04",
136
+ "IGHV1-F2NP*05",
137
+ "IGHV1-L3PD*01",
138
+ "IGHV1-L3PD*02",
139
+ "IGHV1-L3PD*03",
140
+ "IGHV1-N56A*01",
141
+ "IGHV1-N56A*02",
142
+ "IGHV2-2KSE*01",
143
+ "IGHV2-2KSE*02",
144
+ "IGHV2-2KSE*03",
145
+ "IGHV2-2KSE*04",
146
+ "IGHV2-2KSE*05",
147
+ "IGHV2-2KSE*06",
148
+ "IGHV2-2KSE*07",
149
+ "IGHV2-2KSE*08",
150
+ "IGHV2-2KSE*09",
151
+ "IGHV2-2KSE*10",
152
+ "IGHV2-2KSE*11",
153
+ "IGHV2-2KSE*12",
154
+ "IGHV2-4ZTD*01",
155
+ "IGHV2-4ZTD*03",
156
+ "IGHV2-4ZTD*04",
157
+ "IGHV2-4ZTD*05",
158
+ "IGHV2-4ZTD*06",
159
+ "IGHV2-4ZTD*08",
160
+ "IGHV2-HPQR*01",
161
+ "IGHV2-HPQR*02",
162
+ "IGHV2-HPQR*03",
163
+ "IGHV2-HPQR*04",
164
+ "IGHV2-HPQR*05",
165
+ "IGHV3-24HV*01",
166
+ "IGHV3-24HV*02",
167
+ "IGHV3-24HV*03",
168
+ "IGHV3-2WLF*01",
169
+ "IGHV3-3CQT*01",
170
+ "IGHV3-3CQT*02",
171
+ "IGHV3-3CQT*03",
172
+ "IGHV3-3CQT*04",
173
+ "IGHV3-3CQT*05",
174
+ "IGHV3-3CQT*06",
175
+ "IGHV3-3LLP*01",
176
+ "IGHV3-3LLP*02",
177
+ "IGHV3-3LLP*03",
178
+ "IGHV3-3LLP*04",
179
+ "IGHV3-3LLP*05",
180
+ "IGHV3-3LLP*06",
181
+ "IGHV3-3OXN*01",
182
+ "IGHV3-3OXN*02",
183
+ "IGHV3-3OXN*03",
184
+ "IGHV3-3OXN*04",
185
+ "IGHV3-3OXN*05",
186
+ "IGHV3-3OXN*06",
187
+ "IGHV3-3OXN*07",
188
+ "IGHV3-3OXN*08",
189
+ "IGHV3-3OXN*09",
190
+ "IGHV3-3OXN*10",
191
+ "IGHV3-3OXN*11",
192
+ "IGHV3-3OXN*12",
193
+ "IGHV3-3OXN*13",
194
+ "IGHV3-3OXN*14",
195
+ "IGHV3-3OXN*15",
196
+ "IGHV3-3OXN*16",
197
+ "IGHV3-3OXN*17",
198
+ "IGHV3-3OXN*18",
199
+ "IGHV3-3OXN*19",
200
+ "IGHV3-3OXN*20",
201
+ "IGHV3-3OXN*21",
202
+ "IGHV3-3OXN*22",
203
+ "IGHV3-3OXN*23",
204
+ "IGHV3-3OXN*24",
205
+ "IGHV3-3OXN*25",
206
+ "IGHV3-3PAY*01",
207
+ "IGHV3-3PAY*02",
208
+ "IGHV3-3PAY*03",
209
+ "IGHV3-3PAY*04",
210
+ "IGHV3-3PAY*05",
211
+ "IGHV3-3PAY*06",
212
+ "IGHV3-3PAY*07",
213
+ "IGHV3-3PAY*08",
214
+ "IGHV3-3VUL*01",
215
+ "IGHV3-3VUL*02",
216
+ "IGHV3-3VUL*03",
217
+ "IGHV3-3XNV*01",
218
+ "IGHV3-3XNV*03",
219
+ "IGHV3-3XNV*04",
220
+ "IGHV3-3XNV*05",
221
+ "IGHV3-3XNV*06",
222
+ "IGHV3-3XNV*07",
223
+ "IGHV3-3XNV*08",
224
+ "IGHV3-3XNV*09",
225
+ "IGHV3-42YL*01",
226
+ "IGHV3-42YL*02",
227
+ "IGHV3-42YL*03",
228
+ "IGHV3-42YL*04",
229
+ "IGHV3-42YL*05",
230
+ "IGHV3-42YL*06",
231
+ "IGHV3-42YL*08",
232
+ "IGHV3-43PC*01",
233
+ "IGHV3-43PC*02",
234
+ "IGHV3-43PC*03",
235
+ "IGHV3-43PC*04",
236
+ "IGHV3-4JBP*01",
237
+ "IGHV3-4JBP*02",
238
+ "IGHV3-4JBP*03",
239
+ "IGHV3-4JBP*04",
240
+ "IGHV3-4LYT*01",
241
+ "IGHV3-4LYT*02",
242
+ "IGHV3-4LYT*03",
243
+ "IGHV3-4LYT*04",
244
+ "IGHV3-4LYT*05",
245
+ "IGHV3-4LYT*06",
246
+ "IGHV3-4XBI*01",
247
+ "IGHV3-4XBI*02",
248
+ "IGHV3-4XBI*03",
249
+ "IGHV3-4XBI*04",
250
+ "IGHV3-4XBI*05",
251
+ "IGHV3-4XBI*06",
252
+ "IGHV3-55JG*01",
253
+ "IGHV3-55JG*03",
254
+ "IGHV3-55JG*05",
255
+ "IGHV3-6AE4*01",
256
+ "IGHV3-6AE4*02",
257
+ "IGHV3-6AE4*03",
258
+ "IGHV3-6AE4*04",
259
+ "IGHV3-6B3Z*01",
260
+ "IGHV3-6B3Z*02",
261
+ "IGHV3-6B3Z*03",
262
+ "IGHV3-6B3Z*04",
263
+ "IGHV3-6B3Z*05",
264
+ "IGHV3-6B3Z*06",
265
+ "IGHV3-6B3Z*07",
266
+ "IGHV3-6B3Z*10",
267
+ "IGHV3-6B3Z*11",
268
+ "IGHV3-6B3Z*14",
269
+ "IGHV3-6B3Z*15",
270
+ "IGHV3-6B3Z*16",
271
+ "IGHV3-6B3Z*17",
272
+ "IGHV3-6B3Z*18",
273
+ "IGHV3-6B3Z*21",
274
+ "IGHV3-6OT2*02",
275
+ "IGHV3-6OT2*04",
276
+ "IGHV3-6OT2*05",
277
+ "IGHV3-7A77*01",
278
+ "IGHV3-7A77*02",
279
+ "IGHV3-7A77*03",
280
+ "IGHV3-7A77*04",
281
+ "IGHV3-7A77*05",
282
+ "IGHV3-7EYU*01",
283
+ "IGHV3-7EYU*02",
284
+ "IGHV3-7EYU*03",
285
+ "IGHV3-AJDG*01",
286
+ "IGHV3-AJDG*02",
287
+ "IGHV3-AQ77*01",
288
+ "IGHV3-AQ77*02",
289
+ "IGHV3-AQ77*03",
290
+ "IGHV3-AQ77*04",
291
+ "IGHV3-AQ77*05",
292
+ "IGHV3-BR4P*01",
293
+ "IGHV3-FBDZ*01",
294
+ "IGHV3-FBDZ*02",
295
+ "IGHV3-ITHP*01",
296
+ "IGHV3-ITHP*02",
297
+ "IGHV3-ITHP*03",
298
+ "IGHV3-JMEH*01",
299
+ "IGHV3-JMEH*02",
300
+ "IGHV3-JMEH*03",
301
+ "IGHV3-KNBP*01",
302
+ "IGHV3-KQKW*01",
303
+ "IGHV3-KQKW*02",
304
+ "IGHV3-KQKW*04",
305
+ "IGHV3-NYRR*02",
306
+ "IGHV3-OAAD*01",
307
+ "IGHV3-OAAD*02",
308
+ "IGHV4-2352*01",
309
+ "IGHV4-2352*02",
310
+ "IGHV4-2352*03",
311
+ "IGHV4-2352*04",
312
+ "IGHV4-2352*06",
313
+ "IGHV4-2352*07",
314
+ "IGHV4-2352*08",
315
+ "IGHV4-2352*09",
316
+ "IGHV4-2352*10",
317
+ "IGHV4-2352*11",
318
+ "IGHV4-2352*12",
319
+ "IGHV4-2352*13",
320
+ "IGHV4-2352*14",
321
+ "IGHV4-2352*15",
322
+ "IGHV4-2352*16",
323
+ "IGHV4-2352*17",
324
+ "IGHV4-2P6V*01",
325
+ "IGHV4-2P6V*02",
326
+ "IGHV4-2P6V*03",
327
+ "IGHV4-2P6V*04",
328
+ "IGHV4-2P6V*05",
329
+ "IGHV4-2P6V*06",
330
+ "IGHV4-2P6V*07",
331
+ "IGHV4-3GMG*02",
332
+ "IGHV4-3GMG*03",
333
+ "IGHV4-3GMG*04",
334
+ "IGHV4-4MQ4*01",
335
+ "IGHV4-4MQ4*02",
336
+ "IGHV4-4MQ4*03",
337
+ "IGHV4-4RVQ*01",
338
+ "IGHV4-4RVQ*02",
339
+ "IGHV4-4RVQ*03",
340
+ "IGHV4-4SZD*01",
341
+ "IGHV4-4SZD*02",
342
+ "IGHV4-4SZD*03",
343
+ "IGHV4-4SZD*04",
344
+ "IGHV4-4SZD*05",
345
+ "IGHV4-4SZD*06",
346
+ "IGHV4-4SZD*07",
347
+ "IGHV4-4SZD*08",
348
+ "IGHV4-4SZD*09",
349
+ "IGHV4-4SZD*10",
350
+ "IGHV4-4SZD*11",
351
+ "IGHV4-4SZD*12",
352
+ "IGHV4-4SZD*13",
353
+ "IGHV4-6RTC*01",
354
+ "IGHV4-6RTC*02",
355
+ "IGHV4-6RTC*03",
356
+ "IGHV4-6RTC*04",
357
+ "IGHV4-6RTC*05",
358
+ "IGHV4-6RTC*06",
359
+ "IGHV4-6RTC*07",
360
+ "IGHV4-6RTC*08",
361
+ "IGHV4-6RTC*09",
362
+ "IGHV4-6VHY*01",
363
+ "IGHV4-6VHY*02",
364
+ "IGHV4-6VHY*03",
365
+ "IGHV4-6VHY*04",
366
+ "IGHV4-6VHY*05",
367
+ "IGHV4-6VHY*06",
368
+ "IGHV4-6VHY*07",
369
+ "IGHV4-6VHY*08",
370
+ "IGHV4-6VHY*09",
371
+ "IGHV4-6VHY*10",
372
+ "IGHV4-CAUG*01",
373
+ "IGHV4-CAUG*02",
374
+ "IGHV4-CAUG*03",
375
+ "IGHV4-CAUG*04",
376
+ "IGHV4-CAUG*05",
377
+ "IGHV4-CKBR*01",
378
+ "IGHV4-CKBR*02",
379
+ "IGHV4-CKBR*03",
380
+ "IGHV4-CKBR*04",
381
+ "IGHV4-CKBR*05",
382
+ "IGHV4-CKBR*06",
383
+ "IGHV4-EMAS*01",
384
+ "IGHV4-EMAS*02",
385
+ "IGHV4-EMAS*03",
386
+ "IGHV4-EMAS*04",
387
+ "IGHV5-4DTG*01",
388
+ "IGHV5-4DTG*02",
389
+ "IGHV5-4DTG*03",
390
+ "IGHV5-4DTG*04",
391
+ "IGHV5-BB4N*01",
392
+ "IGHV5-BB4N*02",
393
+ "IGHV5-BB4N*03",
394
+ "IGHV5-BB4N*04",
395
+ "IGHV6-CLFH*01",
396
+ "IGHV6-CLFH*02",
397
+ "IGHV7-4XK4*02",
398
+ "IGHV7-4XK4*03",
399
+ "IGHV7-C7GW*01",
400
+ "IGHV7-C7GW*02",
401
+ "IGHV7-C7GW*03",
402
+ "IGHV7-C7GW*04",
403
+ "IGHV7-C7GW*05",
404
+ "IGHV8-DOWS*01",
405
+ "IGHV8-DOWS*02",
406
+ "IGHV8-DOWS*04",
407
+ "IGHV8-DOWS*06",
408
+ "IGHV8-DOWS*07"
409
+ ]
410
+ },
411
+ "chain_type": "BCR_HEAVY",
412
+ "has_d": true,
