rabaevn commited on
Commit
73b0042
·
verified ·
1 Parent(s): 87f6531

Training in progress, step 200, checkpoint

Browse files
.gitattributes CHANGED
@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  tokenizer.json filter=lfs diff=lfs merge=lfs -text
 
 
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  tokenizer.json filter=lfs diff=lfs merge=lfs -text
37
+ last-checkpoint/tokenizer.json filter=lfs diff=lfs merge=lfs -text
last-checkpoint/1_Pooling/config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "word_embedding_dimension": 768,
3
+ "pooling_mode_cls_token": false,
4
+ "pooling_mode_mean_tokens": true,
5
+ "pooling_mode_max_tokens": false,
6
+ "pooling_mode_mean_sqrt_len_tokens": false,
7
+ "pooling_mode_weightedmean_tokens": false,
8
+ "pooling_mode_lasttoken": false,
9
+ "include_prompt": true
10
+ }
last-checkpoint/2_Dense/config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "in_features": 768,
3
+ "out_features": 3072,
4
+ "bias": false,
5
+ "activation_function": "torch.nn.modules.linear.Identity"
6
+ }
last-checkpoint/2_Dense/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d32bfa5f6e9259dec81cc5bba77922c05ea4450cb5363b9e39a8e3b6efee4c13
3
+ size 9437272
last-checkpoint/3_Dense/config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "in_features": 3072,
3
+ "out_features": 768,
4
+ "bias": false,
5
+ "activation_function": "torch.nn.modules.linear.Identity"
6
+ }
last-checkpoint/3_Dense/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d4776c819cbf9cd976833c5cb4487169ab05deae90b4dabcc292d2a9d8737e2
3
+ size 9437272
last-checkpoint/README.md ADDED
@@ -0,0 +1,750 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ license: apache-2.0
5
+ tags:
6
+ - sentence-transformers
7
+ - sentence-similarity
8
+ - feature-extraction
9
+ - dense
10
+ - generated_from_trainer
11
+ - dataset_size:386737
12
+ - loss:CachedMultipleNegativesRankingLoss
13
+ base_model: google/embeddinggemma-300m
14
+ widget:
15
+ - source_sentence: "title: \nCar Wars Arena Game"
16
+ sentences:
17
+ - 'description
18
+
19
+ Cute Duck on Swing can not only be used in car mirrors but can also be hung in
20
+ your garden windowsill kitchen etc in your office or home
21
+
22
+ Drop your favorite perfume oil on the bottom
23
+
24
+ After absorbing the essential oil it will give off a charming fragrance
25
+
26
+ Cute car mirror trim ducks are easy to install Just fasten the lanyard and hang
27
+ it on the car rearview mirrorCute Duck on Swing Car Ornament is perfect for hanging
28
+ car mirrors beautiful decorations in car interior and home decoration The Swing
29
+ Duck reflects the owners personality and aesthetic and harmonises with the interior
30
+
31
+ Reliable material the Cute Duck on Swing car pendant with a lovely shape is mainly
32
+ made of gypsum material stable and smooth solid and sturdy offering a long service
33
+ life And you can also add some perfume to it which will make your driving tour
34
+ more relaxing
35
+
36
+ Quantity and dimension you will receive 1 piece of Cute Duck on Swing car hanging
37
+ ornament in the package The rope measures about 15 inches in length and the rope
38
+ is adjustable according to your needs
39
+
40
+ Safe and comfortable Small and exquisite does not affect the driving line of sight
41
+ The main body part is the right length to avoid hitting the glass when braking
42
+ The Swing Duck will shake when driving which can eliminate the drivers visual
43
+ fatigue very well
44
+
45
+ Aftersales guarantee The product is covered and packaged with bubble film which
46
+ is not easy to be damaged If you have any questions please contact us and we will
47
+ solve it within 24 hours'
48
+ - 'description
49
+
50
+ This new version of Spot It combines different characters from Marvel comics with
51
+ one of the most emblematic family games Players try to spot the single matching
52
+ symbol between two cards in a playful battle of speedy observation Featuring some
53
+ of the most iconic Marvel Super Heroes Iron Man Captain America Black WidowMARVEL
54
+ CARD GAME Spot It hones players observational skills and lightningfast reflexes
55
+ as the whole family enjoys five different games modes that test whos the fastest
56
+ to spot matching symbols and call them out Marvel fans will race to find some
57
+ of the most iconic Super Heroes Iron Man Captain America and Black Widow
58
+
59
+ FAMILY GAME NIGHT Fun symbols and pictures cover every card in a Spot It deck
60
+ making it the perfect game for family game night Each card has exactly one matching
61
+ symbol or picture in common with the other cards in the deck Spot it first and
62
+ you win
63
+
64
+ BEST KIDS GAME Spot It helps develop focus visual perception skills speechlanguage
65
+ skills and fine motor skills Players of all ages will enjoy the engaging tactile
66
+ gameplay
67
+
68
+ PERFECT TRAVEL GAME Spot It Is fast to learn and teach and its portable tin makes
69
+ it the perfect game to take anywhere
70
+
71
+ NUMBER OF PLAYERS AND AVERAGE PLAYTIME Spot It Marvel Emojis is designed for 2
72
+ to 8 players and is the perfect game for kids ages 6 and up The average playtime
73
+ is approximately 15 minutes'
74
+ - 'description
75
+
76
+ To be the best Youve got to beat the best and that means taking to the many
77
+ different autodueling arenas across the Car Wars world and testing yourself under
78
+ all kinds of conditions Car Wars Arenas gives hardcore autoduelists more exciting
79
+ locations to showcase their skills and blow away their opponents The boxed set
80
+ contains five onesided 22 x 34 arena maps and a booklet detailing the special
81
+ features of each arena The rules will also cover revised versions of popular variants
82
+ like corporate team dueling an AADA pro circuit and more The arena designs in
83
+ Car Wars Arenas have appeared in various Car Wars supplements including the Car
84
+ Wars Arena Book and The AADA Duel Circuit LOutrance but only as scaleddown maps
85
+ In this set they come out of the box ready to play in full Car Wars Classic scaleThe
86
+ most dangerous arenas in autoduelling history printed at full scale and ready
87
+ to play
88
+
89
+ Expansion for base game'
90
+ - source_sentence: "title: \nTrains Locomotives Railroad Fan Degree Custom Gag Diploma\
91
+ \ Doctorate Certificate Funny Customized Joke Gift Novelty Item"
92
+ sentences:
93
+ - 'description
94
+
95
+ One customized novelty certificate 85 x 11 inch printed on premium certificate
96
+ paper with official border Includes embossed Gold Seal on certificate Custom produced
97
+ with your own personalized information Any name and any date you chooseFully customized
98
+ with ANY NAME and ANY DATE of your choice
99
+
100
+ Great custom novelty gift with a personalized touch
