austinpatrickm commited on
Commit
7eab079
·
verified ·
1 Parent(s): 0325fda

Upload 10 files

Browse files
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip 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
 
 
33
  *.zip 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
1_Pooling/config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "word_embedding_dimension": 384,
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
+ }
README.md ADDED
@@ -0,0 +1,898 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - sentence-transformers
4
+ - sentence-similarity
5
+ - feature-extraction
6
+ - dense
7
+ - generated_from_trainer
8
+ - dataset_size:29840
9
+ - loss:MultipleNegativesRankingLoss
10
+ base_model: intfloat/multilingual-e5-small
11
+ widget:
12
+ - source_sentence: Can I use Fruity Bass Boost on sounds other than bass and kick
13
+ drums?
14
+ sentences:
15
+ - 'Document_title: Locating FL Studio Installation Files
16
+
17
+ File_name: app_flstudioinstallationfiles.htm
18
+
19
+ Heading_hierarchy: [Locating FL Studio Installation Files -> Multiple Installations]
20
+
21
+ Anchor_id: [none]
22
+
23
+ To avoid installing a Beta, or any new version, over an existing installation,
24
+ you can rename the existing FL Studio application prior to installation of the
25
+ beta or new release. • Close FL Studio. • Open Finder and select the '' Applications
26
+ '' folder in the Sidebar. • Locate and right-click the '' FL
27
+
28
+ Studio N '' application. Where N is the version number. • Select '' Rename ''.
29
+ • Rename the application to something other than '' FL Studio N '', such as ''FL
30
+ Studio N Stable''. • Install the beta, or new release, as normal. It will become
31
+ the default FL Studio installation. • To maintain quick access to
32
+
33
+ the various installations, make Alias shortcuts . • Open Finder and select the
34
+ '' Applications '' folder in the Sidebar. • Right-click the application you would
35
+ like to create a shortcut for and select '' Make alias ''. • Relocate the alias
36
+ to your Desktop. Alternatively, you can drag the application from
37
+
38
+ the Applications directory to your Dock to create a shortcut.'
39
+ - "Document_title: Fruity Bass Boost\nFile_name: plugins/Fruity Bass Boost.htm\n\
40
+ Heading_hierarchy: [Fruity Bass Boost]\nAnchor_id: [none]\nFruity Bass Boost is\
41
+ \ a bass enhancing effect (specialized EQ). You may need to adjust the track volume\
42
+ \ to avoid clipping since this FX works by boosting frequencies. Try it on bass\n\
43
+ \ and kick drum sounds. Yes it's just a fancy 'bass' knob :) [Image: Knob\
44
+ \ controls with frequency display | Ref: img_plug/Fruity_bass_boost.png]"
45
+ - "Document_title: \n Image-Line Remote \nFile_name: il_remote.htm\n\
46
+ Heading_hierarchy: [Image-Line Remote -> Layouts]\nAnchor_id: [none]\nA Layout\
47
+ \ (*.ilr format) is a collection of tabs in an IL\n Remote project.\
48
+ \ The Layout menu is located to the top-left of the\n display while\
49
+ \ in Edit mode. Presets (*.ilrp format) are a\n collection of controls\
50
+ \ saved in a [Container](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/il_remote.htm#IL_Remote_Container)\
51
+ \ that can\n be loaded on a tab. This is the fastest way to build new\
52
+ \ Layouts. [Image: Document icon with layout options menu | Ref: img_shot/ILRemote_menu_layoutsettings.jpg]"
53
+ - source_sentence: How can I tell if my Windows is 32-bit or 64-bit?
54
+ sentences:
55
+ - 'Title: How do I remove (uninstall) an Image-Line application or plugin in macOS
56
+ / OS X?