413
+ "last_updated": "2025-08-02",
414
+ "reference_set": "OGRDB V8 extended",
415
+ "species": "Human"
416
+ },
417
+ "extra": null,
418
+ "format_version": 3,
419
+ "framework": "pytorch",
420
+ "framework_version": "pytorch-2.10.0+cu128",
421
+ "genairr_version": "0.6.3",
422
+ "git_commit": "2f2988e",
423
+ "has_d_gene": true,
424
+ "has_orientation_head": true,
425
+ "j_allele_count": 10,
426
+ "j_allele_latent_size": null,
427
+ "max_seq_length": 576,
428
+ "model_type": "single_chain",
429
+ "notes": null,
430
+ "number_of_chains": null,
431
+ "python_version": "3.12.3",
432
+ "v_allele_count": 342,
433
+ "v_allele_latent_size": null
434
+ }
HUMAN_IGH_EXTENDED_576/dataconfig.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78b38050920cf3cd45cff62b770b9c3495818b2fde50aaa50f6adccf11b70453
3
+ size 16796088
HUMAN_IGH_EXTENDED_576/fingerprint.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ 43d5c77da3d4aefa00a69785c9897837388d2d48cb499046c359d368f36ef5b0
HUMAN_IGH_EXTENDED_576/model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31430081031b9e3da952325dfa47be89aa38975005981191375642eb7196c9d5
3
+ size 11996687
HUMAN_IGH_EXTENDED_576/training_meta.json ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "batch_size": 512,
3
+ "best_epoch": 995,
4
+ "best_loss": 1.9363010847568511,
5
+ "created_utc": "2026-03-03T15:07:27.033542+00:00",
6
+ "epochs_trained": 1000,
7
+ "extra": {
8
+ "notes": "IGH HUMAN_IGH_EXTENDED (V=342, D=37, J=10), CrossAttentionPooling, Uniform mutation (0.01-0.05), Kendall uncertainty weighting, orientation head, sigma schedule cosine 3.0\u21920.75, 6 workers"
9
+ },
10
+ "final_epoch": 999,
11
+ "final_loss": 1.9802256536483764,
12
+ "learning_rate": "1.4999999999999999e-05",
13
+ "metrics_summary": {
14
+ "classification": 0.19060173973441125,
15
+ "d_allele_avg_gt_count": 1.4607844352722168,
16
+ "d_allele_setK_precision": 0.8681917211328976,
17
+ "d_allele_setK_recall": 0.8681917211328976,
18
+ "d_allele_thresh_precision": 0.8330066204071045,
19
+ "d_allele_thresh_recall": 0.8193464279174805,
20
+ "d_allele_top1_acc": 0.8607843518257141,
21
+ "d_allele_top1_conf": 0.8420405983924866,
22
+ "d_end_acc": 0.7196078896522522,
23
+ "d_end_acc_1nt": 0.8294118046760559,
24
+ "d_end_mae": 0.8843137621879578,
25
+ "d_start_acc": 0.7098039388656616,
26
+ "d_start_acc_1nt": 0.8352941870689392,
27
+ "d_start_mae": 0.8235294818878174,
28
+ "indel_count": 0.25063891351222994,
29
+ "indel_count_mae": 0.310779333114624,
30
+ "j_allele_avg_gt_count": 1.0,
31
+ "j_allele_setK_precision": 0.984313725490196,
32
+ "j_allele_setK_recall": 0.984313725490196,
33
+ "j_allele_thresh_precision": 0.9843137860298157,
34
+ "j_allele_thresh_recall": 0.9843137860298157,
35
+ "j_allele_top1_acc": 0.9843137860298157,
36
+ "j_allele_top1_conf": 0.9409783482551575,
37
+ "j_end_acc": 1.0,
38
+ "j_end_acc_1nt": 1.0,
39
+ "j_end_mae": 0.0,
40
+ "j_start_acc": 0.9745098352432251,
41
+ "j_start_acc_1nt": 0.9980392456054688,
42
+ "j_start_mae": 0.027450982481241226,
43
+ "log_var/log_var_d_classification": -0.8180696344375611,
44
+ "log_var/log_var_d_end": 1.1709008902311324,
45
+ "log_var/log_var_d_start": 1.1773365020751954,
46
+ "log_var/log_var_indel": -0.5003187489509583,
47
+ "log_var/log_var_j_classification": -0.9034614259004593,
48
+ "log_var/log_var_j_end": 0.8181240218877792,
49
+ "log_var/log_var_j_start": 0.8338462087512016,
50
+ "log_var/log_var_junction_region": -1.8439301317930221,
51
+ "log_var/log_var_mutation": -3.0,
52
+ "log_var/log_var_orientation": -3.0,
53
+ "log_var/log_var_productivity": -3.0,
54
+ "log_var/log_var_v_classification": -0.9100403302907943,
55
+ "log_var/log_var_v_end": 0.9727652844786644,
56
+ "log_var/log_var_v_start": 0.9656437847018242,
57
+ "loss": 1.9802256536483764,
58
+ "lr": 1.4999999999999999e-05,
59
+ "mutation_rate": -1.40669819355011,
60
+ "mutation_rate_mae": 0.004147736355662346,
61
+ "orientation": -1.4999999988079071,
62
+ "orientation_acc": 1.0,
63
+ "orientation_acc/complement": 0.9999999403953552,
64
+ "orientation_acc/normal": 1.0,
65
+ "orientation_acc/revcomp": 1.0,
66
+ "orientation_acc/reversed": 1.0,
67
+ "productive": -1.2632013711333274,
68
+ "productive_acc": 1.0,
69
+ "raw/d_allele": 0.22103169716894627,
70
+ "raw/d_end": 1.6159657567739487,
71
+ "raw/d_start": 1.6268917560577392,
72
+ "raw/frame_loss": 0.21468963261693716,
73
+ "raw/hinge_loss": 0.00041688800696647375,
74
+ "raw/indel_count": 0.30365239836275576,
75
+ "raw/iou_loss": 0.15741808891296385,
76
+ "raw/j_allele": 0.2025365286320448,
77
+ "raw/j_end": 1.1331209546327592,
78
+ "raw/j_start": 1.1501597076654435,
79
+ "raw/junction_region": 0.07937577821314334,
80
+ "raw/len_loss": 1.0310878744721412,
81
+ "raw/mutation_rate": 0.004645223214756697,
82
+ "raw/ordering_loss": 0.0011530860774928442,
83
+ "raw/orientation": 6.311059674857233e-11,
84
+ "raw/productive": 0.011789509553054814,
85
+ "raw/regularization": 0.029999999329447746,
86
+ "raw/v_allele": 0.20122628405690193,
87
+ "raw/v_end": 1.3210634535551071,
88
+ "raw/v_start": 1.3098637717962265,
89
+ "segmentation": 5.678884587287903,
90
+ "sigma": 0.75,
91
+ "v_allele_avg_gt_count": 2.401960849761963,
92
+ "v_allele_setK_precision": 0.9143497134673606,
93
+ "v_allele_setK_recall": 0.9143497134673606,
94
+ "v_allele_thresh_precision": 0.8683497309684753,
95
+ "v_allele_thresh_recall": 0.8631495237350464,
96
+ "v_allele_top1_acc": 0.9098039865493774,
97
+ "v_allele_top1_conf": 0.8282101154327393,
98
+ "v_end_acc": 0.8666667342185974,
99
+ "v_end_acc_1nt": 0.9313725829124451,
100
+ "v_end_mae": 0.2764706015586853,
101
+ "v_start_acc": 0.8941177129745483,
102
+ "v_start_acc_1nt": 0.9529412388801575,
103
+ "v_start_mae": 0.24901962280273438
104
+ },
105
+ "mixed_precision": true,
106
+ "optimizer_class": "AdamW",
107
+ "samples_per_epoch": 102400,
108
+ "wall_time_seconds": 75114
109
+ }
HUMAN_IGH_OGRDB_576/README.md ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # AlignAIRR Model Bundle
2
+
3
+ Format Version: 3
4
+ Model Type: single_chain
5
+ Max Sequence Length: 576
6
+
7
+ ## Usage
8
+ ```python
9
+ from AlignAIR.Models import SingleChainAlignAIR
10
+ model = SingleChainAlignAIR.from_pretrained('PATH_TO_BUNDLE')
11
+ ```
12
+
13
+ ## Integrity
14
+ SHA256 fingerprint stored in fingerprint.txt. Do not modify bundle contents manually.