101
+
102
+ Show your uniqueness or celebrate someones favorite interest talent or hobby
103
+
104
+ Top quality certificate also includes official embossed GOLD SEAL for added appeal
105
+
106
+ Personalized certificate makes a thoughtful gift for any occasion'
107
+ - 'description
108
+
109
+ Magic The Gathering Ashiok Nightmare Muse Theros Beyond DeathName Ashiok Nightmare
110
+ Muse
111
+
112
+ Set Theros Beyond Death
113
+
114
+ A single individual card from the Magic the Gathering MTG trading and collectible
115
+ card game TCGCCG'
116
+ - 'description
117
+
118
+ Air Dancers inflatable tube man 20ft blue custom embroidered with WINDOW TINT
119
+ down the center in white lettering This same message is embroidered on the second
120
+ side as well This Air Dancers inflatable tube man is compatible with all 18 diameter
121
+ Velcro mount blowers Blower not included Spend the extra money to let your customers
122
+ know what you are promotingDesigned to Grab Attention the dynamic waving motion
123
+ fringed hair eyes and iconic face attract customers to your business and bring
124
+ attention to whatever youre promoting
125
+
126
+ Long Lasting Durable Material LookOurWay Air Dancers are constructed with high
127
+ strength polyamide nylon silk with added tarpaulin ensuring a longlasting flexible
128
+ ripproof product
129
+
130
+ CostEffective Marketing Tool Promote your business or grand opening event with
131
+ our affordable dancing tube man Air Dancers are a great way to get your business
132
+ noticed at a low cost
133
+
134
+ Perfect Promotional Gift Is anyone close to you starting their new business or
135
+ hosting an event Surprise them with this innovative advertising gift and help
136
+ them attract new customers in style
137
+
138
+ Product Specifications Air Dancers inflatable tube man attachment height 20 Feet Attachment
139
+ diameter 18 inches Compatible with all 18 inch diameter velcro mount blowers Blower
140
+ Not Included'
141
+ - source_sentence: "title: \nHobbyPark Aluminum Axle Carriers Knuckle LR Replacement\
142
+ \ of 5334 for Traxxas 110 ERevo BrushlessRevo 33SummitEMaxxTMaxx RC Monster Truck\
143
+ \ 4Pack Blue"
144
+ sentences:
145
+ - 'description
146
+
147
+ Hunt for eggs with this fun LEGO BrickHeadz Easter Bunny construction character
148
+ Check out its cute cheeks and enormous movable ears This perfect seasonal gift
149
+ also comes with a detachable carrot and bucket plus 2 Easter eggs and flowers
150
+ and stands on a buildable collectors baseplate with a seasonal calendar and BrickHeadz
151
+ logo for easy display in the home office or anywhere they likeBuildable LEGO BrickHeadz
152
+ Easter Bunny construction character features decorated eyes movable ears and a
153
+ detachable carrot and bucket
154
+
155
+ Also includes 2 buildable Easter eggs and flowers
156
+
157
+ Each LEGO BrickHeadz construction character comes with its own buildable collectors
158
+ baseplate featuring a seasonal calendar and BrickHeadz logo
159
+
160
+ Have fun growing your LEGO BrickHeadz collection with other characters from your
161
+ favorite films TV series games and comics
162
+
163
+ Mash up your LEGO BrickHeadz construction characters to create supercool hybrids
164
+ or your own amazing characters
165
+
166
+ Stands over 4inch 12cm tall without baseplate and baseplate measures over 4inch
167
+ 12cm wide and 1inch 5cm deep'
168
+ - 'description
169
+
170
+ LITTLE RETRO TOYS THAT PROMISE LOTS OF FUN
171
+
172
+ Ol School Entertainment
173
+
174
+ Simple yet addictively fun uncomplicated yet thoroughly entertaining these rubber
175
+ ball poppers keep sparking smiles over and over again Inspired by their 90s counterparts
176
+ these retro popup toys fuel the fun with a simple action Just flip them inside
177
+ out place them on a flat surface wait for a second or two and watch them let out
178
+ a delightful pop and jump off the ground Little ones will love the fun assortment
179
+ of color and superhero terms such as BAM POW BOP sayings among others Theyre perfect
180
+ for fun indoors outdoors and away from home
181
+
182
+ Great for Gifting
183
+
184
+ Whether youre looking for party favors for your little ones birthday party unique
185
+ party supplies prizes for that kids carnival or contest or good behavior incentives
186
+ for your star students these rubber poppers are guaranteed to be a hit Every set
187
+ not only comes with 12 poppers to give you enough to dish around but also a 100
188
+ moneyback guarantee to give you total peace of mind Inspire a rush of smiles and
189
+ giggles without having anything to lose
190
+
191
+ Heres why youll love these popper toys
192
+
193
+ Come in a pack of 12 to give you great value
194
+
195
+ Includes a fun assortment of superhero terms for even more fun
196
+
197
+ Very easy to use by kids as young as 3
198
+
199
+ Great for keeping little ones occupied when traveling
200
+
201
+ Make great party favors for boys and girls
202
+
203
+ Backed by a nohassle 100 moneyback guarantee
204
+
205
+ Spread smiles and delight with these superfun retro poppersBULK PACK OF 12 Get
206
+ the best bang for your buck as you treat the kiddos to a total blast Every set
207
+ comes with 12 superhero themed poppers They include terms such as BAM POW BOP
208
+ sayings among others Assortment may vary
209
+
210
+ AWESOME RETRO FUN A blast from the past that the little ones will simply love
211
+ These half rubber ball poppers are simple but oh so fun Flip them inside out place
212
+ them on a flat surface and watch them pop Just what you need to spark instant
213
+ smiles
214
+
215
+ TAKE THEM ON THE GO Pocketsized fun to make time away from home less of a drag
216
+ These popup rubber ball toys measure 175 in diameter Slip one in your pocket whip
217
+ it out whenever boredom starts to set in and brighten up any slow day with old
218
+ school fun
219
+
220
+ COOL PARTY FAVORS Looking for birthday party favors for boys and girls Goodie
221
+ treat bag fillers that suit just about any theme The jumping ball toys are guaranteed
222
+ to be a hit Theyre also great as piata fillers good behavior incentives and stocking
223
+ stuffers For kids 3
224
+
225
+ BUY RISKFREE We fully stand behind our products with a best satisfaction and 100
226
+ moneyback guarantee Not satisfied with these superhero sayings poppers Well send
227
+ you a replacement or issue a full refund Click Add to Cart now to get your set
228
+ riskfree'
229
+ - 'description
230
+
231
+ Note The color of the item may vary slightly due to photography and your own computer
232
+ Theres some little error from human measuring
233
+
234
+ Customer Satisfaction Guarantee
235
+
236
+ If you are not 100 completely satisfied with our products please do not hesitate
