Views
No views yet
1merge_method: passthrough
2slices:
3- sources:
4 - layer_range: [0, 19]
5 model: unsloth/Mistral-Small-Instruct-2409
6# Original L19
7- sources:
8 - layer_range: [19, 20]
9 model: unsloth/Mistral-Small-Instruct-2409
10# Dupe A of L19
11- sources:
12 - layer_range: [19, 20]
13 model: unsloth/Mistral-Small-Instruct-2409
14 parameters:
15 scale:
16 - filter: o_proj
17 value: 0.0
18 - filter: down_proj
19 value: 0.0
20 - value: 1.0
21# Dupe B of L19
22- sources:
23 - layer_range: [19, 20]
24 model: unsloth/Mistral-Small-Instruct-2409
25 parameters:
26 scale:
27 - filter: o_proj
28 value: 0.0
29 - filter: down_proj
30 value: 0.0
31 - value: 1.0
32# Original L20
33- sources:
34 - layer_range: [20, 21]
35 model: unsloth/Mistral-Small-Instruct-2409
36# Dupe A of L20
37- sources:
38 - layer_range: [20, 21]
39 model: unsloth/Mistral-Small-Instruct-2409
40 parameters:
41 scale:
42 - filter: o_proj
43 value: 0.0
44 - filter: down_proj
45 value: 0.0
46 - value: 1.0
47# Dupe B of L20
48- sources:
49 - layer_range: [20, 21]
50 model: unsloth/Mistral-Small-Instruct-2409
51 parameters:
52 scale:
53 - filter: o_proj
54 value: 0.0
55 - filter: down_proj
56 value: 0.0
57 - value: 1.0
58# Original L21
59- sources:
60 - layer_range: [21, 22]
61 model: unsloth/Mistral-Small-Instruct-2409
62# Dupe A of L21
63- sources:
64 - layer_range: [21, 22]
65 model: unsloth/Mistral-Small-Instruct-2409
66 parameters:
67 scale:
68 - filter: o_proj
69 value: 0.0
70 - filter: down_proj
71 value: 0.0
72 - value: 1.0
73# Dupe B of L21
74- sources:
75 - layer_range: [21, 22]
76 model: unsloth/Mistral-Small-Instruct-2409
77 parameters:
78 scale:
79 - filter: o_proj
80 value: 0.0
81 - filter: down_proj
82 value: 0.0
83 - value: 1.0
84# Original L22
85- sources:
86 - layer_range: [22, 23]
87 model: unsloth/Mistral-Small-Instruct-2409
88# Dupe A of L22
89- sources:
90 - layer_range: [22, 23]
91 model: unsloth/Mistral-Small-Instruct-2409
92 parameters:
93 scale:
94 - filter: o_proj
95 value: 0.0
96 - filter: down_proj
97 value: 0.0
98 - value: 1.0
99# Dupe B of L22
100- sources:
101 - layer_range: [22, 23]
102 model: unsloth/Mistral-Small-Instruct-2409
103 parameters:
104 scale:
105 - filter: o_proj
106 value: 0.0
107 - filter: down_proj
108 value: 0.0
109 - value: 1.0
110# Original L23
111- sources:
112 - layer_range: [23, 24]
113 model: unsloth/Mistral-Small-Instruct-2409
114# Dupe A of L23
115- sources:
116 - layer_range: [23, 24]
117 model: unsloth/Mistral-Small-Instruct-2409
118 parameters:
119 scale:
120 - filter: o_proj
121 value: 0.0
122 - filter: down_proj
123 value: 0.0
124 - value: 1.0
125# Dupe B of L23
126- sources:
127 - layer_range: [23, 24]
128 model: unsloth/Mistral-Small-Instruct-2409
129 parameters:
130 scale:
131 - filter: o_proj
132 value: 0.0
133 - filter: down_proj
134 value: 0.0
135 - value: 1.0
136# Original L24
137- sources:
138 - layer_range: [24, 25]
139 model: unsloth/Mistral-Small-Instruct-2409
140# Dupe A of L24
141- sources:
142 - layer_range: [24, 25]
143 model: unsloth/Mistral-Small-Instruct-2409
144 parameters:
145 scale:
146 - filter: o_proj
147 value: 0.0