HUMAN_IGH_OGRDB_576/VERSION ADDED
@@ -0,0 +1 @@
 
 
1
+ FORMAT_VERSION=3
HUMAN_IGH_OGRDB_576/config.json ADDED
@@ -0,0 +1,283 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignairr_version": "2.0.2",
3
+ "chain_types": null,
4
+ "created_utc": "2026-03-01T15:40:13.636358+00:00",
5
+ "d_allele_count": 34,
6
+ "d_allele_latent_size": null,
7
+ "dataconfig_hash": "f9e3e35bd4a48280b15ec4817d1238a7f495f3452af1201d2ceb7d004e6830e0",
8
+ "dataconfig_metadata": {
9
+ "allele_counts": {
10
+ "d": 33,
11
+ "j": 7,
12
+ "v": 198
13
+ },
14
+ "allele_names": {
15
+ "d": [
16
+ "IGHD1-1*01",
17
+ "IGHD1-14*01",
18
+ "IGHD1-20*01",
19
+ "IGHD1-26*01",
20
+ "IGHD1-7*01",
21
+ "IGHD2-15*01",
22
+ "IGHD2-2*01",
23
+ "IGHD2-2*02",
24
+ "IGHD2-21*01",
25
+ "IGHD2-21*02",
26
+ "IGHD2-8*01",
27
+ "IGHD2-8*02",
28
+ "IGHD3-10*01",
29
+ "IGHD3-10*03",
30
+ "IGHD3-16*02",
31
+ "IGHD3-16*03",
32
+ "IGHD3-22*01",
33
+ "IGHD3-3*01",
34
+ "IGHD3-9*01",
35
+ "IGHD4-11*01",
36
+ "IGHD4-17*01",
37
+ "IGHD4-23*01",
38
+ "IGHD4-4*01",
39
+ "IGHD5-12*01",
40
+ "IGHD5-18*01",
41
+ "IGHD5-18*02",
42
+ "IGHD5-24*01",
43
+ "IGHD5-5*01",
44
+ "IGHD6-13*01",
45
+ "IGHD6-19*01",
46
+ "IGHD6-25*01",
47
+ "IGHD6-6*01",
48
+ "IGHD7-27*01"
49
+ ],
50
+ "j": [
51
+ "IGHJ1*01",
52
+ "IGHJ2*01",
53
+ "IGHJ3*02",
54
+ "IGHJ4*02",
55
+ "IGHJ5*02",
56
+ "IGHJ6*02",
57
+ "IGHJ6*03"
58
+ ],
59
+ "v": [
60
+ "IGHVF1-G1*01",
61
+ "IGHVF1-G1*02",
62
+ "IGHVF1-G1*03",
63
+ "IGHVF1-G2*01",
64
+ "IGHVF1-G2*02",
65
+ "IGHVF1-G3*01",
66
+ "IGHVF1-G3*02",
67
+ "IGHVF1-G3*03",
68
+ "IGHVF1-G3*04",
69
+ "IGHVF1-G3*05",
70
+ "IGHVF1-G3*06",
71
+ "IGHVF10-G33*01",
72
+ "IGHVF10-G33*02",
73
+ "IGHVF10-G33*03",
74
+ "IGHVF10-G33*04",
75
+ "IGHVF10-G34*01",
76
+ "IGHVF10-G34*02",
77
+ "IGHVF10-G35*01",
78
+ "IGHVF10-G35*02",
79
+ "IGHVF10-G35*03",
80
+ "IGHVF10-G35*04",
81
+ "IGHVF10-G36*01",
82
+ "IGHVF10-G37*01",
83
+ "IGHVF10-G37*02",
84
+ "IGHVF10-G37*03",
85
+ "IGHVF10-G37*04",
86
+ "IGHVF10-G37*05",
87
+ "IGHVF10-G37*06",
88
+ "IGHVF10-G37*07",
89
+ "IGHVF10-G37*08",
90
+ "IGHVF10-G37*09",
91
+ "IGHVF10-G37*10",
92
+ "IGHVF10-G37*11",
93
+ "IGHVF10-G38*01",
94
+ "IGHVF10-G38*02",
95
+ "IGHVF10-G38*03",
96
+ "IGHVF10-G38*04",
97
+ "IGHVF10-G38*05",
98
+ "IGHVF10-G38*06",
99
+ "IGHVF10-G39*01",
100
+ "IGHVF10-G40*01",
101
+ "IGHVF10-G40*02",
102
+ "IGHVF10-G40*03",
103
+ "IGHVF10-G40*04",
104
+ "IGHVF10-G40*05",
105
+ "IGHVF10-G40*06",
106
+ "IGHVF10-G40*07",
107
+ "IGHVF10-G41*01",
108
+ "IGHVF10-G41*02",
109
+ "IGHVF10-G41*03",
110
+ "IGHVF10-G42*01",
111
+ "IGHVF10-G42*02",
112
+ "IGHVF10-G42*03",
113
+ "IGHVF10-G42*04",
114
+ "IGHVF10-G42*05",
115
+ "IGHVF10-G43*01",
116
+ "IGHVF10-G43*02",
117
+ "IGHVF10-G43*03",
118
+ "IGHVF10-G44*01",
119
+ "IGHVF10-G44*02",
120
+ "IGHVF10-G44*03",
121
+ "IGHVF10-G44*04",
122
+ "IGHVF10-G45*01",
123
+ "IGHVF10-G46*01",
124
+ "IGHVF10-G46*02",
125
+ "IGHVF10-G46*03",
126
+ "IGHVF10-G47*01",
127
+ "IGHVF10-G48*01",
128
+ "IGHVF10-G48*02",
129
+ "IGHVF10-G49*01",
130
+ "IGHVF10-G49*02",
131
+ "IGHVF10-G49*03",
132
+ "IGHVF10-G49*04",
133
+ "IGHVF10-G50*01",
134
+ "IGHVF10-G50*02",
135
+ "IGHVF10-G50*03",
136
+ "IGHVF10-G50*04",
137
+ "IGHVF10-G51*01",
138
+ "IGHVF10-G51*02",
139
+ "IGHVF10-G51*03",
140
+ "IGHVF10-G51*04",
141
+ "IGHVF10-G51*05",
142
+ "IGHVF10-G52*01",
143
+ "IGHVF10-G52*02",
144
+ "IGHVF10-G52*03",
145
+ "IGHVF10-G52*04",
146
+ "IGHVF10-G52*05",
147
+ "IGHVF10-G52*06",
148
+ "IGHVF2-G4*01",
149
+ "IGHVF2-G4*02",
150
+ "IGHVF3-G10*01",
151
+ "IGHVF3-G10*02",
152
+ "IGHVF3-G10*03",
153
+ "IGHVF3-G10*04",
154
+ "IGHVF3-G10*05",
155
+ "IGHVF3-G10*06",
156
+ "IGHVF3-G10*07",
157
+ "IGHVF3-G11*01",
158
+ "IGHVF3-G11*02",
159
+ "IGHVF3-G11*03",
160
+ "IGHVF3-G11*04",
161
+ "IGHVF3-G11*05",
162
+ "IGHVF3-G11*06",
163
+ "IGHVF3-G12*01",
164
+ "IGHVF3-G12*02",
165
+ "IGHVF3-G12*03",
166
+ "IGHVF3-G12*04",
167
+ "IGHVF3-G5*01",
168
+ "IGHVF3-G6*01",
169
+ "IGHVF3-G6*02",
170
+ "IGHVF3-G6*03",
171
+ "IGHVF3-G6*04",
172
+ "IGHVF3-G6*05",
173
+ "IGHVF3-G6*06",
174
+ "IGHVF3-G6*07",
175
+ "IGHVF3-G6*08",
176
+ "IGHVF3-G7*01",
177
+ "IGHVF3-G7*02",
178
+ "IGHVF3-G7*03",
179
+ "IGHVF3-G7*04",
180
+ "IGHVF3-G7*05",
181
+ "IGHVF3-G8*01",
182
+ "IGHVF3-G8*02",
183
+ "IGHVF3-G8*03",
184
+ "IGHVF3-G8*04",
185
+ "IGHVF3-G9*01",
186
+ "IGHVF3-G9*02",
187
+ "IGHVF3-G9*03",
188
+ "IGHVF3-G9*04",
189
+ "IGHVF3-G9*04_A319G",
190
+ "IGHVF4-G13*01",
191
+ "IGHVF4-G13*02",
192
+ "IGHVF4-G14*01",
193
+ "IGHVF4-G14*02",
194
+ "IGHVF4-G14*03",
195
+ "IGHVF4-G14*04",
196
+ "IGHVF5-G15*01",
197
+ "IGHVF5-G16*01",
198
+ "IGHVF5-G16*02",
199
+ "IGHVF5-G16*03",
200
+ "IGHVF5-G16*04",
201
+ "IGHVF6-G17*01",
202
+ "IGHVF6-G18*01",
203
+ "IGHVF6-G19*01",
204
+ "IGHVF6-G19*02",
205
+ "IGHVF6-G19*03",
206
+ "IGHVF6-G20*01",
207
+ "IGHVF6-G20*02",
208
+ "IGHVF6-G21*01",
209
+ "IGHVF6-G21*02",
210
+ "IGHVF6-G21*03",
211
+ "IGHVF6-G21*04",
212
+ "IGHVF6-G21*05",
213
+ "IGHVF6-G21*06",
214
+ "IGHVF6-G21*07",
215
+ "IGHVF6-G21*08",
216
+ "IGHVF6-G21*09",
217
+ "IGHVF6-G21*10",
218
+ "IGHVF6-G21*11",
219
+ "IGHVF6-G21*12",
220
+ "IGHVF6-G21*13",
221
+ "IGHVF6-G21*14",
222
+ "IGHVF6-G21*15",
223
+ "IGHVF6-G21*16",
224
+ "IGHVF6-G22*01",
225
+ "IGHVF6-G22*02",
226
+ "IGHVF6-G22*03",
227
+ "IGHVF6-G23*01",
228
+ "IGHVF6-G23*02",
229
+ "IGHVF6-G23*03",
230
+ "IGHVF6-G24*01",
231
+ "IGHVF6-G24*02",
232
+ "IGHVF6-G24*03",
233
+ "IGHVF6-G24*04",
234
+ "IGHVF6-G24*05",
235
+ "IGHVF6-G25*01",
236
+ "IGHVF6-G25*02",
237
+ "IGHVF6-G25*03",
238
+ "IGHVF6-G26*01",
239
+ "IGHVF6-G26*02",
240
+ "IGHVF6-G26*03",
241
+ "IGHVF6-G26*04",
242
+ "IGHVF7-G27*01",
243
+ "IGHVF7-G27*02",
244
+ "IGHVF7-G27*03",
245
+ "IGHVF7-G27*04",
246
+ "IGHVF8-G28*01",
247
+ "IGHVF8-G29*01",
248
+ "IGHVF8-G29*02",
249
+ "IGHVF8-G30*01",
250
+ "IGHVF8-G30*02",
251
+ "IGHVF8-G30*03",
252
+ "IGHVF8-G31*01",
253
+ "IGHVF8-G31*02",
254
+ "IGHVF9-G32*01",
255
+ "IGHVF9-G32*02",
256
+ "IGHVF9-G32*03",
257
+ "IGHVF9-G32*04"
258
+ ]
259
+ },
260
+ "chain_type": "BCR_HEAVY",
261
+ "has_d": true,
262
+ "last_updated": "2025-02-01",
263
+ "reference_set": "OGRDB V8",
264
+ "species": "Human"
265
+ },
266
+ "extra": null,
267
+ "format_version": 3,
268
+ "framework": "pytorch",
269
+ "framework_version": "pytorch-2.10.0+cu128",
270
+ "genairr_version": "0.6.3",
271
+ "git_commit": "2f2988e",
272
+ "has_d_gene": true,
273
+ "has_orientation_head": true,
274
+ "j_allele_count": 7,
275
+ "j_allele_latent_size": null,
276
+ "max_seq_length": 576,
277
+ "model_type": "single_chain",
278
+ "notes": null,
279
+ "number_of_chains": null,
280
+ "python_version": "3.12.3",
281
+ "v_allele_count": 198,
282
+ "v_allele_latent_size": null
283
+ }
HUMAN_IGH_OGRDB_576/dataconfig.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:faeeaadca3860540baec789ae130f1a6f38a29273867b8a971568937c4fa1b2f
3
+ size 6451007
HUMAN_IGH_OGRDB_576/fingerprint.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ 0c8df7b9790a94a7cf0f7ab9c5a69eaa7b2b39891b1f9d10cc2279b873e054d6
HUMAN_IGH_OGRDB_576/model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d31aab26a45fc26de13118f8e83a5ea5976c134b6a821c5c61b2d55f74405333
3
+ size 10451535
HUMAN_IGH_OGRDB_576/training_meta.json ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "batch_size": 512,
3
+ "best_epoch": 995,
4
+ "best_loss": 1.8584578084945678,
5
+ "created_utc": "2026-03-01T15:40:13.692179+00:00",