237
+ to contact us to request a refund or exchange
238
+
239
+ Item
240
+
241
+ Aluminum Axle Carriers Knuckle LR
242
+
243
+ Package Include
244
+
245
+ 4 pieces 2x Right 2xLeft
246
+
247
+ Features
248
+
249
+ Made Of Quality Aluminum AlloyReplacement of Part 53345334R CNC machined for precision
250
+
251
+ Compatible with
252
+
253
+ Traxxas 110 EMaxx Brushless EMaxx ERevo Brushless ERevo TMaxx RC Monster Truck
254
+ Car
255
+
256
+ Note
257
+
258
+ Check your specific models manual for compatibilityor Contact usCNC machined for
259
+ precisionIncreased strength and precisionEasy upgrade from the original part
260
+
261
+ Compatible with Traxxas 110 EMaxx Brushless EMaxx ERevo Brushless ERevo TMaxx
262
+ RC Monster Truck Car
263
+
264
+ Replacement of Part 53345334R
265
+
266
+ Package Include 4 pieces As shown in the picture
267
+
268
+ Produced By HobbyPark Well made and durable100 New'
269
+ - source_sentence: "title: \nSkyrocket Beastie Buds Interactive Electronic Toy for\
270
+ \ Boys and Girls Snap Dragon Dancing Slug Chomping Plant with Attitude"
271
+ sentences:
272
+ - "description\nSpecification\nMaterial metal Color as shown Quantity 1x dollhouse\
273
+ \ coffee pot\nNote\n Keep the miniatures dry and cool avoid sunlight Small parts\
274
+ \ please use under the supervision of adults Customer Service We have a 100 satisfaction\
275
+ \ guarantee If you have any question before or after purchasing please feel free\
276
+ \ to contact us\nOdoria Miniature\nMeet more miniatures in our shop\nOdoria\n\
277
+ \ Youve never seen these beforeScale 112 miniature dollhouse accessories\nPackage\
278
+ \ Included 1x miniature coffee pot Not Including Other Items\nDimension LxWxH\
279
+ \ 11x06x11 inch\nApplication Suitable for 112 scale miniature kitchen decoration\
280
+ \ delicate ornaments for dollhouse additional items to make your dollhouse more\
281
+ \ complete\nHigh Quality Exquisite details perfect as gift for friends It will\
282
+ \ bring more fun to your dollhouse decoration"
283
+ - 'description
284
+
285
+ From the anime series Demon Slayer Kimetsu no Yaiba comes a DX figma of Inosuke
286
+ Hashibira Using the smooth yet poseable joints of figma you can create a variety
287
+ of poses from the series A flexible plastic is used for important areas allowing
288
+ proportions to be kept without compromising posability He comes with three face
289
+ plates including a standard face an unmasked face and an angry face Optional parts
290
+ include his Nichirin Blade expression effect stickers his boar mask and attacking
291
+ effect parts for recreating his Beast Breathing techniques An articulated figma
292
+ stand is included to display the figma in a variety of posesA Max Factory import
293
+
294
+ From the hit anime Demon Slayer Kimetsu no Yaiba
295
+
296
+ Comes with three face plates including a standard expression unmasked face and
297
+ an angry face
298
+
299
+ Optional parts include his Nichirin Blade expression effect stickers his boar
300
+ mask and attacking effect parts
301
+
302
+ Articulated figma stand is included for multiple posing options'
303
+ - 'description
304
+
305
+ Feeeeed Me These arent your mommas spring flowers Beastie Buds are alive Theyre
306
+ snarky theyre cantankerous and theyre HUNGRY These funny interactive plants are
307
+ real beasts to take care of Pet their heads and they rise from the dirt Rub the
308
+ wrong wayLook outthey might bite Beastie Buds dont like sun and water they prefer
309
+ eating slugs and bugs Gross They mumble and grunt but music can soothe these savage
310
+ beasts Plug in a boom box and theyll dance to the beat of their favorite tunes
311
+ Beastie Buds the plants with attitudeInteractive plant creature that reacts to
312
+ your touch
313
+
314
+ Feed him slug balls again and again
315
+
316
+ Includes boom box to make him dance
317
+
318
+ 50 sounds reactions'
319
+ - source_sentence: "title: \nHatchimal Pixies Riders Lilac Luna Pixie and Swanling\
320
+ \ Glider Set with Mystery Feature"
321
+ sentences:
322
+ - 'description
323
+
324
+ From the Manufacturer
325
+
326
+ A fun and girly way to liven up your wallssatin soft sculpture for wall or ceiling
327
+ decoration5 color striped satin rainbowSatin soft sculpture for wall or ceiling
328
+ decoration
329
+
330
+ 5 color striped satin rainbow
331
+
332
+ Blue satin clouds with heart pattern
333
+
334
+ Satin flower and heart motifs with Satin ribbons
335
+
336
+ Crystal beaded fringe'
337
+ - 'description
338
+
339
+ Introducing an allnew mystical connection between Hatchimals and Pixies Pixies
340
+ Riders These beautiful Pixies feature fluttery wings poseable heads and legs and
341
+ come with a matching Glider they can really ride When these stunning Hatchimals
342
+ duos ride together they measure 35 inches tall Pixies Riders have a magical unboxing
343
+ experience First hatch the heart to reveal your perfect Pixie Next slide the box
344
+ out of its sleeve and open the enchanted castle inside Discover your gorgeous
345
+ Hatchimals Glider necklaces and tiaras for your Pixie and Glider a display stand
346
+ and an exclusive Hatchtopia Life code to use in the free app compatible on iOS
347
+ and Android devices Once youve unboxed your fantastical Pixie and Glider the box
348
+ becomes a palace playset unique to your Pixies Riders duo You can also flip the
349
+ heart over and use it to display your new friends in front of their castle When
350
+ youre finished playing close up the box and take it with you anywhere you go With
351
+ 10 Pixies Riders to collect including two special editions each sold separately
352
+ which mystical pair will be your favorite Each Pixies Riders duo has a unique
353
+ feature like metallic glowinthedark color change and more Add all of the majestic
354
+ Pixies Riders to your collection and ride into adventurePIXIES RIDER AND HATCHIMALS
355
+ GLIDER This Pixie has fluttery wings poseable legs and comes with a matching Glider
356
+ she can really ride Both have beautiful mystical details and poseable heads
357
+
358
+ MYSTERY FEATURES AND MATCHING ACCESSORIES Each duo includes 4 matching accessories
359
+ and 1 of 10 unique features to discover like fuzzy color change metallic and more
360
+ What feature will you reveal
361
+
362
+ MAGICAL UNBOXING The box becomes a playset and a beautiful display for your Pixies
363
+ Riders Each box has different artwork to enhance your magical Hatchimals storytelling
364
+ Collect all 10 castles
365
+
366
+ Includes 1 Hatchimals Pixie 1 Hatchimals Glider 4 Accessories 1 Display Stand
367
+ 1 Checklist 1 Instruction Sheet 1 Hatchtopia Life App Token'
368
+ - 'description
369
+
370
+ The CRIMSON GUARD are the elite shock troops of the COBRA legions All Sieges must