148 - filter: down_proj
149 value: 0.0
150 - value: 1.0
151# Dupe B of L24
152- sources:
153 - layer_range: [24, 25]
154 model: unsloth/Mistral-Small-Instruct-2409
155 parameters:
156 scale:
157 - filter: o_proj
158 value: 0.0
159 - filter: down_proj
160 value: 0.0
161 - value: 1.0
162# Original L25
163- sources:
164 - layer_range: [25, 26]
165 model: unsloth/Mistral-Small-Instruct-2409
166# Dupe A of L25
167- sources:
168 - layer_range: [25, 26]
169 model: unsloth/Mistral-Small-Instruct-2409
170 parameters:
171 scale:
172 - filter: o_proj
173 value: 0.0
174 - filter: down_proj
175 value: 0.0
176 - value: 1.0
177# Dupe B of L25
178- sources:
179 - layer_range: [25, 26]
180 model: unsloth/Mistral-Small-Instruct-2409
181 parameters:
182 scale:
183 - filter: o_proj
184 value: 0.0
185 - filter: down_proj
186 value: 0.0
187 - value: 1.0
188# Original L26
189- sources:
190 - layer_range: [26, 27]
191 model: unsloth/Mistral-Small-Instruct-2409
192# Dupe A of L26
193- sources:
194 - layer_range: [26, 27]
195 model: unsloth/Mistral-Small-Instruct-2409
196 parameters:
197 scale:
198 - filter: o_proj
199 value: 0.0
200 - filter: down_proj
201 value: 0.0
202 - value: 1.0
203# Dupe B of L26
204- sources:
205 - layer_range: [26, 27]
206 model: unsloth/Mistral-Small-Instruct-2409
207 parameters:
208 scale:
209 - filter: o_proj
210 value: 0.0
211 - filter: down_proj
212 value: 0.0
213 - value: 1.0
214# Original L27
215- sources:
216 - layer_range: [27, 28]
217 model: unsloth/Mistral-Small-Instruct-2409
218# Dupe A of L27
219- sources:
220 - layer_range: [27, 28]
221 model: unsloth/Mistral-Small-Instruct-2409
222 parameters:
223 scale:
224 - filter: o_proj
225 value: 0.0
226 - filter: down_proj
227 value: 0.0
228 - value: 1.0
229# Dupe B of L27
230- sources:
231 - layer_range: [27, 28]
232 model: unsloth/Mistral-Small-Instruct-2409
233 parameters:
234 scale:
235 - filter: o_proj
236 value: 0.0
237 - filter: down_proj
238 value: 0.0
239 - value: 1.0
240# Original L28
241- sources:
242 - layer_range: [28, 29]
243 model: unsloth/Mistral-Small-Instruct-2409
244# Dupe A of L28
245- sources:
246 - layer_range: [28, 29]
247 model: unsloth/Mistral-Small-Instruct-2409
248 parameters:
249 scale:
250 - filter: o_proj
251 value: 0.0
252 - filter: down_proj
253 value: 0.0
254 - value: 1.0
255# Dupe B of L28
256- sources:
257 - layer_range: [28, 29]
258 model: unsloth/Mistral-Small-Instruct-2409
259 parameters:
260 scale:
261 - filter: o_proj
262 value: 0.0
263 - filter: down_proj
264 value: 0.0
265 - value: 1.0
266# Original L29
267- sources:
268 - layer_range: [29, 30]
269 model: unsloth/Mistral-Small-Instruct-2409
270# Dupe A of L29
271- sources:
272 - layer_range: [29, 30]
273 model: unsloth/Mistral-Small-Instruct-2409
274 parameters:
275 scale:
276 - filter: o_proj
277 value: 0.0
278 - filter: down_proj
279 value: 0.0
280 - value: 1.0
281# Dupe B of L29
282- sources:
283 - layer_range: [29, 30]
284 model: unsloth/Mistral-Small-Instruct-2409
285 parameters:
286 scale:
287 - filter: o_proj
288 value: 0.0
289 - filter: down_proj
290 value: 0.0
291 - value: 1.0
292# Original L30
293- sources:
294 - layer_range: [30, 31]
295 model: unsloth/Mistral-Small-Instruct-2409
296# Dupe A of L30
297- sources:
298 - layer_range: [30, 31]
299 model: unsloth/Mistral-Small-Instruct-2409
300 parameters:
301 scale:
302 - filter: o_proj
303 value: 0.0
304 - filter: down_proj
305 value: 0.0
306 - value: 1.0
307# Dupe B of L30
308- sources:
309 - layer_range: [30, 31]
310 model: unsloth/Mistral-Small-Instruct-2409
311 parameters:
312 scale:
313 - filter: o_proj
314 value: 0.0
315 - filter: down_proj
316 value: 0.0
317 - value: 1.0
318# Original L31
319- sources:
320 - layer_range: [31, 32]
321 model: unsloth/Mistral-Small-Instruct-2409
322# Dupe A of L31
323- sources:
324 - layer_range: [31, 32]
325 model: unsloth/Mistral-Small-Instruct-2409
326 parameters:
327 scale:
328 - filter: o_proj
329 value: 0.0
330 - filter: down_proj
331 value: 0.0
332 - value: 1.0
333# Dupe B of L31
334- sources:
335 - layer_range: [31, 32]
336 model: unsloth/Mistral-Small-Instruct-2409
337 parameters:
338 scale:
339 - filter: o_proj
340 value: 0.0
341 - filter: down_proj
342 value: 0.0
343 - value: 1.0
344# Original L32
345- sources:
346 - layer_range: [32, 33]
347 model: unsloth/Mistral-Small-Instruct-2409
348# Dupe A of L32
349- sources:
350 - layer_range: [32, 33]
351 model: unsloth/Mistral-Small-Instruct-2409
352 parameters:
353 scale:
354 - filter: o_proj
355 value: 0.0
356 - filter: down_proj
357 value: 0.0
358 - value: 1.0
359# Dupe B of L32
360- sources:
361 - layer_range: [32, 33]
362 model: unsloth/Mistral-Small-Instruct-2409
363 parameters:
364 scale:
365 - filter: o_proj
366 value: 0.0
367 - filter: down_proj
368 value: 0.0
369 - value: 1.0
370# Original L33
371- sources:
372 - layer_range: [33, 34]
373 model: unsloth/Mistral-Small-Instruct-2409
374# Dupe A of L33
375- sources:
376 - layer_range: [33, 34]
377 model: unsloth/Mistral-Small-Instruct-2409
378 parameters:
379 scale:
380 - filter: o_proj
381 value: 0.0
382 - filter: down_proj
383 value: 0.0
384 - value: 1.0
385# Dupe B of L33
386- sources:
387 - layer_range: [33, 34]
388 model: unsloth/Mistral-Small-Instruct-2409
389 parameters:
390 scale:
391 - filter: o_proj
392 value: 0.0
393 - filter: down_proj
394 value: 0.0
395 - value: 1.0
396# Original L34
397- sources:
398 - layer_range: [34, 35]
399 model: unsloth/Mistral-Small-Instruct-2409
400# Dupe A of L34
401- sources:
402 - layer_range: [34, 35]
403 model: unsloth/Mistral-Small-Instruct-2409
404 parameters:
405 scale:
406 - filter: o_proj
407 value: 0.0
408 - filter: down_proj
409 value: 0.0
410 - value: 1.0
411# Dupe B of L34
412- sources:
413 - layer_range: [34, 35]
414 model: unsloth/Mistral-Small-Instruct-2409
415 parameters:
416 scale:
417 - filter: o_proj
418 value: 0.0
419 - filter: down_proj
420 value: 0.0
421 - value: 1.0
422# Original L35
423- sources:
424 - layer_range: [35, 36]
425 model: unsloth/Mistral-Small-Instruct-2409
426# Dupe A of L35
427- sources:
428 - layer_range: [35, 36]
429 model: unsloth/Mistral-Small-Instruct-2409
430 parameters:
431 scale:
432 - filter: o_proj
433 value: 0.0
434 - filter: down_proj
435 value: 0.0
436 - value: 1.0
437# Dupe B of L35