57
+
58
+ Answer: The FL Studio manual explains this procedure under the section [Locate
59
+ the FL Studio Installation](https://www.image-line.com/support/flstudio_online_manual/html/app_flstudioinstallationfiles.htm)
60
+ for Windows and macOS. Image-Line Support Team'
61
+ - "Document_title: Fruity Vocoder\nFile_name: plugins/Fruity Vocoder.htm\nHeading_hierarchy:\
62
+ \ [Fruity Vocoder -> Setup]\nAnchor_id: [none]\n• Load: an instance of Fruity\
63
+ \ Vocoder in a mixer track effect slot (don't use the master). • Routing: By default,\
64
+ \ the left channel is the modulator and the right channel is the carrier. This\
65
+ \ means the vocoder can only process a mono input sound, although the output is\
66
+ \ stereo.\n The modulator/carrier channel assignment can be swapped from\
67
+ \ the MIX section. There are several ways to create a modulator/carrier combination\
68
+ \ in the L/R channels of a Mixer Track: ➤ Use panning - Route the carrier and\
69
+ \ source [Instrument\nChannels](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/channelrack.htm)\
70
+ \ to the Mixer track containing the Fruity Vocoder and pan the [Channel Panning](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/channelrack.htm)\
71
+ \ knob for the\nmodulator source hard-left and carrier source hard-right. ➤ Single\
72
+ \ channel muting - Use the [Fruity Mute 2](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/Fruity%20Mute%202.htm)\
73
+ \ plugin to mute the L and R channels of a 2nd and 3rd mixer track carrying\n\
74
+ the modulator and carrier respectively then route these to the mixer track with\
75
+ \ Fruity Vocoder. • Carrier sound - The most effective carriers (the sound you\
76
+ \ hear vocoded) are those covering the whole frequency range. Pad sounds often\
77
+ \ make good carriers, you can also try mixing in\n some white noise if you\
78
+ \ need to hear speech (s & t sounds) vocoded more clearly. Remember, the vocoder\
79
+ \ acts like a series of gates allowing through those frequencies it detects in\
80
+ \ the modulator. If a frequency\n gate opens and there is no carrier sound\
81
+ \ in that band, that part of the 'tonal' character in the vocoded sound will be\
82
+ \ lost when compared to the original. • Modulator sound - Using some compression\
83
+ \ on the modulator usually helps sound (and speech in particular) to vocode more\
84
+ \ clearly."
85
+ - 'To find out if your computer is running a 32-bit or 64-bit version of Windows
86
+ in Windows 8 from the metro (tiled) interface type ''Windows Version'' and enter.
87
+ Click ''See if you have a 32-bit or 64-bit version of Windows. In Vista or 7:
88
+ Open System by clicking the Start button, right-clicking Computer, and
89
+
90
+ then clicking Properties. Under System, you can view the system type. If your
91
+ computer is running Windows XP, do the following: Click Start. Right-click My
92
+ Computer, and then click Properties. If you don''t see "x64 Edition" listed, then
93
+ you''re running the 32-bit version of Windows XP. We recommend a minimum of 4
94
+ GB of RAM , preferably more. 8 to 16 GB is usually enough, very few projects will
95
+ use more than 16 GB. You may also like to check the '' [What computer should I
96
+ get for music making](https://support.image-line.com/action/knowledgebase/?ans=214)
97
+ '' article. Why does the
98
+
99
+ 64 Bit version install into Program Files (x86)? - Both the 32 and 64 Bit versions
100
+ of FL Studio are installed to C:\Program Files (x86)\Image-Line\FL Studio. FL
101
+ Studio 32 Bit = FL.exe and FL Studio 64 Bit = FL64.exe. This really, absolutely,
102
+ doesn''t matter, but people seem to fret about it, so let''s just keep this between
103
+ us and not start a global panic, OK? Program Files (x86) and Program Files(x64)
104
+ folders are not really important, you can point installers anywhere. A future
105
+ release will probably install to the standard Windows locations. We are doing
106
+ this for two reasons 1. To minimize the
107
+
108
+ FL Studio footprint when both versions are installed as they share a lot of content
109
+ which takes up disk space. It''s an additional 140 MB when combined vs an additional
110
+ 950 MB if installed separately. 2. For the FL Studio 12 series of releases we
111
+ wanted people to have both 32 and 64 Bit version installed, for technical support
112
+ & testing purposes. About FL Studio 32 vs 64 Bit Should I use FL Studio 32 or
113
+ 64 Bit? - The answer depends mainly on your 3rd party VST library . 32 Bit plugins
114
+ work best in FL Studio 32 Bit and 64 Bit plugins work best in FL Studio 64 Bit.
115
+ Because of this, we
116
+
117
+ recommend you load projects made in FL Studio 32 Bit and 64 Bit in their respective
118
+ native versions. Loading a 32 Bit project in FL Studio 64 Bit (and vice versa)
119
+ before you have updated your plugin library to 64 Bit will result in many bridged
120
+ VST plugins. Bridging uses more CPU and can reduce stability of FL Studio when
121
+ plugins behave badly (if they do, please report it to [Techsupport](https://forum.image-line.com/viewforum.php?f=200)
122
+ ). If you still use lots of 32 Bit plugins, perhaps 64 Bit versions are not available,
123
+ then we recommend you use FL Studio 32 Bit with those plugins.
124
+
125
+ Remember, you can use both FL Studio 32 and 64 Bit depending on your project and the plugins it uses. If you want to make a permanent switch to FL Studio 64 Bit you should update ALL your 3rd party VST plugins to 64 Bit format and make sure you use 64 Bit plugins moving forward with new projects. What''s the difference between FL Studio 32 and 64 Bit? – It''s all about the maximum memory available to FL Studio. 32 Bit applications can access a maximum of 4 GB while 64 Bit applications are limited to [8
126
+
127
+ TB](https://docs.microsoft.com/nl-nl/windows/win32/memory/memory-limits-for-windows-releases?redirectedfrom=MSDN)
128
+ . However, FL Studio 32 Bit includes several [memory management tools](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/app_opt.htm)
129
+ that allow it to bypass the 32 Bit 4 GB limitation by loading samples and VST
130
+ plugins in separate processes that don’t add to FL Studios total memory count.