371
+ hold a degree in either law or accounting and must be in top physical condition
372
+ Final stages of training take place in the deepest recesses of COBRA Headquarters
373
+ and are purported to involve an initiation ceremony too hideous to describeBuilding
374
+ the perfect army just got a little easier Great for the real GI Joe collectorCelebrate
375
+ 25 years of the ultimate action team with these articulated action figures Figures
376
+ also comes with interchangeable weaponsGI Joe 25th Anniversary 3 34 Action Figure
377
+ Collection from Hasbro
378
+
379
+ This Toys R Us exclusive trooper builder set includes 5 Crimson Guard action figures
380
+
381
+ For Ages 5 Up'
382
+ pipeline_tag: sentence-similarity
383
+ library_name: sentence-transformers
384
+ ---
385
+
386
+ # EmbeddingGemma-300m trained on toys and games
387
+
388
+ This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [google/embeddinggemma-300m](https://huggingface.co/google/embeddinggemma-300m). It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
389
+
390
+ ## Model Details
391
+
392
+ ### Model Description
393
+ - **Model Type:** Sentence Transformer
394
+ - **Base model:** [google/embeddinggemma-300m](https://huggingface.co/google/embeddinggemma-300m) <!-- at revision 57c266a740f537b4dc058e1b0cda161fd15afa75 -->
395
+ - **Maximum Sequence Length:** 256 tokens
396
+ - **Output Dimensionality:** 768 dimensions
397
+ - **Similarity Function:** Cosine Similarity
398
+ <!-- - **Training Dataset:** Unknown -->
399
+ - **Language:** en
400
+ - **License:** apache-2.0
401
+
402
+ ### Model Sources
403
+
404
+ - **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
405
+ - **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers)
406
+ - **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)
407
+
408
+ ### Full Model Architecture
409
+
410
+ ```
411
+ SentenceTransformer(
412
+ (0): Transformer({'max_seq_length': 256, 'do_lower_case': False, 'architecture': 'Gemma3TextModel'})
413
+ (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
414
+ (2): Dense({'in_features': 768, 'out_features': 3072, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity'})
415
+ (3): Dense({'in_features': 3072, 'out_features': 768, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity'})
416
+ (4): Normalize()
417
+ )
418
+ ```
419
+
420
+ ## Usage
421
+
422
+ ### Direct Usage (Sentence Transformers)
423
+
424
+ First install the Sentence Transformers library:
425
+
426
+ ```bash
427
+ pip install -U sentence-transformers
428
+ ```
429
+
430
+ Then you can load this model and run inference.
431
+ ```python
432
+ from sentence_transformers import SentenceTransformer
433
+
434
+ # Download from the 🤗 Hub
435
+ model = SentenceTransformer("rabaevn/EncodeRec_Toys")
436
+ # Run inference
437
+ queries = [
438
+ "title: \nHatchimal Pixies Riders Lilac Luna Pixie and Swanling Glider Set with Mystery Feature",
439
+ ]
440
+ documents = [
441
+ 'description\nIntroducing an allnew mystical connection between Hatchimals and Pixies Pixies Riders These beautiful Pixies feature fluttery wings poseable heads and legs and come with a matching Glider they can really ride When these stunning Hatchimals duos ride together they measure 35 inches tall Pixies Riders have a magical unboxing experience First hatch the heart to reveal your perfect Pixie Next slide the box out of its sleeve and open the enchanted castle inside Discover your gorgeous Hatchimals Glider necklaces and tiaras for your Pixie and Glider a display stand and an exclusive Hatchtopia Life code to use in the free app compatible on iOS and Android devices Once youve unboxed your fantastical Pixie and Glider the box becomes a palace playset unique to your Pixies Riders duo You can also flip the heart over and use it to display your new friends in front of their castle When youre finished playing close up the box and take it with you anywhere you go With 10 Pixies Riders to collect including two special editions each sold separately which mystical pair will be your favorite Each Pixies Riders duo has a unique feature like metallic glowinthedark color change and more Add all of the majestic Pixies Riders to your collection and ride into adventurePIXIES RIDER AND HATCHIMALS GLIDER This Pixie has fluttery wings poseable legs and comes with a matching Glider she can really ride Both have beautiful mystical details and poseable heads\nMYSTERY FEATURES AND MATCHING ACCESSORIES Each duo includes 4 matching accessories and 1 of 10 unique features to discover like fuzzy color change metallic and more What feature will you reveal\nMAGICAL UNBOXING The box becomes a playset and a beautiful display for your Pixies Riders Each box has different artwork to enhance your magical Hatchimals storytelling Collect all 10 castles\nIncludes 1 Hatchimals Pixie 1 Hatchimals Glider 4 Accessories 1 Display Stand 1 Checklist 1 Instruction Sheet 1 Hatchtopia Life App Token',
442
+ 'description\nThe CRIMSON GUARD are the elite shock troops of the COBRA legions All Sieges must hold a degree in either law or accounting and must be in top physical condition Final stages of training take place in the deepest recesses of COBRA Headquarters and are purported to involve an initiation ceremony too hideous to describeBuilding the perfect army just got a little easier Great for the real GI Joe collectorCelebrate 25 years of the ultimate action team with these articulated action figures Figures also comes with interchangeable weaponsGI Joe 25th Anniversary 3 34 Action Figure Collection from Hasbro\nThis Toys R Us exclusive trooper builder set includes 5 Crimson Guard action figures\nFor Ages 5 Up',
443
+ 'description\nFrom the Manufacturer\nA fun and girly way to liven up your wallssatin soft sculpture for wall or ceiling decoration5 color striped satin rainbowSatin soft sculpture for wall or ceiling decoration\n5 color striped satin rainbow\nBlue satin clouds with heart pattern\nSatin flower and heart motifs with Satin ribbons\nCrystal beaded fringe',
444
+ ]
445
+ query_embeddings = model.encode_query(queries)
446
+ document_embeddings = model.encode_document(documents)
447
+ print(query_embeddings.shape, document_embeddings.shape)
448
+ # [1, 768] [3, 768]
449
+
450
+ # Get the similarity scores for the embeddings
451
+ similarities = model.similarity(query_embeddings, document_embeddings)
452
+ print(similarities)
453
+ # tensor([[ 0.6404, -0.0838, 0.1479]])
454
+ ```
455
+
456
+ <!--
457
+ ### Direct Usage (Transformers)
458
+
459
+ <details><summary>Click to see the direct usage in Transformers</summary>
460
+
461
+ </details>
462
+ -->
463
+
464
+ <!--
465
+ ### Downstream Usage (Sentence Transformers)
466
+
467
+ You can finetune this model on your own dataset.