6
+ "epochs_trained": 1000,
7
+ "extra": {
8
+ "notes": "v3: DeepSegHead + ConditionedSegHead + JunctionRegionHead + AlignAIRLoss (hybrid CE+Cramer, inv-freq class weights, reading frame loss, Short-D concordance)"
9
+ },
10
+ "final_epoch": 999,
11
+ "final_loss": 1.885474544763565,
12
+ "learning_rate": "0.0",
13
+ "metrics_summary": {
14
+ "classification": 0.1835135055333376,
15
+ "d_allele_avg_gt_count": 1.5322896242141724,
16
+ "d_allele_setK_precision": 0.8797246295778584,
17
+ "d_allele_setK_recall": 0.8797246295778584,
18
+ "d_allele_thresh_precision": 0.843770444393158,
19
+ "d_allele_thresh_recall": 0.8331026434898376,
20
+ "d_allele_top1_acc": 0.8727984428405762,
21
+ "d_allele_top1_conf": 0.8446706533432007,
22
+ "d_end_acc": 0.7514677047729492,
23
+ "d_end_acc_1nt": 0.8414872884750366,
24
+ "d_end_mae": 1.039138913154602,
25
+ "d_start_acc": 0.7573385238647461,
26
+ "d_start_acc_1nt": 0.8277886509895325,
27
+ "d_start_mae": 1.189823865890503,
28
+ "indel_count": 0.1536969581246376,
29
+ "indel_count_mae": 0.2501048743724823,
30
+ "j_allele_avg_gt_count": 1.0,
31
+ "j_allele_setK_precision": 1.0,
32
+ "j_allele_setK_recall": 1.0,
33
+ "j_allele_thresh_precision": 1.0,
34
+ "j_allele_thresh_recall": 1.0,
35
+ "j_allele_top1_acc": 1.0,
36
+ "j_allele_top1_conf": 0.9496095776557922,
37
+ "j_end_acc": 1.0,
38
+ "j_end_acc_1nt": 1.0,
39
+ "j_end_mae": 0.0,
40
+ "j_start_acc": 0.9373776912689209,
41
+ "j_start_acc_1nt": 0.9784735441207886,
42
+ "j_start_mae": 0.11741682887077332,
43
+ "log_var/log_var_d_classification": -0.8107569217681885,
44
+ "log_var/log_var_d_end": 1.2113195657730103,
45
+ "log_var/log_var_d_start": 1.1995137929916382,
46
+ "log_var/log_var_indel": -0.6938347220420837,
47
+ "log_var/log_var_j_classification": -0.9232622981071472,
48
+ "log_var/log_var_j_end": 0.82975834608078,
49
+ "log_var/log_var_j_start": 0.9077933430671692,
50
+ "log_var/log_var_junction_region": -1.6838476657867432,
51
+ "log_var/log_var_mutation": -3.0,
52
+ "log_var/log_var_orientation": -3.0,
53
+ "log_var/log_var_productivity": -3.0,
54
+ "log_var/log_var_v_classification": -0.9191080927848816,
55
+ "log_var/log_var_v_end": 0.9017267227172852,
56
+ "log_var/log_var_v_start": 0.9413631558418274,
57
+ "loss": 1.885474544763565,
58
+ "lr": 0.0,
59
+ "mutation_rate": -1.3141397738456726,
60
+ "mutation_rate_mae": 0.009754992090165615,
61
+ "orientation": -1.5,
62
+ "orientation_acc": 1.0,
63
+ "orientation_acc/complement": 1.0,
64
+ "orientation_acc/normal": 0.9999999403953552,
65
+ "orientation_acc/revcomp": 1.0,
66
+ "orientation_acc/reversed": 1.0,
67
+ "productive": -1.345355775654316,
68
+ "productive_acc": 1.0,
69
+ "raw/d_allele": 0.2223323979973793,
70
+ "raw/d_end": 1.6471605092287063,
71
+ "raw/d_start": 1.630733658671379,
72
+ "raw/frame_loss": 0.12696342319250106,
73
+ "raw/hinge_loss": 0.00035467982070258586,
74
+ "raw/indel_count": 0.25013513743877414,
75
+ "raw/iou_loss": 0.21144418716430663,
76
+ "raw/j_allele": 0.19859787605702878,
77
+ "raw/j_end": 1.1314865028858185,
78
+ "raw/j_start": 1.205588527917862,
79
+ "raw/junction_region": 0.09296022888273,
80
+ "raw/len_loss": 0.5996542032063007,
81
+ "raw/mutation_rate": 0.009253435865975917,
82
+ "raw/ordering_loss": 0.003418139982386492,
83
+ "raw/orientation": 0.0,
84
+ "raw/productive": 0.007699282681714976,
85
+ "raw/regularization": 0.029999999329447746,
86
+ "raw/v_allele": 0.19943240255117417,
87
+ "raw/v_end": 1.2012043470144271,
88
+ "raw/v_start": 1.2461054134368896,
89
+ "segmentation": 5.677759671211243,
90
+ "sigma": 0.75,
91
+ "v_allele_avg_gt_count": 1.281800389289856,
92
+ "v_allele_setK_precision": 0.9921722113502935,
93
+ "v_allele_setK_recall": 0.9921722113502935,
94
+ "v_allele_thresh_precision": 0.9846706390380859,
95
+ "v_allele_thresh_recall": 0.9868884682655334,
96
+ "v_allele_top1_acc": 0.9941291809082031,
97
+ "v_allele_top1_conf": 0.9441247582435608,
98
+ "v_end_acc": 0.951076328754425,
99
+ "v_end_acc_1nt": 0.9882583022117615,
100
+ "v_end_mae": 0.06457925587892532,
101
+ "v_start_acc": 0.9178082346916199,
102
+ "v_start_acc_1nt": 0.9823874831199646,
103
+ "v_start_mae": 0.13111546635627747
104
+ },
105
+ "mixed_precision": true,
106
+ "optimizer_class": "AdamW",
107
+ "samples_per_epoch": 102400,
108
+ "wall_time_seconds": 67419
109
+ }
HUMAN_IGK_OGRDB_576/README.md ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # AlignAIRR Model Bundle
2
+
3
+ Format Version: 3
4
+ Model Type: single_chain
5
+ Max Sequence Length: 576
6
+
7
+ ## Usage
8
+ ```python
9
+ from AlignAIR.Models import SingleChainAlignAIR
10
+ model = SingleChainAlignAIR.from_pretrained('PATH_TO_BUNDLE')
11
+ ```
12
+
13
+ ## Integrity
14
+ SHA256 fingerprint stored in fingerprint.txt. Do not modify bundle contents manually.
HUMAN_IGK_OGRDB_576/VERSION ADDED
@@ -0,0 +1 @@
 
 
1
+ FORMAT_VERSION=3
HUMAN_IGK_OGRDB_576/config.json ADDED
@@ -0,0 +1,218 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignairr_version": "2.0.2",
3
+ "chain_types": null,
4
+ "created_utc": "2026-03-04T13:32:53.815450+00:00",
5
+ "d_allele_count": null,
6
+ "d_allele_latent_size": null,
7
+ "dataconfig_hash": "635c92a75e067e8e32caf3d28079afd822793b257404ffbf1185a8649f2e869f",
8
+ "dataconfig_metadata": {
9
+ "allele_counts": {
10
+ "j": 8,
11
+ "v": 168
12
+ },
13
+ "allele_names": {
14
+ "j": [
15
+ "IGKJ1-5PHE*01",
16
+ "IGKJ2-27FL*01",
17
+ "IGKJ2-27FL*02",
18
+ "IGKJ2-27FL*03",
19
+ "IGKJ2-27FL*04",
20
+ "IGKJ3-DD27*01",
21
+ "IGKJ4-ZOFQ*01",
22
+ "IGKJ5-CLVV*01"
23
+ ],
24
+ "v": [
25
+ "IGKV1-2YQT*01",
26
+ "IGKV1-2YQT*02",
27
+ "IGKV1-2YQT*03",
28
+ "IGKV1-2YQT*04",
29
+ "IGKV1-2YQT*05",
30
+ "IGKV1-2YQT*06",
31
+ "IGKV1-2YQT*07",
32
+ "IGKV1-2YQT*08",
33
+ "IGKV1-2YQT*09",
34
+ "IGKV1-3EZB*01",
35
+ "IGKV1-3EZB*02",
36
+ "IGKV1-3EZB*03",
37
+ "IGKV1-3EZB*04",
38
+ "IGKV1-3EZB*05",
39
+ "IGKV1-5TCS*03",
40
+ "IGKV1-5TCS*04",
41
+ "IGKV1-5TCS*05",
42
+ "IGKV1-5TCS*07",
43
+ "IGKV1-6FMK*01",
44
+ "IGKV1-6FMK*02",
45
+ "IGKV1-6FMK*03",
46
+ "IGKV1-6FMK*04",
47
+ "IGKV1-6FMK*05",
48
+ "IGKV1-6FMK*06",
49
+ "IGKV1-6FMK*07",
50
+ "IGKV1-6FMK*08",
51
+ "IGKV1-6FMK*09",
52
+ "IGKV1-6HUK*01",
53
+ "IGKV1-6HUK*02",
54
+ "IGKV1-6HUK*05",
55
+ "IGKV1-6QNL*01",
56
+ "IGKV1-6QNL*02",
57
+ "IGKV1-6QNL*03",
58
+ "IGKV1-6QNL*04",
59
+ "IGKV1-6QNL*05",
60
+ "IGKV1-6QNL*06",
61
+ "IGKV1-747T*01",
62
+ "IGKV1-747T*02",
63
+ "IGKV1-747T*03",
64
+ "IGKV1-747T*04",
65
+ "IGKV1-747T*05",
66
+ "IGKV1-747T*06",
67
+ "IGKV1-747T*07",
68
+ "IGKV1-747T*08",
69
+ "IGKV1-AZST*01",
70
+ "IGKV1-AZST*02",
71
+ "IGKV1-AZST*03",
72
+ "IGKV1-AZST*04",
73
+ "IGKV1-AZST*05",
74
+ "IGKV1-BJHA*01",
75
+ "IGKV1-BJHA*02",
76
+ "IGKV1-BJHA*03",
77
+ "IGKV1-CEQR*01",
78
+ "IGKV1-CEQR*02",
79
+ "IGKV1-CSWN*01",
80
+ "IGKV1-CSWN*02",
81
+ "IGKV1-CSWN*03",
82
+ "IGKV1-EBUV*01",
83
+ "IGKV1-EBUV*02",
84
+ "IGKV1-EBUV*03",
85
+ "IGKV1-EBUV*04",
86
+ "IGKV1-EBUV*05",
87
+ "IGKV1-F5PO*01",
88
+ "IGKV1-F5PO*03",
89
+ "IGKV1-F5PO*04",
90
+ "IGKV1-F5PO*05",
91
+ "IGKV1-F5PO*07",
92
+ "IGKV1-K4UB*01",
93
+ "IGKV1-K4UB*02",
94
+ "IGKV1-K4UB*03",
95
+ "IGKV1-LLLI*02",
96
+ "IGKV1-PLJP*01",
97
+ "IGKV2-22MS*01",
98
+ "IGKV2-22MS*02",
99
+ "IGKV2-22MS*04",
100
+ "IGKV2-22MS*05",
101
+ "IGKV2-22MS*06",
102
+ "IGKV2-22MS*07",
103
+ "IGKV2-22MS*08",
104
+ "IGKV2-22MS*09",
105
+ "IGKV2-22MS*10",
106