131
+ So does that mean there is no difference?'
132
+ - source_sentence: Is there any reason to still use the 32 Bit version if I have a
133
+ 64 Bit Windows system?
134
+ sentences:
135
+ - "Document_title: MIDI Setup\nFile_name: app_wiz5.htm\nHeading_hierarchy: [MIDI\
136
+ \ Setup -> Input - Connecting External Controller/s]\nAnchor_id: [none]\nIf you\
137
+ \ have a [USB/MIDI keyboard or controller](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/basics_controllers.htm)\
138
+ \ that you would like to control using FL Studio, you will need to change some\
139
+ \ settings in the Input section. This contains a list of controller devices\n\
140
+ that have been detected by Windows (see section 1). To activate a controller in\
141
+ \ FL Studio - • Open FL Studio and make sure the ' Enable MIDI remote control\
142
+ \ ' option is selected in the [Options menu](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/menu_options.htm)\n\
143
+ , MIDI input is disabled otherwise. • Click on the device in the Input list, so\
144
+ \ that it is highlighted. If your device is listed in the Controller type drop-down\
145
+ \ menu, select it there also. If your controller does not appear by name this\
146
+ \ \n means it does not have custom drivers. If so, use the ' USB Audio Device\
147
+ \ (generic controller) '. If you do not see any options, you will need to re/install\
148
+ \ the appropriate drivers for your controller. • With the controller highlighted\
149
+ \ select the Enable switch below the Input list. Each device\nselected in the\
150
+ \ Input list must be selected/deselected separately. • If successful the MIDI\
151
+ \ activity light on the [Main Panel](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/toolbar_panels.htm#panel_hint)\
152
+ \ will blink each time you play notes or move knobs on your\ncontroller. NOTE:\
153
+ \ If the ' [Debug log](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/envsettings_debug.htm)\
154
+ \ ' is open any MIDI data received by FL Studio is displayed and logged , useful\
155
+ \ for learning the MIDI CC numbers of controls or\n troubleshooting connections.\
156
+ \ [Image: Top window menu bar, MIDI icon | Ref: img_shot/midi_activity.png] NOTE:\
157
+ \ For more detailed information on the rest of the settings in this section and\
158
+ \ further details on supported controllers go to the [MIDI\nSettings](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/envsettings_midi.htm)\
159
+ \ page."
160
+ - 'What''s the difference between FL Studio 32 and 64 Bit? – It''s all about the maximum memory available to FL Studio. 32 Bit applications can access a maximum of 4 GB while 64 Bit applications are limited to [8
161
+
162
+ TB](https://docs.microsoft.com/nl-nl/windows/win32/memory/memory-limits-for-windows-releases?redirectedfrom=MSDN)
163
+ . However, FL Studio 32 Bit includes several [memory management tools](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/app_opt.htm)
164
+ that allow it to bypass the 32 Bit 4 GB limitation by loading samples and VST
165
+ plugins in separate processes that don’t add to FL Studios total memory count.
166
+ So does that mean there is no difference? No, while FL Studio 32 Bit can access
167
+ as much memory as a 64 Bit application, the core FL Studio 32
168
+
169
+ Bit process is still limited to 4 GB, and so out-of-memory errors can occur when
170
+ editing very long audio files in Edison, or when the memory management techniques
171
+ are not used. This won''t happen with FL Studio 64 Bit. Additionally, FL Studio
172
+ 64 Bit also includes a complete update of most plugins to native 64 Bit format.