468
+
469
+ <details><summary>Click to expand</summary>
470
+
471
+ </details>
472
+ -->
473
+
474
+ <!--
475
+ ### Out-of-Scope Use
476
+
477
+ *List how the model may foreseeably be misused and address what users ought not to do with the model.*
478
+ -->
479
+
480
+ <!--
481
+ ## Bias, Risks and Limitations
482
+
483
+ *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*
484
+ -->
485
+
486
+ <!--
487
+ ### Recommendations
488
+
489
+ *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*
490
+ -->
491
+
492
+ ## Training Details
493
+
494
+ ### Training Dataset
495
+
496
+ #### Unnamed Dataset
497
+
498
+ * Size: 386,737 training samples
499
+ * Columns: <code>title</code> and <code>description</code>
500
+ * Approximate statistics based on the first 1000 samples:
501
+ | | title | description |
502
+ |:--------|:----------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------|
503
+ | type | string | string |
504
+ | details | <ul><li>min: 8 tokens</li><li>mean: 26.17 tokens</li><li>max: 70 tokens</li></ul> | <ul><li>min: 14 tokens</li><li>mean: 163.64 tokens</li><li>max: 256 tokens</li></ul> |
505
+ * Samples:
506
+ | title | description |
507
+ |:---------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
508
+ | <code>title: <br>Meeruz 5Pcs Magic Butterfly Flying in The Book Fairy Rubber Band Powered Wind Up Butterfly Toy Great Surprise for Wedding and Birthday Gifts</code> | <code>description<br>Wind up this clever little device and hide it inside a book or greeting card When it is set free the butterfly will spin and fly up to 20 feet in the air Amazing These are made adorable stocking stuffers for the kids They are so much fun to use You can place one in a book a card or wherever you please and have it fly out when somebody opens it which is very fun and a cute surpriseMagic flying butterflies small and flexible watch their face as heshe opens the wish card there will be a beautiful butterfly flying out which can gives her eyes a bright and pleasant surprise<br>You can put it in cards books letters you name it<br>Flies a great distance in the air<br>Prank your parents kids colleagues and more<br>NOTE Designs will come at random</code> |
509
+ | <code>title: <br>GDOOL 12010 Wheel Hex Accessories Spare Parts for 116 16890A 16890 16889 16889A RC Cars</code> | <code>description<br>12010 Wheel Hex Accessories Spare Parts for 116 16889 16890 RC CarsOriginal Accessories and Qaulity GuaranteeIt work with any 116 4WD 16889 16890 RC Monster Trucks<br>100 Brand NewIt is perfect for vehicle type cars practical and durable for longterm use<br>Item No12010 Wheel Hex<br>High QualityIt made of High Quality ABS Plastic 100 environmentfriendly and it is a good choice of your truck<br>Good Aftersales service30Day Money Back Guarantee and Ready to respond within 24 hours</code> |
510
+ | <code>title: <br>Turnigy 2700mAh 3S 20C Lipo Pack Suitable for Quanum Nova Phantom QR X350</code> | <code>description<br>Turnigy batteries are known the world over for performance reliability and price Its no surprise to us that Turnigy Lipoly packs are the goto pack for those in the know Turnigy batteries deliver the full rated capacity at a price everyone can afford Turnigy batteries are equipped with heavy duty discharge leads to minimise resistance and sustain high current loads Turnigy batteries stand up to the punishing extremes of aerobatic flight and RC vehicles Each pack is equipped with gold plated connectors and JSTXH style balance connectors All Turnigy Lipoly batteries packs are assembled using IR matched cellsYou wont find a better deal in Lipoly batteries anywhereSpecMinimum Capacity 2700mAhConfiguration 3S1P 111v 3CellConstant Discharge 20CPeak Discharge 10sec 40CPack Weight 205gPack Size 105 x 35 x 29mmCharge Plug JSTXHDischarge plug XT60Note For use with Walkera QR X350 a battery plug adapter is required See related items below</code> |
511
+ * Loss: [<code>CachedMultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cachedmultiplenegativesrankingloss) with these parameters:
512
+ ```json
513
+ {
514
+ "scale": 20.0,
515
+ "similarity_fct": "cos_sim",
516
+ "mini_batch_size": 8,
517
+ "gather_across_devices": false
518
+ }
519
+ ```
520
+
521
+ ### Evaluation Dataset
522
+
523
+ #### Unnamed Dataset
524
+
525
+ * Size: 82,872 evaluation samples
526
+ * Columns: <code>title</code> and <code>description</code>
527
+ * Approximate statistics based on the first 1000 samples:
528
+ | | title | description |
529
+ |:--------|:----------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------|
530
+ | type | string | string |
531
+ | details | <ul><li>min: 9 tokens</li><li>mean: 25.53 tokens</li><li>max: 77 tokens</li></ul> | <ul><li>min: 16 tokens</li><li>mean: 163.11 tokens</li><li>max: 256 tokens</li></ul> |
532
+ * Samples:
533
+ | title | description |
534
+ |:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
535