438- sources:
439 - layer_range: [35, 36]
440 model: unsloth/Mistral-Small-Instruct-2409
441 parameters:
442 scale:
443 - filter: o_proj
444 value: 0.0
445 - filter: down_proj
446 value: 0.0
447 - value: 1.0
448# Original L36
449- sources:
450 - layer_range: [36, 37]
451 model: unsloth/Mistral-Small-Instruct-2409
452# Dupe A of L36
453- sources:
454 - layer_range: [36, 37]
455 model: unsloth/Mistral-Small-Instruct-2409
456 parameters:
457 scale:
458 - filter: o_proj
459 value: 0.0
460 - filter: down_proj
461 value: 0.0
462 - value: 1.0
463# Dupe B of L36
464- sources:
465 - layer_range: [36, 37]
466 model: unsloth/Mistral-Small-Instruct-2409
467 parameters:
468 scale:
469 - filter: o_proj
470 value: 0.0
471 - filter: down_proj
472 value: 0.0
473 - value: 1.0
474# Original L37
475- sources:
476 - layer_range: [37, 38]
477 model: unsloth/Mistral-Small-Instruct-2409
478# Dupe A of L37
479- sources:
480 - layer_range: [37, 38]
481 model: unsloth/Mistral-Small-Instruct-2409
482 parameters:
483 scale:
484 - filter: o_proj
485 value: 0.0
486 - filter: down_proj
487 value: 0.0
488 - value: 1.0
489# Dupe B of L37
490- sources:
491 - layer_range: [37, 38]
492 model: unsloth/Mistral-Small-Instruct-2409
493 parameters:
494 scale:
495 - filter: o_proj
496 value: 0.0
497 - filter: down_proj
498 value: 0.0
499 - value: 1.0
500# Original L38
501- sources:
502 - layer_range: [38, 39]
503 model: unsloth/Mistral-Small-Instruct-2409
504# Dupe A of L38
505- sources:
506 - layer_range: [38, 39]
507 model: unsloth/Mistral-Small-Instruct-2409
508 parameters:
509 scale:
510 - filter: o_proj
511 value: 0.0
512 - filter: down_proj
513 value: 0.0
514 - value: 1.0
515# Dupe B of L38
516- sources:
517 - layer_range: [38, 39]
518 model: unsloth/Mistral-Small-Instruct-2409
519 parameters:
520 scale:
521 - filter: o_proj
522 value: 0.0
523 - filter: down_proj
524 value: 0.0
525 - value: 1.0
526# Original L39
527- sources:
528 - layer_range: [39, 40]
529 model: unsloth/Mistral-Small-Instruct-2409
530# Dupe A of L39
531- sources:
532 - layer_range: [39, 40]
533 model: unsloth/Mistral-Small-Instruct-2409
534 parameters:
535 scale:
536 - filter: o_proj
537 value: 0.0
538 - filter: down_proj
539 value: 0.0
540 - value: 1.0
541# Dupe B of L39
542- sources:
543 - layer_range: [39, 40]
544 model: unsloth/Mistral-Small-Instruct-2409
545 parameters:
546 scale:
547 - filter: o_proj
548 value: 0.0
549 - filter: down_proj
550 value: 0.0
551 - value: 1.0
552# Original L40
553- sources:
554 - layer_range: [40, 41]
555 model: unsloth/Mistral-Small-Instruct-2409
556# Dupe A of L40
557- sources:
558 - layer_range: [40, 41]
559 model: unsloth/Mistral-Small-Instruct-2409
560 parameters:
561 scale:
562 - filter: o_proj
563 value: 0.0
564 - filter: down_proj
565 value: 0.0
566 - value: 1.0
567# Dupe B of L40
568- sources:
569 - layer_range: [40, 41]
570 model: unsloth/Mistral-Small-Instruct-2409
571 parameters:
572 scale:
573 - filter: o_proj
574 value: 0.0
575 - filter: down_proj
576 value: 0.0
577 - value: 1.0
578# ... REPEAT UNTIL 41
579- sources:
580 - layer_range: [41, 55]
581 model: unsloth/Mistral-Small-Instruct-2409