+ "IGKV2-2F4I*01",
107
+ "IGKV2-2F4I*02",
108
+ "IGKV2-2F4I*03",
109
+ "IGKV2-2F4I*04",
110
+ "IGKV2-2F4I*05",
111
+ "IGKV2-2F4I*06",
112
+ "IGKV2-2F4I*07",
113
+ "IGKV2-2NPS*02",
114
+ "IGKV2-2NPS*03",
115
+ "IGKV2-2NPS*04",
116
+ "IGKV2-2NPS*05",
117
+ "IGKV2-2NPS*06",
118
+ "IGKV2-6U5O*01",
119
+ "IGKV2-6U5O*02",
120
+ "IGKV2-AQII*02",
121
+ "IGKV2-AQII*03",
122
+ "IGKV2-AQII*04",
123
+ "IGKV2-AQII*05",
124
+ "IGKV2-AQII*06",
125
+ "IGKV2-AQII*07",
126
+ "IGKV2-DJXV*01",
127
+ "IGKV2-DJXV*02",
128
+ "IGKV2-DJXV*03",
129
+ "IGKV2-DJXV*05",
130
+ "IGKV2-DJXV*06",
131
+ "IGKV2-DJXV*07",
132
+ "IGKV2-DJXV*08",
133
+ "IGKV2-GVWW*01",
134
+ "IGKV2-GVWW*02",
135
+ "IGKV2-GVWW*03",
136
+ "IGKV2-GVWW*04",
137
+ "IGKV2-GVWW*05",
138
+ "IGKV2-GVWW*06",
139
+ "IGKV2-GVWW*07",
140
+ "IGKV2-GVWW*08",
141
+ "IGKV3-3IAK*01",
142
+ "IGKV3-3IAK*02",
143
+ "IGKV3-3IAK*03",
144
+ "IGKV3-3IAK*04",
145
+ "IGKV3-3IAK*05",
146
+ "IGKV3-3IAK*06",
147
+ "IGKV3-3IAK*07",
148
+ "IGKV3-3IAK*08",
149
+ "IGKV3-3IAK*09",
150
+ "IGKV3-3IAK*10",
151
+ "IGKV3-3IAK*11",
152
+ "IGKV3-3IAK*12",
153
+ "IGKV3-3IAK*13",
154
+ "IGKV3-3IAK*14",
155
+ "IGKV3-3IAK*15",
156
+ "IGKV3-3P4Y*01",
157
+ "IGKV3-3P4Y*02",
158
+ "IGKV3-3P4Y*03",
159
+ "IGKV3-3P4Y*04",
160
+ "IGKV3-3P4Y*05",
161
+ "IGKV3-5PXT*01",
162
+ "IGKV3-5PXT*02",
163
+ "IGKV3-5PXT*03",
164
+ "IGKV3-5PXT*04",
165
+ "IGKV3-5PXT*05",
166
+ "IGKV3-5PXT*06",
167
+ "IGKV3-5PXT*07",
168
+ "IGKV3-5PXT*09",
169
+ "IGKV3-CJIB*01",
170
+ "IGKV3-CJIB*02",
171
+ "IGKV3-CJIB*03",
172
+ "IGKV3-CJIB*04",
173
+ "IGKV3-CJIB*05",
174
+ "IGKV3-CJIB*06",
175
+ "IGKV4-7TN7*01",
176
+ "IGKV4-7TN7*02",
177
+ "IGKV4-7TN7*03",
178
+ "IGKV4-7TN7*04",
179
+ "IGKV5-D65I*01",
180
+ "IGKV5-D65I*02",
181
+ "IGKV5-D65I*03",
182
+ "IGKV5-D65I*04",
183
+ "IGKV6-227Y*01",
184
+ "IGKV6-227Y*02",
185
+ "IGKV6-227Y*03",
186
+ "IGKV6-227Y*04",
187
+ "IGKV6-227Y*05",
188
+ "IGKV6-227Y*06",
189
+ "IGKV6-227Y*07",
190
+ "IGKV6-5WQW*01",
191
+ "IGKV6-5WQW*02",
192
+ "IGKV6-5WQW*03"
193
+ ]
194
+ },
195
+ "chain_type": "BCR_LIGHT_KAPPA",
196
+ "has_d": false,
197
+ "last_updated": "2025-07-15",
198
+ "reference_set": "OGRDB V2 & V3 extended",
199
+ "species": "Human"
200
+ },
201
+ "extra": null,
202
+ "format_version": 3,
203
+ "framework": "pytorch",
204
+ "framework_version": "pytorch-2.10.0+cu128",
205
+ "genairr_version": "0.6.3",
206
+ "git_commit": "2f2988e",
207
+ "has_d_gene": false,
208
+ "has_orientation_head": true,
209
+ "j_allele_count": 8,
210
+ "j_allele_latent_size": null,
211
+ "max_seq_length": 576,
212
+ "model_type": "single_chain",
213
+ "notes": null,
214
+ "number_of_chains": null,
215
+ "python_version": "3.12.3",
216
+ "v_allele_count": 168,
217
+ "v_allele_latent_size": null
218
+ }
HUMAN_IGK_OGRDB_576/dataconfig.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d72aec84874d99e547164306543f22cf6e033b7b2a0eb351b2b85c5d5796361c
3
+ size 3680760
HUMAN_IGK_OGRDB_576/fingerprint.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ 43cff24a37fd1f856d0828ae4820db7192c2e0d2efd9faf58c258a5144803e74
HUMAN_IGK_OGRDB_576/model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:431f33b0bb8175f393bb50f69b7c18c26a2a3b4ffff50b0fa814ffbc05d60e5f
3
+ size 8670647
HUMAN_IGK_OGRDB_576/training_meta.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "batch_size": 512,
3
+ "best_epoch": 992,
4
+ "best_loss": -1.4399833059310914,
5
+ "created_utc": "2026-03-04T13:32:53.816574+00:00",
6
+ "epochs_trained": 1000,
7
+ "extra": {
8
+ "notes": "IGK HUMAN_IGK_OGRDB (V=168, D=0, J=8), light chain (no D segment), CrossAttentionPooling, Uniform mutation (0.01-0.05), Kendall uncertainty weighting, orientation head, sigma schedule cosine 3.0\u21920.75, 6 workers"
9
+ },
10
+ "final_epoch": 999,
11
+ "final_loss": -1.4222590343654156,
12
+ "learning_rate": "1.4999999999999999e-05",
13
+ "metrics_summary": {
14
+ "classification": 0.10841067031025886,
15
+ "indel_count": 0.16907258480787277,
16
+ "indel_count_mae": 0.26504307985305786,
17
+ "j_allele_avg_gt_count": 1.0,
18
+ "j_allele_setK_precision": 0.978515625,
19
+ "j_allele_setK_recall": 0.978515625,
20
+ "j_allele_thresh_precision": 0.97265625,
21
+ "j_allele_thresh_recall": 0.97265625,
22
+ "j_allele_top1_acc": 0.978515625,
23
+ "j_allele_top1_conf": 0.9289813041687012,
24
+ "j_end_acc": 1.0,
25
+ "j_end_acc_1nt": 1.0,
26
+ "j_end_mae": 0.0,
27
+ "j_start_acc": 0.9453125,
28
+ "j_start_acc_1nt": 0.982421875,
29
+ "j_start_mae": 0.080078125,
30
+ "log_var/log_var_indel": -0.6573320901393891,
31
+ "log_var/log_var_j_classification": -0.873119498193264,
32
+ "log_var/log_var_j_end": 0.8165233558416367,
33
+ "log_var/log_var_j_start": 0.8649461326003075,
34
+ "log_var/log_var_junction_region": -3.0,
35
+ "log_var/log_var_mutation": -3.0,
36
+ "log_var/log_var_orientation": -3.0,
37
+ "log_var/log_var_productivity": -3.0,
38
+ "log_var/log_var_v_classification": -0.9093747818470002,
39
+ "log_var/log_var_v_end": 0.925546005666256,
40
+ "log_var/log_var_v_start": 0.9423904371261597,
41
+ "loss": -1.4222590343654156,
42
+ "lr": 1.4999999999999999e-05,
43
+ "mutation_rate": -1.4390911811590195,
44
+ "mutation_rate_mae": 0.003021292155608535,
45
+ "orientation": -1.5,
46
+ "orientation_acc": 1.0,
47
+ "orientation_acc/complement": 1.0,
48
+ "orientation_acc/normal": 1.0,
49
+ "orientation_acc/revcomp": 1.0,
50
+ "orientation_acc/reversed": 1.0,
51
+ "productive": -1.3008609154820443,
52
+ "productive_acc": 1.0,
53
+ "raw/frame_loss": 0.1685053576156497,
54
+ "raw/hinge_loss": 0.0003186190128326416,
55
+ "raw/indel_count": 0.2579441211372614,
56
+ "raw/iou_loss": 0.007318592965602875,
57
+ "raw/j_allele": 0.20862680852413176,
58
+ "raw/j_end": 1.1313094753026962,
59
+ "raw/j_start": 1.1874751782417297,
60
+ "raw/junction_region": 0.009267570574302227,
61
+ "raw/len_loss": 0.3159478325396776,
62
+ "raw/mutation_rate": 0.0030324714817106723,
63
+ "raw/ordering_loss": 0.001164478063583374,
64
+ "raw/orientation": 0.0,
65
+ "raw/productive": 0.009914551141846459,
66
+ "raw/regularization": 0.03999999910593033,
67
+ "raw/v_allele": 0.2014396557956934,
68
+ "raw/v_end": 1.2592456096410751,
69
+ "raw/v_start": 1.281528382897377,
70
+ "segmentation": 2.5002098369598387,
71
+ "sigma": 0.75,
72
+ "v_allele_avg_gt_count": 1.73828125,
73
+ "v_allele_setK_precision": 0.9382440476190477,
74
+ "v_allele_setK_recall": 0.9382440476190477,
75
+ "v_allele_thresh_precision": 0.9306513667106628,
76
+ "v_allele_thresh_recall": 0.9378255605697632,
77
+ "v_allele_top1_acc": 0.939453125,
78
+ "v_allele_top1_conf": 0.8940415382385254,
79
+ "v_end_acc": 0.91015625,
80
+ "v_end_acc_1nt": 0.955078125,
81
+ "v_end_mae": 0.15625,
82
+ "v_start_acc": 0.908203125,
83
+ "v_start_acc_1nt": 0.953125,
84
+ "v_start_mae": 0.2578125
85
+ },
86
+ "mixed_precision": true,
87
+ "optimizer_class": "AdamW",
88
+ "samples_per_epoch": 102400,
89
+ "wall_time_seconds": 54802
90
+ }
HUMAN_IGL_OGRDB_576/README.md ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # AlignAIRR Model Bundle
2
+
3
+ Format Version: 3
4
+ Model Type: single_chain
5
+ Max Sequence Length: 576
6
+
7
+ ## Usage
8
+ ```python
9
+ from AlignAIR.Models import SingleChainAlignAIR
10
+ model = SingleChainAlignAIR.from_pretrained('PATH_TO_BUNDLE')
11
+ ```
12
+
13
+ ## Integrity
14
+ SHA256 fingerprint stored in fingerprint.txt. Do not modify bundle contents manually.