+ | <code>title: <br>Blue and Black Silver Balloon garland kit 140pcs Royal blue and silver starburst Disco ball balloons for men 30th Birthday Party graduation 80s 90s Disco theme Decorations</code> | <code>description<br> Blue Black silver Balloon Dark Blue Chrome Silver Disco Theme Balloons garland KIT including 130Pcs balloons in 4 sizes 5101218 inchand Large 27inch silver starburst 5pcs 22inch Silver Disco Balloons 4 balloon toolsenough for 1216FT Garland Classics color theme For 80s90sY2k Disco Music theme PartyLets lose yourself to Dance One more timeDisco Blue Black silver Balloon Dark Blue Chrome Silver Disco Theme Balloons garland KIT including 130Pcs balloons in 4 sizes 5101218 inchand Large 27inch silver starburst 5pcs 22inch Silver Disco Balloons 4 balloon toolsenough for 1216FT Garland Classics color theme For 80s90sY2k Disco Music theme PartyLets lose yourself to Dance One more time<br>Kozee Reliable Color We insist on 100 real photography by using Color Correction CardProviding True Color of every single balloonswhat you have to do is trust your color insprition and ideaReliable Consistent Color Balloons to Make your Party Decor Perfect<br>Create your own Themed decorcan be used fo...</code> |
536
+ | <code>title: <br>Brown Resin Hand Painted Basketball Shaped Piggy Bank</code> | <code>description<br>Brown resin hand painted basketball shaped piggy bank Includes gift box 3 12H X 3 34Diameter3 12H X 3 34Diameter<br>basketball shaped piggy bank<br>Burton Burton<br>Sports Champ<br>Resin</code> |
537
+ | <code>title: <br>Epic ProRaw Wired Video AdapterPart 94 Drone Flyer</code> | <code>description<br>This gorgeous DJI osmo proraw wired video adapter part 94 has the finest details and highest quality you will find anywhere DJI osmo proraw wired video adapter part 94 is truly remarkable Product details condition brand newPerfect purchase for any hobby<br>Great craftsmanship<br>Must buy item</code> |
538
+ * Loss: [<code>CachedMultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cachedmultiplenegativesrankingloss) with these parameters:
539
+ ```json
540
+ {
541
+ "scale": 20.0,
542
+ "similarity_fct": "cos_sim",
543
+ "mini_batch_size": 8,
544
+ "gather_across_devices": false
545
+ }
546
+ ```
547
+
548
+ ### Training Hyperparameters
549
+ #### Non-Default Hyperparameters
550
+
551
+ - `per_device_train_batch_size`: 128
552
+ - `per_device_eval_batch_size`: 128
553
+ - `learning_rate`: 2e-05
554
+ - `num_train_epochs`: 1
555
+ - `warmup_ratio`: 0.1
556
+ - `fp16`: True
557
+ - `push_to_hub`: True
558
+ - `hub_model_id`: rabaevn/EncodeRec_Toys
559
+ - `hub_strategy`: checkpoint
560
+ - `prompts`: {'title': 'task: search result | query: ', 'description': 'title: none | text: '}
561
+ - `batch_sampler`: no_duplicates
562
+ - `router_mapping`: {'title': 'query', 'description': 'document'}
563
+
564
+ #### All Hyperparameters
565
+ <details><summary>Click to expand</summary>
566
+
567
+ - `overwrite_output_dir`: False
568
+ - `do_predict`: False
569
+ - `eval_strategy`: no
570
+ - `prediction_loss_only`: True
571
+ - `per_device_train_batch_size`: 128
572
+ - `per_device_eval_batch_size`: 128
573
+ - `per_gpu_train_batch_size`: None
574
+ - `per_gpu_eval_batch_size`: None
575
+ - `gradient_accumulation_steps`: 1
576
+ - `eval_accumulation_steps`: None
577
+ - `torch_empty_cache_steps`: None
578
+ - `learning_rate`: 2e-05
579
+ - `weight_decay`: 0.0
580
+ - `adam_beta1`: 0.9
581
+ - `adam_beta2`: 0.999
582
+ - `adam_epsilon`: 1e-08
583
+ - `max_grad_norm`: 1.0
584
+ - `num_train_epochs`: 1
585
+ - `max_steps`: -1
586
+ - `lr_scheduler_type`: linear
587
+ - `lr_scheduler_kwargs`: {}
588
+ - `warmup_ratio`: 0.1
589
+ - `warmup_steps`: 0
590
+ - `log_level`: passive
591
+ - `log_level_replica`: warning
592
+ - `log_on_each_node`: True
593
+ - `logging_nan_inf_filter`: True
594
+ - `save_safetensors`: True
595
+ - `save_on_each_node`: False
596
+ - `save_only_model`: False
597
+ - `restore_callback_states_from_checkpoint`: False
598
+ - `no_cuda`: False
599
+ - `use_cpu`: False
600
+ - `use_mps_device`: False
601
+ - `seed`: 42
602
+ - `data_seed`: None
603
+ - `jit_mode_eval`: False
604
+ - `use_ipex`: False
605
+ - `bf16`: False
606
+ - `fp16`: True
607
+ - `fp16_opt_level`: O1
608
+ - `half_precision_backend`: auto
609
+ - `bf16_full_eval`: False
610
+ - `fp16_full_eval`: False
611
+ - `tf32`: None
612
+ - `local_rank`: 0
613
+ - `ddp_backend`: None
614
+ - `tpu_num_cores`: None
615
+ - `tpu_metrics_debug`: False
616
+ - `debug`: []
617
+ - `dataloader_drop_last`: False
618
+ - `dataloader_num_workers`: 0
619
+ - `dataloader_prefetch_factor`: None
620
+ - `past_index`: -1
621
+ - `disable_tqdm`: False
622
+ - `remove_unused_columns`: True
623
+ - `label_names`: None
624
+ - `load_best_model_at_end`: False
625
+ - `ignore_data_skip`: False
626
+ - `fsdp`: []
627
+ - `fsdp_min_num_params`: 0
628
+ - `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
629
+ - `fsdp_transformer_layer_cls_to_wrap`: None
630
+ - `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
631
+ - `deepspeed`: None
632
+ - `label_smoothing_factor`: 0.0
633
+ - `optim`: adamw_torch_fused
634
+ - `optim_args`: None
635
+ - `adafactor`: False
636
+ - `group_by_length`: False
637
+ - `length_column_name`: length
638
+ - `ddp_find_unused_parameters`: None
639
+ - `ddp_bucket_cap_mb`: None
640
+ - `ddp_broadcast_buffers`: False
641
+ - `dataloader_pin_memory`: True
642
+ - `dataloader_persistent_workers`: False
643
+ - `skip_memory_metrics`: True
644
+ - `use_legacy_prediction_loop`: False
645
+ - `push_to_hub`: True
646
+ - `resume_from_checkpoint`: None
647
+ - `hub_model_id`: rabaevn/EncodeRec_Toys
648
+ - `hub_strategy`: checkpoint
649
+ - `hub_private_repo`: None
650
+ - `hub_always_push`: False
651
+ - `hub_revision`: None
652
+ - `gradient_checkpointing`: False
653
+ - `gradient_checkpointing_kwargs`: None
654
+ - `include_inputs_for_metrics`: False
655
+ - `include_for_metrics`: []
656
+ - `eval_do_concat_batches`: True
657
+ - `fp16_backend`: auto
658
+ - `push_to_hub_model_id`: None
659
+ - `push_to_hub_organization`: None
660
+ - `mp_parameters`:
661
+ - `auto_find_batch_size`: False
662
+ - `full_determinism`: False
663
+ - `torchdynamo`: None
664
+ - `ray_scope`: last
665
+ - `ddp_timeout`: 1800
666
+ - `torch_compile`: False
667
+ - `torch_compile_backend`: None
668
+ - `torch_compile_mode`: None
669
+ - `include_tokens_per_second`: False
670
+ - `include_num_input_tokens_seen`: False
671
+ - `neftune_noise_alpha`: None
672
+ - `optim_target_modules`: None
673
+ - `batch_eval_metrics`: False
674
+ - `eval_on_start`: False
675
+ - `use_liger_kernel`: False
676
+ - `liger_kernel_config`: None
677
+ - `eval_use_gather_object`: False
678
+ - `average_tokens_across_devices`: False
679
+ - `prompts`: {'title': 'task: search result | query: ', 'description': 'title: none | text: '}
680
+ - `batch_sampler`: no_duplicates
681
+ - `multi_dataset_batch_sampler`: proportional
682
+ - `router_mapping`: {'title': 'query', 'description': 'document'}
683
+ - `learning_rate_mapping`: {}
684
+
685
+ </details>
686
+
687
+ ### Training Logs
688
+ | Epoch | Step | Training Loss |
689
+ |:------:|:----:|:-------------:|
690
+ | 0.0165 | 50 | 0.411 |
691
+ | 0.0331 | 100 | 0.1448 |
692
+ | 0.0496 | 150 | 0.1364 |
693
+ | 0.0662 | 200 | 0.1064 |
694
+
695
+
696
+ ### Framework Versions
697
+ - Python: 3.12.7
698
+ - Sentence Transformers: 5.1.0
699
+ - Transformers: 4.55.2
700
+ - PyTorch: 2.8.0+cu126
701
+ - Accelerate: 1.10.0
702
+ - Datasets: 4.1.1
703
+ - Tokenizers: 0.21.4
704
+
705
+ ## Citation
706
+
707
+ ### BibTeX
708
+
709
+ #### Sentence Transformers
710
+ ```bibtex
711
+ @inproceedings{reimers-2019-sentence-bert,
712
+ title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
713
+ author = "Reimers, Nils and Gurevych, Iryna",
714
+ booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
715
+ month = "11",
716
+ year = "2019",
717
+ publisher = "Association for Computational Linguistics",
718
+ url = "https://arxiv.org/abs/1908.10084",
719
+ }
720
+ ```
721
+
722
+ #### CachedMultipleNegativesRankingLoss
723
+ ```bibtex
724
+ @misc{gao2021scaling,
725
+ title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
726
+ author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
727
+ year={2021},
728
+ eprint={2101.06983},
729
+ archivePrefix={arXiv},
730
+ primaryClass={cs.LG}
731
+ }
732
+ ```
733
+
734
+ <!--
735
+ ## Glossary
736
+
737
+ *Clearly define terms in order to be accessible across audiences.*
738
+ -->
739
+
740
+ <!--
741
+ ## Model Card Authors
742
+
743
+ *Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*
744
+ -->
745
+
746
+ <!--
747
+ ## Model Card Contact
748
+
749
+ *Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*
750
+ -->
last-checkpoint/config.json ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_sliding_window_pattern": 6,
3
+ "architectures": [
4
+ "Gemma3TextModel"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "attn_logit_softcapping": null,
9
+ "bos_token_id": 2,
10
+ "dtype": "float32",
11
+ "eos_token_id": 1,
12
+ "final_logit_softcapping": null,
13
+ "head_dim": 256,
14
+ "hidden_activation": "gelu_pytorch_tanh",
15
+ "hidden_size": 768,
16
+ "initializer_range": 0.02,
17
+ "intermediate_size": 1152,
18
+ "layer_types": [
19
+ "sliding_attention",
20
+ "sliding_attention",
21
+ "sliding_attention",
22
+ "sliding_attention",
23
+ "sliding_attention",
24
+ "full_attention",
25
+ "sliding_attention",
26
+ "sliding_attention",
27
+ "sliding_attention",
28
+ "sliding_attention",
29
+ "sliding_attention",
30
+ "full_attention",
31
+ "sliding_attention",
32
+ "sliding_attention",
33
+ "sliding_attention",
34
+ "sliding_attention",
35
+ "sliding_attention",
36
+ "full_attention",
37
+ "sliding_attention",
38
+ "sliding_attention",
39
+ "sliding_attention",
40
+ "sliding_attention",
41
+ "sliding_attention",
42
+ "full_attention"
43
+ ],
44
+ "max_position_embeddings": 2048,
45
+ "model_type": "gemma3_text",
46
+ "num_attention_heads": 3,
47
+ "num_hidden_layers": 24,
48
+ "num_key_value_heads": 1,
49
+ "pad_token_id": 0,
50
+ "query_pre_attn_scalar": 256,
51
+ "rms_norm_eps": 1e-06,
52
+ "rope_local_base_freq": 10000.0,
53
+ "rope_scaling": null,
54
+ "rope_theta": 1000000.0,
55
+ "sliding_window": 512,
56
+ "torch_dtype": "float32",
57
+ "transformers_version": "4.55.2",
58
+ "use_bidirectional_attention": true,
59
+ "use_cache": true,
60
+ "vocab_size": 262144