HUMAN_IGL_OGRDB_576/VERSION ADDED
@@ -0,0 +1 @@
 
 
1
+ FORMAT_VERSION=3
HUMAN_IGL_OGRDB_576/config.json ADDED
@@ -0,0 +1,233 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignairr_version": "2.0.2",
3
+ "chain_types": null,
4
+ "created_utc": "2026-03-05T07:23:10.005123+00:00",
5
+ "d_allele_count": null,
6
+ "d_allele_latent_size": null,
7
+ "dataconfig_hash": "4f519381775241d12127b33bcb204af05294e206e444d9b274b09e0e5d973078",
8
+ "dataconfig_metadata": {
9
+ "allele_counts": {
10
+ "j": 10,
11
+ "v": 181
12
+ },
13
+ "allele_names": {
14
+ "j": [
15
+ "IGLJ1-VFD4*01",
16
+ "IGLJ2/3-HDWH*01",
17
+ "IGLJ2/3-JWS7*02",
18
+ "IGLJ3-JWS7*01",
19
+ "IGLJ4-4HLU*01",
20
+ "IGLJ5-SBBZ*01",
21
+ "IGLJ5-SBBZ*02",
22
+ "IGLJ6-QPXP*01",
23
+ "IGLJ7-GIJA*01",
24
+ "IGLJ7-GIJA*03"
25
+ ],
26
+ "v": [
27
+ "IGLV1-2CIH*01",
28
+ "IGLV1-2CIH*02",
29
+ "IGLV1-2CIH*04",
30
+ "IGLV1-2CIH*05",
31
+ "IGLV1-2FD4*01",
32
+ "IGLV1-2FD4*05",
33
+ "IGLV1-53QP*01",
34
+ "IGLV1-53QP*02",
35
+ "IGLV1-53QP*03",
36
+ "IGLV1-53QP*04",
37
+ "IGLV1-53QP*05",
38
+ "IGLV1-6ZIO*01",
39
+ "IGLV1-6ZIO*02",
40
+ "IGLV1-6ZIO*03",
41
+ "IGLV1-6ZIO*04",
42
+ "IGLV1-6ZIO*05",
43
+ "IGLV1-6ZIO*06",
44
+ "IGLV1-6ZIO*07",
45
+ "IGLV1-6ZIO*08",
46
+ "IGLV1-6ZIO*09",
47
+ "IGLV1-6ZIO*10",
48
+ "IGLV1-6ZIO*11",
49
+ "IGLV1-6ZIO*12",
50
+ "IGLV1-6ZIO*13",
51
+ "IGLV1-6ZIO*14",
52
+ "IGLV1-ARVS*01",
53
+ "IGLV1-ARVS*02",
54
+ "IGLV1-ARVS*03",
55
+ "IGLV1-ARVS*04",
56
+ "IGLV1-ARVS*05",
57
+ "IGLV1-ARVS*06",
58
+ "IGLV1-CAX2*01",
59
+ "IGLV1-CAX2*02",
60
+ "IGLV1-CAX2*03",
61
+ "IGLV1-CAX2*04",
62
+ "IGLV1-CAX2*05",
63
+ "IGLV1-CAX2*06",
64
+ "IGLV1-CAX2*07",
65
+ "IGLV10-32HX*01",
66
+ "IGLV10-32HX*02",
67
+ "IGLV10-32HX*03",
68
+ "IGLV10-32HX*04",
69
+ "IGLV10-32HX*05",
70
+ "IGLV10-32HX*06",
71
+ "IGLV10-32HX*07",
72
+ "IGLV10-32HX*09",
73
+ "IGLV10-32HX*10",
74
+ "IGLV11-B3GZ*03",
75
+ "IGLV11-B3GZ*07",
76
+ "IGLV2-2CD2*01",
77
+ "IGLV2-2CD2*02",
78
+ "IGLV2-2CD2*03",
79
+ "IGLV2-2CD2*04",
80
+ "IGLV2-2CD2*05",
81
+ "IGLV2-2CD2*06",
82
+ "IGLV2-2CD2*07",
83
+ "IGLV2-2CD2*08",
84
+ "IGLV2-2CD2*09",
85
+ "IGLV2-35RK*01",
86
+ "IGLV2-35RK*04",
87
+ "IGLV2-3CMY*01",
88
+ "IGLV2-3CMY*02",
89
+ "IGLV2-3CMY*03",
90
+ "IGLV2-3CMY*04",
91
+ "IGLV2-3CMY*05",
92
+ "IGLV2-3CMY*06",
93
+ "IGLV2-3CMY*07",
94
+ "IGLV2-3CMY*08",
95
+ "IGLV2-3CMY*09",
96
+ "IGLV2-3CMY*10",
97
+ "IGLV2-3CMY*11",
98
+ "IGLV2-3CMY*12",
99
+ "IGLV2-4IUE*01",
100
+ "IGLV2-4IUE*02",
101
+ "IGLV2-4IUE*03",
102
+ "IGLV2-4IUE*04",
103
+ "IGLV2-4IUE*05",
104
+ "IGLV2-4IUE*06",
105
+ "IGLV2-4IUE*07",
106
+ "IGLV2-4OG4*01",
107
+ "IGLV2-4OG4*03",
108
+ "IGLV2-4OG4*04",
109
+ "IGLV2-4OG4*06",
110
+ "IGLV2-4OG4*07",
111
+ "IGLV2-6Y26*01",
112
+ "IGLV2-6Y26*02",
113
+ "IGLV2-6Y26*03",
114
+ "IGLV3-2W4X*02",
115
+ "IGLV3-2W4X*03",
116
+ "IGLV3-2W4X*04",
117
+ "IGLV3-2WXG*01",
118
+ "IGLV3-2WXG*02",
119
+ "IGLV3-2WXG*03",
120
+ "IGLV3-2WXG*04",
121
+ "IGLV3-2WXG*05",
122
+ "IGLV3-2WXG*06",
123
+ "IGLV3-2WXG*07",
124
+ "IGLV3-3UTT*01",
125
+ "IGLV3-3UTT*02",
126
+ "IGLV3-3VOM*01",
127
+ "IGLV3-3VOM*02",
128
+ "IGLV3-3VOM*03",
129
+ "IGLV3-3VOM*06",
130
+ "IGLV3-3VOM*07",
131
+ "IGLV3-3ZIC*01",
132
+ "IGLV3-3ZIC*02",
133
+ "IGLV3-3ZIC*03",
134
+ "IGLV3-3ZIC*04",
135
+ "IGLV3-6ROY*01",
136
+ "IGLV3-6ROY*02",
137
+ "IGLV3-6ROY*03",
138
+ "IGLV3-6ROY*04",
139
+ "IGLV3-BXUO*01",
140
+ "IGLV3-BXUO*02",
141
+ "IGLV3-BXUO*03",
142
+ "IGLV3-BXUO*04",
143
+ "IGLV3-CKAL*01",
144
+ "IGLV3-CKAL*02",
145
+ "IGLV3-CKAL*03",
146
+ "IGLV3-CKAL*04",
147
+ "IGLV3-CKAL*05",
148
+ "IGLV3-CKAL*06",
149
+ "IGLV3-ITC7*01",
150
+ "IGLV3-J2PC*01",
151
+ "IGLV3-J2PC*02",
152
+ "IGLV3-J2PC*03",
153
+ "IGLV3-TIMK*01",
154
+ "IGLV3-TIMK*02",
155
+ "IGLV4-3HWN*01",
156
+ "IGLV4-3HWN*02",
157
+ "IGLV4-3HWN*04",
158
+ "IGLV4-3HWN*05",
159
+ "IGLV4-3HWN*06",
160
+ "IGLV4-45SV*01",
161
+ "IGLV4-45SV*02",
162
+ "IGLV4-45SV*03",
163
+ "IGLV4-7C4U*01",
164
+ "IGLV4-7C4U*02",
165
+ "IGLV4-7C4U*03",
166
+ "IGLV5-24IV*01",
167
+ "IGLV5-24IV*02",
168
+ "IGLV5-24IV*03",
169
+ "IGLV5-24IV*04",
170
+ "IGLV5-24IV*05",
171
+ "IGLV5-24IV*06",
172
+ "IGLV5-24IV*07",
173
+ "IGLV5-24IV*08",
174
+ "IGLV5-6HXL*01",
175
+ "IGLV5-6HXL*02",
176
+ "IGLV5-7RPJ*01",
177
+ "IGLV5-7RPJ*02",
178
+ "IGLV5-7RPJ*03",
179
+ "IGLV5-7RPJ*04",
180
+ "IGLV5-7RPJ*05",
181
+ "IGLV5-BWX3*01",
182
+ "IGLV5-FX2U*01",
183
+ "IGLV5-L7W6*01",
184
+ "IGLV5-L7W6*02",
185
+ "IGLV5-L7W6*03",
186
+ "IGLV6-A5V2*01",
187
+ "IGLV6-A5V2*02",
188
+ "IGLV6-A5V2*03",
189
+ "IGLV6-A5V2*04",
190
+ "IGLV6-A5V2*05",
191
+ "IGLV6-A5V2*06",
192
+ "IGLV7-5ZGD*01",
193
+ "IGLV7-5ZGD*02",
194
+ "IGLV7-5ZGD*03",
195
+ "IGLV7-5ZGD*04",
196
+ "IGLV7-5ZGD*05",
197
+ "IGLV7-76FE*01",
198
+ "IGLV7-76FE*02",
199
+ "IGLV7-76FE*03",
200
+ "IGLV8-DFVG*01",
201
+ "IGLV8-DFVG*02",
202
+ "IGLV8-DFVG*03",
203
+ "IGLV8-VQDK*01",
204
+ "IGLV9-3GPH*01",
205
+ "IGLV9-3GPH*02",
206
+ "IGLV9-3GPH*03",
207
+ "IGLV9-3GPH*04"
208
+ ]
209
+ },
210
+ "chain_type": "BCR_LIGHT_LAMBDA",
211
+ "has_d": false,
212
+ "last_updated": "2025-07-15",
213
+ "reference_set": "OGRDB V2 & V3 extended",
214
+ "species": "Human"
215
+ },
216
+ "extra": null,
217
+ "format_version": 3,
218
+ "framework": "pytorch",
219
+ "framework_version": "pytorch-2.10.0+cu128",
220
+ "genairr_version": "0.6.3",
221
+ "git_commit": "2f2988e",
222
+ "has_d_gene": false,
223
+ "has_orientation_head": true,
224
+ "j_allele_count": 10,
225
+ "j_allele_latent_size": null,
226
+ "max_seq_length": 576,
227
+ "model_type": "single_chain",
228
+ "notes": null,
229
+ "number_of_chains": null,
230
+ "python_version": "3.12.3",
231
+ "v_allele_count": 181,
232
+ "v_allele_latent_size": null
233
+ }
HUMAN_IGL_OGRDB_576/dataconfig.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f63852e7f327e7013c8387cec8d156caf565e612f7dbf3ea4b090671a27deea1
3
+ size 4307184
HUMAN_IGL_OGRDB_576/fingerprint.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ ce7e20e699f7bf4a5aff95accca609fe22ad01b1e0c8812e336441e64c9e5da4
HUMAN_IGL_OGRDB_576/model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a15fb7e1db01b37f4a6da90bd5516113356891265aeb7c7fb613887e1a697876
3
+ size 8834807
HUMAN_IGL_OGRDB_576/training_meta.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "batch_size": 512,
3
+ "best_epoch": 456,
4
+ "best_loss": -0.13777857705950736,
5
+ "created_utc": "2026-03-05T07:23:10.006254+00:00",
6
+ "epochs_trained": 458,
7
+ "extra": {
8
+ "notes": "IGL HUMAN_IGL_OGRDB (V=181, D=0, J=10), light chain (no D segment), CrossAttentionPooling, Uniform mutation (0.01-0.05), Kendall uncertainty weighting, orientation head, sigma schedule cosine 3.0\u21920.75, 6 workers"
9
+ },
10
+ "final_epoch": 457,
11
+ "final_loss": -0.13029090009629726,
12
+ "learning_rate": "0.00017033236044328875",
13
+ "metrics_summary": {
14
+ "classification": 0.09708564773201943,
15
+ "indel_count": 0.2226593554019928,
16
+ "indel_count_mae": 0.30126291513442993,
17
+ "j_allele_avg_gt_count": 1.0,
18
+ "j_allele_setK_precision": 0.978515625,
19
+ "j_allele_setK_recall": 0.978515625,
20
+ "j_allele_thresh_precision": 0.974609375,
21
+ "j_allele_thresh_recall": 0.978515625,
22
+ "j_allele_top1_acc": 0.978515625,
23
+ "j_allele_top1_conf": 0.9293403625488281,
24
+ "j_end_acc": 0.998046875,
25
+ "j_end_acc_1nt": 1.0,
26
+ "j_end_mae": 0.001953125,
27
+ "j_start_acc": 0.93359375,
28
+ "j_start_acc_1nt": 0.978515625,
29
+ "j_start_mae": 0.11328125,
30
+ "log_var/log_var_indel": -0.5561221012473107,
31
+ "log_var/log_var_j_classification": -0.8964999657869339,
32
+ "log_var/log_var_j_end": 1.4468038487434387,
33
+ "log_var/log_var_j_start": 1.4558982169628143,
34
+ "log_var/log_var_junction_region": -3.0,
35
+ "log_var/log_var_mutation": -3.0,
36
+ "log_var/log_var_orientation": -3.0,
37
+ "log_var/log_var_productivity": -3.0,
38
+ "log_var/log_var_v_classification": -0.909313616156578,
39
+ "log_var/log_var_v_end": 1.468509510755539,
40
+ "log_var/log_var_v_start": 1.4938469630479814,
41
+ "loss": -0.13029090009629726,
42
+ "lr": 0.00017049625113573977,
43
+ "mutation_rate": -1.4223034703731536,
44
+ "mutation_rate_mae": 0.003901875577867031,
45
+ "orientation": -1.4999748712778092,
46
+ "orientation_acc": 1.0,
47
+ "orientation_acc/complement": 1.0,
48
+ "orientation_acc/normal": 1.0,
49
+ "orientation_acc/revcomp": 0.9999999403953552,
50
+ "orientation_acc/reversed": 1.0,
51
+ "productive": -1.26833356320858,
52
+ "productive_acc": 0.994140625,
53
+ "raw/frame_loss": 0.1913288650661707,
54
+ "raw/hinge_loss": 0.00015748143196105956,
55
+ "raw/indel_count": 0.28712632328271864,
56
+ "raw/iou_loss": 0.009658987820148467,
57
+ "raw/j_allele": 0.20400346301496028,
58
+ "raw/j_end": 2.124738314151764,
59
+ "raw/j_start": 2.144471411705017,
60
+ "raw/junction_region": 0.010529722792562098,
61
+ "raw/len_loss": 0.46119194373488426,
62
+ "raw/mutation_rate": 0.0038682823721319435,
63
+ "raw/ordering_loss": 0.0009547075629234313,
64
+ "raw/orientation": 1.2510997341794173e-06,
65
+ "raw/productive": 0.011533992557087914,
66
+ "raw/regularization": 0.03999999910593033,
67
+ "raw/v_allele": 0.20139127671718599,
68
+ "raw/v_end": 2.1715712904930116,
69
+ "raw/v_start": 2.225696895122528,
70
+ "segmentation": 3.700576022863388,
71
+ "sigma": 2.0249105240388445,
72
+ "v_allele_avg_gt_count": 1.54296875,
73
+ "v_allele_setK_precision": 0.9485677083333333,
74
+ "v_allele_setK_recall": 0.9485677083333333,
75
+ "v_allele_thresh_precision": 0.92431640625,
76
+ "v_allele_thresh_recall": 0.9188213348388672,
77
+ "v_allele_top1_acc": 0.947265625,
78
+ "v_allele_top1_conf": 0.8756070137023926,
79
+ "v_end_acc": 0.791015625,
80
+ "v_end_acc_1nt": 0.93359375,
81
+ "v_end_mae": 0.322265625,
82
+ "v_start_acc": 0.7890625,
83
+ "v_start_acc_1nt": 0.9375,
84
+ "v_start_mae": 0.470703125
85
+ },
86
+ "mixed_precision": true,
87
+ "optimizer_class": "AdamW",
88
+ "samples_per_epoch": 102400,
89
+ "wall_time_seconds": 28409
90
+ }
HUMAN_TCRB_IMGT_576/README.md ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # AlignAIRR Model Bundle
2
+
3
+ Format Version: 3
4
+ Model Type: single_chain
5
+ Max Sequence Length: 576
6
+
7
+ ## Usage
8
+ ```python
9
+ from AlignAIR.Models import SingleChainAlignAIR
10
+ model = SingleChainAlignAIR.from_pretrained('PATH_TO_BUNDLE')
11
+ ```
12
+
13
+ ## Integrity
14
+ SHA256 fingerprint stored in fingerprint.txt. Do not modify bundle contents manually.
HUMAN_TCRB_IMGT_576/VERSION ADDED
@@ -0,0 +1 @@
 
 
1
+ FORMAT_VERSION=3
HUMAN_TCRB_IMGT_576/config.json ADDED
@@ -0,0 +1,192 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alignairr_version": "2.0.2",
3
+ "chain_types": null,
4
+ "created_utc": "2026-03-06T09:19:44.000587+00:00",
5
+ "d_allele_count": 4,
6
+ "d_allele_latent_size": null,
7
+ "dataconfig_hash": "adabac1ea19b736dc1f7e01cee18ba3581ed7d5549d3aafb4824337b018c1b2a",
8
+ "dataconfig_metadata": {
9
+ "allele_counts": {
10
+ "d": 3,
11
+ "j": 14,
12
+ "v": 130
13
+ },
14
+ "allele_names": {
15
+ "d": [
16
+ "TRBD1*01",
17
+ "TRBD2*01",
18
+ "TRBD2*02"
19
+ ],
20
+ "j": [
21
+ "TRBJ1-1*01",
22
+ "TRBJ1-2*01",
23
+ "TRBJ1-3*01",
24
+ "TRBJ1-4*01",
25
+ "TRBJ1-5*01",
26
+ "TRBJ1-6*01",
27
+ "TRBJ1-6*02",
28
+ "TRBJ2-1*01",
29
+ "TRBJ2-2*01",
30
+ "TRBJ2-3*01",
31
+ "TRBJ2-4*01",
32
+ "TRBJ2-5*01",
33
+ "TRBJ2-6*01",
34
+ "TRBJ2-7*01"
35
+ ],
36
+ "v": [
37
+ "TRBV1*01",
38
+ "TRBV10-1*01",
39
+ "TRBV10-1*02",
40
+ "TRBV10-2*01",
41
+ "TRBV10-2*02",
42
+ "TRBV10-3*01",
43
+ "TRBV10-3*02",
44
+ "TRBV10-3*03",
45
+ "TRBV10-3*04",
46
+ "TRBV11-1*01",
47
+ "TRBV11-2*01",
48
+ "TRBV11-2*02",
49
+ "TRBV11-2*03",
50
+ "TRBV11-3*01",
51
+ "TRBV11-3*02",
52
+ "TRBV11-3*03",
53
+ "TRBV11-3*04",
54
+ "TRBV12-1*01",
55
+ "TRBV12-2*01",
56
+ "TRBV12-3*01",
57
+ "TRBV12-4*01",
58
+ "TRBV12-4*02",
59
+ "TRBV12-5*01",
60
+ "TRBV13*01",
61
+ "TRBV13*02",
62
+ "TRBV14*01",
63
+ "TRBV14*02",
64
+ "TRBV15*01",
65
+ "TRBV15*02",
66
+ "TRBV15*03",
67
+ "TRBV16*01",
68
+ "TRBV16*02",
69
+ "TRBV16*03",
70
+ "TRBV18*01",
71
+ "TRBV19*01",
72
+ "TRBV19*02",
73
+ "TRBV19*03",
74
+ "TRBV2*01",
75
+ "TRBV2*02",
76
+ "TRBV2*03",
77
+ "TRBV20-1*01",
78
+ "TRBV20-1*02",
79
+ "TRBV20-1*03",
80
+ "TRBV20-1*04",
81
+ "TRBV20-1*05",
82
+ "TRBV20-1*06",
83
+ "TRBV20-1*07",
84
+ "TRBV21-1*01",
85
+ "TRBV21-1*02",
86
+ "TRBV23-1*01",
87
+ "TRBV24-1*01",
88
+ "TRBV24-1*02",
89
+ "TRBV25-1*01",
90
+ "TRBV27*01",
91
+ "TRBV28*01",
92
+ "TRBV29-1*01",
93
+ "TRBV29-1*02",
94
+ "TRBV29-1*03",
95
+ "TRBV3-1*01",
96
+ "TRBV3-1*02",
97
+ "TRBV3-2*01",
98
+ "TRBV3-2*02",
99
+ "TRBV3-2*03",
100
+ "TRBV30*01",
101
+ "TRBV30*02",
102
+ "TRBV30*03",
103
+ "TRBV30*04",
104
+ "TRBV30*05",
105
+ "TRBV4-1*01",
106
+ "TRBV4-1*02",
107
+ "TRBV4-2*01",
108
+ "TRBV4-2*02",
109
+ "TRBV4-3*01",
110
+ "TRBV4-3*02",
111
+ "TRBV4-3*03",
112
+ "TRBV4-3*04",
113
+ "TRBV5-1*01",
114
+ "TRBV5-1*02",
115
+ "TRBV5-3*01",
116
+ "TRBV5-3*02",
117
+ "TRBV5-4*01",
118
+ "TRBV5-4*02",
119
+ "TRBV5-4*03",
120
+ "TRBV5-4*04",
121
+ "TRBV5-5*01",
122
+ "TRBV5-5*02",
123
+ "TRBV5-5*03",
124
+ "TRBV5-6*01",
125
+ "TRBV5-7*01",
126
+ "TRBV5-8*01",
127
+ "TRBV5-8*02",
128
+ "TRBV6-1*01",
129
+ "TRBV6-23*01",
130
+ "TRBV6-4*01",
131
+ "TRBV6-4*02",
132
+ "TRBV6-5*01",
133
+ "TRBV6-6*01",
134
+ "TRBV6-6*02",
135
+ "TRBV6-6*03",
136
+ "TRBV6-6*04",
137
+ "TRBV6-6*05",
138
+ "TRBV6-7*01",
139
+ "TRBV6-8*01",
140
+ "TRBV6-9*01",
141
+ "TRBV7-1*01",
142
+ "TRBV7-2*01",
143
+ "TRBV7-2*02",
144
+ "TRBV7-2*03",
145
+ "TRBV7-2*04",
146
+ "TRBV7-3*01",
147
+ "TRBV7-3*04",
148
+ "TRBV7-3*05",
149
+ "TRBV7-4*01",
150
+ "TRBV7-6*01",
151
+ "TRBV7-6*02",
152
+ "TRBV7-7*01",
153
+ "TRBV7-7*02",
154
+ "TRBV7-8*01",
155
+ "TRBV7-8*02",
156
+ "TRBV7-8*03",
157
+ "TRBV7-9*01",
158
+ "TRBV7-9*02",
159
+ "TRBV7-9*03",
160
+ "TRBV7-9*04",
161
+ "TRBV7-9*05",
162
+ "TRBV7-9*06",
163
+ "TRBV7-9*07",
164
+ "TRBV9*01",
165
+ "TRBV9*02",
166
+ "TRBV9*03"
167
+ ]
168
+ },
169
+ "chain_type": "TCR_BETA",
170
+ "has_d": true,
171
+ "last_updated": "2025-03-01",
172
+ "reference_set": "IMGT 2022",
173
+ "species": "Human"
174
+ },
175
+ "extra": null,
176
+ "format_version": 3,
177
+ "framework": "pytorch",
178
+ "framework_version": "pytorch-2.10.0+cu128",
179
+ "genairr_version": "0.6.3",
180
+ "git_commit": "2f2988e",
181
+ "has_d_gene": true,
182
+ "has_orientation_head": true,
183
+ "j_allele_count": 14,
184
+ "j_allele_latent_size": null,
185
+ "max_seq_length": 576,
186
+ "model_type": "single_chain",
187
+ "notes": null,
188
+ "number_of_chains": null,
189
+ "python_version": "3.12.3",
190
+ "v_allele_count": 130,
191
+ "v_allele_latent_size": null
192
+ }
HUMAN_TCRB_IMGT_576/dataconfig.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b515060219674840319fae5562c9ec7cd3098102d63b0d1bf4ab73146f405da5
3
+ size 2161540
HUMAN_TCRB_IMGT_576/fingerprint.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ 2f1e07d3515b6475a122654e1738402f8527f7d513098f506b666f3a0aeff7c7
HUMAN_TCRB_IMGT_576/model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3391ffe358452e925a0ed2fc52b20bea86a81f6c41266b4c143f02dc650f83a2
3
+ size 9705679
HUMAN_TCRB_IMGT_576/training_meta.json ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "batch_size": 512,
3
+ "best_epoch": 989,
4
+ "best_loss": 2.4513723576068878,
5
+ "created_utc": "2026-03-06T09:19:44.001784+00:00",
6
+ "epochs_trained": 1000,
7
+ "extra": {
8
+ "notes": "TCRB HUMAN_TCRB_IMGT (V=130, D=3, J=14), TCR beta chain, CrossAttentionPooling, Uniform mutation (0.01-0.05), Kendall uncertainty weighting, orientation head, sigma schedule cosine 3.0\u21920.75, 6 workers"
9
+ },
10
+ "final_epoch": 999,
11
+ "final_loss": 2.4653986859321595,
12
+ "learning_rate": "1.4999999999999999e-05",
13
+ "metrics_summary": {
14
+ "classification": 0.45702348902821543,
15
+ "d_allele_avg_gt_count": 1.091796875,
16
+ "d_allele_setK_precision": 0.908203125,
17
+ "d_allele_setK_recall": 0.908203125,
18
+ "d_allele_thresh_precision": 0.8984375,
19
+ "d_allele_thresh_recall": 0.8981119394302368,
20
+ "d_allele_top1_acc": 0.91015625,
21
+ "d_allele_top1_conf": 0.8624882698059082,
22
+ "d_end_acc": 0.796875,
23
+ "d_end_acc_1nt": 0.87890625,
24
+ "d_end_mae": 0.548828125,
25
+ "d_start_acc": 0.78125,
26
+ "d_start_acc_1nt": 0.87890625,
27
+ "d_start_mae": 0.564453125,
28
+ "indel_count": 0.2868698977679014,
29
+ "indel_count_mae": 0.31625959277153015,
30
+ "j_allele_avg_gt_count": 1.0,
31
+ "j_allele_setK_precision": 0.99609375,
32
+ "j_allele_setK_recall": 0.99609375,
33
+ "j_allele_thresh_precision": 0.99609375,
34
+ "j_allele_thresh_recall": 0.99609375,
35
+ "j_allele_top1_acc": 0.99609375,
36
+ "j_allele_top1_conf": 0.9468851089477539,
37
+ "j_end_acc": 1.0,
38
+ "j_end_acc_1nt": 1.0,
39
+ "j_end_mae": 0.0,
40
+ "j_start_acc": 0.888671875,
41
+ "j_start_acc_1nt": 0.9296875,
42
+ "j_start_mae": 0.275390625,
43
+ "log_var/log_var_d_classification": -0.48912922754883764,
44
+ "log_var/log_var_d_end": 1.0789354532957076,
45
+ "log_var/log_var_d_start": 1.1259199941158295,
46
+ "log_var/log_var_indel": -0.42073034107685087,
47
+ "log_var/log_var_j_classification": -0.9194741976261139,
48
+ "log_var/log_var_j_end": 0.8163983130455017,
49
+ "log_var/log_var_j_start": 0.9725048473477363,
50