61
+ }
last-checkpoint/config_sentence_transformers.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "SentenceTransformer",
3
+ "__version__": {
4
+ "sentence_transformers": "5.1.0",
5
+ "transformers": "4.55.2",
6
+ "pytorch": "2.8.0+cu126"
7
+ },
8
+ "prompts": {
9
+ "query": "task: search result | query: ",
10
+ "document": "title: none | text: ",
11
+ "BitextMining": "task: search result | query: ",
12
+ "Clustering": "task: clustering | query: ",
13
+ "Classification": "task: classification | query: ",
14
+ "InstructionRetrieval": "task: code retrieval | query: ",
15
+ "MultilabelClassification": "task: classification | query: ",
16
+ "PairClassification": "task: sentence similarity | query: ",
17
+ "Reranking": "task: search result | query: ",
18
+ "Retrieval": "task: search result | query: ",
19
+ "Retrieval-query": "task: search result | query: ",
20
+ "Retrieval-document": "title: none | text: ",
21
+ "STS": "task: sentence similarity | query: ",
22
+ "Summarization": "task: summarization | query: "
23
+ },
24
+ "default_prompt_name": null,
25
+ "similarity_fn_name": "cosine"
26
+ }
last-checkpoint/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43fd0b9b8721c9ea306dd8f4764df5603c03aa7978abc57d73a44d69f7b42300
3
+ size 1211486072
last-checkpoint/modules.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "idx": 0,
4
+ "name": "0",
5
+ "path": "",
6
+ "type": "sentence_transformers.models.Transformer"
7
+ },
8
+ {
9
+ "idx": 1,
10
+ "name": "1",
11
+ "path": "1_Pooling",
12
+ "type": "sentence_transformers.models.Pooling"
13
+ },
14
+ {
15
+ "idx": 2,
16
+ "name": "2",
17
+ "path": "2_Dense",
18
+ "type": "sentence_transformers.models.Dense"
19
+ },
20
+ {
21
+ "idx": 3,
22
+ "name": "3",
23
+ "path": "3_Dense",
24
+ "type": "sentence_transformers.models.Dense"
25
+ },
26
+ {
27
+ "idx": 4,
28
+ "name": "4",
29
+ "path": "4_Normalize",
30
+ "type": "sentence_transformers.models.Normalize"
31
+ }
32
+ ]
last-checkpoint/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e061afeaae568c474e3f579c8acb359c0b7dd5ce97fa97f1910796df7e71b90
3
+ size 2460923467
last-checkpoint/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef60e3e5f5a51d7d33cfe02db83de9c51e04f4df7d88af6871672fd4589ce3bc
3
+ size 14645
last-checkpoint/scaler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0cd605bcdfda1a9d9eac4f3ea7ab051df8ad1e55668c146cc899ab908c9d1ebe
3
+ size 1383
last-checkpoint/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fc057fbca03ad393f5dd382b07edb53abde6442bd92793df58cb201522d6453
3
+ size 1465
last-checkpoint/sentence_bert_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "max_seq_length": 256,
3
+ "do_lower_case": false
4
+ }
last-checkpoint/special_tokens_map.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "boi_token": "<start_of_image>",
3
+ "bos_token": {
4
+ "content": "<bos>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false
9
+ },
10
+ "eoi_token": "<end_of_image>",
11
+ "eos_token": {
12
+ "content": "<eos>",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false
17
+ },
18
+ "image_token": "<image_soft_token>",
19
+ "pad_token": {
20
+ "content": "<pad>",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false
25
+ },
26
+ "unk_token": {
27
+ "content": "<unk>",
28
+ "lstrip": false,
29
+ "normalized": false,
30
+ "rstrip": false,
31
+ "single_word": false
32
+ }
33
+ }
last-checkpoint/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2039575615d76b80e31a8358daccc19a0052a448e08259bd7e039cb2232e33d
3
+ size 33385261
last-checkpoint/tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff
 
last-checkpoint/trainer_state.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.06618133686300463,
6
+ "eval_steps": 500,
7
+ "global_step": 200,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.01654533421575116,
14
+ "grad_norm": 8.903882026672363,
15
+ "learning_rate": 3.2343234323432342e-06,
16
+ "loss": 0.411,
17
+ "step": 50
18
+ },
19
+ {
20
+ "epoch": 0.03309066843150232,
21
+ "grad_norm": 14.360084533691406,
22
+ "learning_rate": 6.534653465346535e-06,
23
+ "loss": 0.1448,
24
+ "step": 100
25
+ },
26
+ {
27
+ "epoch": 0.04963600264725347,
28
+ "grad_norm": 15.013335227966309,
29
+ "learning_rate": 9.834983498349836e-06,
30
+ "loss": 0.1364,
31
+ "step": 150
32
+ },
33
+ {
34
+ "epoch": 0.06618133686300463,
35
+ "grad_norm": 8.873584747314453,
36
+ "learning_rate": 1.3135313531353136e-05,
37
+ "loss": 0.1064,
38
+ "step": 200
39
+ }
40
+ ],
41
+ "logging_steps": 50,
42
+ "max_steps": 3022,
43
+ "num_input_tokens_seen": 0,
44
+ "num_train_epochs": 1,
45
+ "save_steps": 200,
46
+ "stateful_callbacks": {
47
+ "TrainerControl": {
48
+ "args": {
49
+ "should_epoch_stop": false,
50
+ "should_evaluate": false,
51
+ "should_log": false,
52
+ "should_save": true,
53
+ "should_training_stop": false
54
+ },
55
+ "attributes": {}
56
+ }
57
+ },
58
+ "total_flos": 0.0,
59
+ "train_batch_size": 128,
60
+ "trial_name": null,
61
+ "trial_params": null
62
+ }
last-checkpoint/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a39f24ed7b2e6658a68bef8ea47abb8ca62bb850caad070b292c2d5c537faf8
3
+ size 6289