+ "log_var/log_var_junction_region": -1.5473183602094651,
51
+ "log_var/log_var_mutation": -3.0,
52
+ "log_var/log_var_orientation": -3.0,
53
+ "log_var/log_var_productivity": -3.0,
54
+ "log_var/log_var_v_classification": -0.9029719170928001,
55
+ "log_var/log_var_v_end": 1.012418593764305,
56
+ "log_var/log_var_v_start": 1.00545921087265,
57
+ "loss": 2.4653986859321595,
58
+ "lr": 1.4999999999999999e-05,
59
+ "mutation_rate": -1.4241796892881393,
60
+ "mutation_rate_mae": 0.0037339571863412857,
61
+ "orientation": -1.4999981981515884,
62
+ "orientation_acc": 1.0,
63
+ "orientation_acc/complement": 1.0,
64
+ "orientation_acc/normal": 1.0,
65
+ "orientation_acc/revcomp": 0.9999999403953552,
66
+ "orientation_acc/reversed": 1.0,
67
+ "productive": -1.258803428709507,
68
+ "productive_acc": 0.998046875,
69
+ "raw/d_allele": 0.3075159652531147,
70
+ "raw/d_end": 1.4717569786310196,
71
+ "raw/d_start": 1.5389642435312272,
72
+ "raw/frame_loss": 0.3637680508941412,
73
+ "raw/hinge_loss": 0.0016483981907367707,
74
+ "raw/indel_count": 0.3264681005477905,
75
+ "raw/iou_loss": 0.17637170106172562,
76
+ "raw/j_allele": 0.19935014188289643,
77
+ "raw/j_end": 1.1311669862270355,
78
+ "raw/j_start": 1.3211141312122345,
79
+ "raw/junction_region": 0.10658789347857237,
80
+ "raw/len_loss": 1.078130791783333,
81
+ "raw/mutation_rate": 0.0037748710624873638,
82
+ "raw/ordering_loss": 0.0017432171106338502,
83
+ "raw/orientation": 8.977160144474361e-08,
84
+ "raw/productive": 0.012008470067521558,
85
+ "raw/regularization": 0.029999999329447746,
86
+ "raw/v_allele": 0.2026680625230074,
87
+ "raw/v_end": 1.3741984504461289,
88
+ "raw/v_start": 1.3644738376140595,
89
+ "segmentation": 5.874486618041992,
90
+ "sigma": 0.75,
91
+ "v_allele_avg_gt_count": 1.453125,
92
+ "v_allele_setK_precision": 0.970703125,
93
+ "v_allele_setK_recall": 0.970703125,
94
+ "v_allele_thresh_precision": 0.9624255895614624,
95
+ "v_allele_thresh_recall": 0.9584728479385376,
96
+ "v_allele_top1_acc": 0.97265625,
97
+ "v_allele_top1_conf": 0.9090657234191895,
98
+ "v_end_acc": 0.837890625,
99
+ "v_end_acc_1nt": 0.908203125,
100
+ "v_end_mae": 0.306640625,
101
+ "v_start_acc": 0.8828125,
102
+ "v_start_acc_1nt": 0.939453125,
103
+ "v_start_mae": 0.33203125
104
+ },
105
+ "mixed_precision": true,
106
+ "optimizer_class": "AdamW",
107
+ "samples_per_epoch": 102400,
108
+ "wall_time_seconds": 61645
109
+ }
README.md CHANGED
@@ -1,3 +1,108 @@
1
- ---
2
- license: gpl-3.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: en
3
+ license: gpl-3.0
4
+ tags:
5
+ - immunoinformatics
6
+ - antibody
7
+ - TCR
8
+ - AIRR
9
+ - sequence-alignment
10
+ - bioinformatics
11
+ - pytorch
12
+ library_name: alignair
13
+ pipeline_tag: token-classification
14
+ ---
15
+
16
+ # AlignAIR Pretrained Models
17
+
18
+ **AlignAIR** is a deep learning tool for aligning immunoglobulin (IG) and T-cell receptor (TCR) sequences to germline gene databases. It simultaneously predicts V/D/J gene assignments, segment boundaries, mutation rates, and productivity β€” all in a single forward pass.
19
+
20
+ ## Available Models
21
+
22
+ | Model | Chain | Germline DB | V Alleles | D Alleles | J Alleles | Size |
23
+ |-------|-------|-------------|-----------|-----------|-----------|------|
24
+ | `HUMAN_IGH_OGRDB_576` | IGH (Heavy) | OGRDB | 198 | 33 | 7 | 17 MB |
25
+ | `HUMAN_IGH_EXTENDED_576` | IGH (Heavy) | Extended | 342 | 37 | 10 | 28 MB |
26
+ | `HUMAN_IGK_OGRDB_576` | IGK (Kappa) | OGRDB | 168 | β€” | 8 | 12 MB |
27
+ | `HUMAN_IGL_OGRDB_576` | IGL (Lambda) | OGRDB | 181 | β€” | 10 | 13 MB |
28
+ | `HUMAN_TCRB_IMGT_576` | TCRB (Beta) | IMGT | 130 | 3 | 14 | 12 MB |
29
+
30
+ All models use a maximum sequence length of 576 nucleotides and were trained on 1000 epochs of synthetic data generated by [GenAIRR](https://github.com/MuteJester/GenAIRR).
31
+
32
+ ## Quick Start
33
+
34
+ ```bash
35
+ pip install alignair[hub]
36
+ ```
37
+
38
+ ### Python API
39
+
40
+ ```python
41
+ from AlignAIR.Models import SingleChainAlignAIR
42
+ from AlignAIR.Hub import get_model_path
43
+
44
+ # Download and load a model (cached automatically)
45
+ model_path = get_model_path("igh") # or "HUMAN_IGH_OGRDB_576"
46
+ model = SingleChainAlignAIR.from_pretrained(model_path)
47
+ ```
48
+
49
+ ### CLI
50
+
51
+ ```bash
52
+ # Run inference with a pretrained model
53
+ alignair --model-dir HUMAN_IGH_OGRDB_576 input_sequences.csv -o results/
54
+ ```
55
+
56
+ ## Benchmark Results (100K synthetic sequences)
57
+
58
+ | Model | AlignAIR V | IgBLAST V | AlignAIR D | IgBLAST D | AlignAIR J | IgBLAST J | AlignAIR Speed |
59
+ |-------|-----------|-----------|-----------|-----------|-----------|-----------|----------------|
60
+ | IGH OGRDB | 94.1% | 95.5% | 81.7% | 69.8% | 99.3% | 99.5% | 4,272 seq/s |
61
+ | IGH Extended | 92.3% | 93.9% | 88.5% | 82.6% | 98.7% | 98.4% | 4,245 seq/s |
62
+ | IGK OGRDB | 94.6% | 95.4% | β€” | β€” | 97.2% | 96.0% | 4,807 seq/s |
63
+ | IGL OGRDB | 93.9% | 95.3% | β€” | β€” | 98.4% | 96.7% | 5,384 seq/s |
64
+ | TCRB IMGT | 96.5% | 96.2% | 89.6% | 76.3% | 99.6% | 99.1% | 4,317 seq/s |
65
+
66
+ Speed measured on NVIDIA RTX 3090 Ti (GPU) vs IgBLAST 1.22.0 (8 CPU threads).
67
+
68
+ ## Model Architecture
69
+
70
+ Each model is a `SingleChainAlignAIR` module combining:
71
+ - **Nucleotide embedding** (5β†’64 dim) with center-padded tokenization
72
+ - **Spatial segmentation** via 9-layer dilated convolutions (receptive field = 1023 nt)
73
+ - **Conditioned boundary heads** with chain decoding (v_start β†’ v_end β†’ d_start β†’ ...)
74
+ - **Classification heads** for V/D/J allele assignment
75
+ - **Analysis heads** for mutation rate and productivity prediction
76
+ - **In-model orientation correction** (4-class: forward, reverse-complement, complement, reverse)
77
+
78
+ ## Bundle Format
79
+
80
+ Each model directory contains:
81
+ - `model.pt` β€” PyTorch state dict
82
+ - `config.json` β€” Architecture hyperparameters
83
+ - `dataconfig.pkl` β€” Germline allele database (GenAIRR DataConfig)
84
+ - `training_meta.json` β€” Training provenance
85
+ - `VERSION` β€” Bundle format version
86
+ - `fingerprint.txt` β€” SHA-256 integrity hash
87
+
88
+ ## Citation
89
+
90
+ If you use AlignAIR in your research, please cite:
91
+
92
+ ```bibtex
93
+ @article{alignair2024,
94
+ title={AlignAIR: Deep Learning for Multi-Task Immunoglobulin Sequence Alignment},
95
+ author={Konstantinovsky, Thomas and Peres, Ayelet},
96
+ year={2024}
97
+ }
98
+ ```
99
+
100
+ ## License
101
+
102
+ GPL-3.0. See [LICENSE](https://github.com/MuteJester/AlignAIR/blob/main/LICENSE).
103
+
104
+ ## Links
105
+
106
+ - [GitHub Repository](https://github.com/MuteJester/AlignAIR)
107
+ - [Documentation](https://mutejester.github.io/AlignAIR/)
108
+ - [PyPI Package](https://pypi.org/project/AlignAIR/)