173
+ Finally FL Studio 64 Bit can run 64 Bit VST plugins without ''bridging'' them
174
+ and this saves CPU relative to FL Studio 32 Bit running 64 Bit plugins. Are FL
175
+ Studio 32 and 64 Bit projects compatible? - Yes. You can load projects made with
176
+ FL Studio 32 Bit in FL Studio 64
177
+
178
+ Bit and vice versa. For example, when you load a FL Studio 32 Bit project in FL
179
+ Studio 64 Bit, all plugins are automatically loaded with 64 Bit versions, so long
180
+ as the 64 Bit version exists and has the same name (for VSTs this may not always
181
+ be the case, but you can usually rename the .dll). When loading 64 Bit projects
182
+ in the FL Studio 32 Bit the reverse is also true. If the bit-equivalent plugin
183
+ is not available, then the bit-original plugin is loaded and bridged. Where a
184
+ project contains plugins mentioned in the ''Limitations'' section, you will get
185
+ a missing plugin error message, but the
186
+
187
+ project will still load. NOTE: Loading older 32 Bit projects in FL Studio 64 Bit,
188
+ where 64 Bit versions of your plugins are not available, will result in lots of
189
+ ''bridged'' plugins. This can result in higher than normal CPU usage. In some
190
+ cases this can also cause stability problems - if this happens please report it
191
+ to [Techsupport](https://forum.image-line.com/viewforum.php?f=200) . How many
192
+ plugins can I bridge? - Bridging plugins uses about 2% CPU extra per-plugin. So
193
+ 2 or 3 x 64 Bit plugins in a 32 Bit project, where it makes sense, is no problem
194
+ at all. But using 20 x 64 Bit plugins in a 32
195
+
196
+ Bit project or 20 x 32 Bit plugins in a 64 Bit project is just going to wasting
197
+ CPU and can also compromise program stability. Only bridge plugins where the 32/64
198
+ Bit version is not available OR in FL Studio 32 Bit where you want to bridge a
199
+ Sampler plugin (Kontakt for example) to free up memory. Some ''common sense''
200
+ and a little planning goes a long way to creating a smooth experience. I have
201
+ Windows 64 Bit should I use FL Studio 64 Bit? - Windows 64 can run both 32 and
202
+ 64 Bit applications. There is hardly any reason anymore to use a 32 bit version
203
+ of FL Studio in a 64 bit
204
+
205
+ Windows. If you still have a couple of 32 bit plugins in your library, bridging
206
+ them is probably going to work just fine. Windows 32 Bit can''t run 64 Bit applications,
207
+ so in that case you must use 32 bit FL Studio 20.8.4, or earlier. Can I use 32
208
+ Bit plugins in FL Studio 64 Bit - Yes you can. You can also use 64 Bit plugins
209
+ in FL Studio 32 Bit. This is possible with a ''bit bridge''. This is code that
210
+ allows 32 and 64 Bit applications to share data. There is about 2% CPU hit per
211
+ plugin when bridging, so we recommend minimizing the number of bridged plugins
212
+ in your projects. As you can see, a few
213
+
214
+ bridged plugins won''t matter but bridging 20 or 30 plugins will waste a lot of
215
+ CPU.'
216
+ - 'The slow cache file isn''t needed. WAV files, maximum size - FL Studio currently
217
+ supports WAV files up to 2 GB in size. The WAV format itself has a size limitation
218
+ of 4 GB due to its use of a 32-bit unsigned integer for the file size header.
219
+ To accommodate larger files, we now support the RF64 format, which addresses this
220
+ limitation and allows for files larger than 4 GB. Future updates may include additional
221
+ format support to further enhance file handling capabilities. Why is
222
+
223
+ this so complicated? - It''s actually not so complex. If you switch to FL Studio
224
+ 64 Bit then you should also update your VST library to 64 Bit. If you are working
225
+ on older projects made in FL Studio 32 Bit, it’s probably a good idea to use the
226
+ 32 Bit version of FL Studio. Generally, try to minimise the number of bridged
227
+ plugins regardless of the FL Studio version. Problems and troubleshooting How
228
+ do I associate FL Studio projects with the 32 or 64 Bit version - Open FL Studio
229
+ 32 or 64 Bit, depending on what version you want to use as the default with projects,
230
+ and select Options > General
231
+
232
+ Options > Associate project files with application Audio files (wav) are missing - It''s likely the 64 Bit [Ogg codec](https://xiph.org/vorbis/) has failed to install. Some of the samples are compressed in Ogg format, so if they can''t be decompressed they will be listed as missing. Solution: Browse to ..\Program Files (x86)\Image-Line\FL Studio\System\Installers\VorbisACM_x64 where you will find ''install.bat''. Right-click it and ''Run as administrator''. That should reinstall the 64 Bit ogg codec. My 32/64 Bit bridged plugin is crashing - Please report it to [Techsupport
233
+
234
+ here](https://forum.image-line.com/viewforum.php?f=200) , we always fix reproducible
235
+ bugs with the 32 and 64 Bit bridges. BUT we can''t fix what we don''t know about,
236
+ so please report it. If a bridged plugin crashes you may see errors associated
237
+ with ilbridge.exe rather than the plugin itself. Example errors are Exceptions,
238
+ Access Violations etc. When I edit the controls of some VST plugins FL Studio
239
+ stops playback - This can happen when you edit the controls on a bridged plugin.
240
+ Since the bridged plugin is a separate process (program) ''Auto close'' kicks
241
+ in and FL Studio releases
242
+
243
+ control over the audio interface to the plugin and playback stops. You have two
244
+ options, 1. Disable '' [Auto close device](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/envsettings_audio.htm)
245
+ '' on the Audio settings OR 2. Make sure there are no bridged plugins being used
246
+ in your project (update the plugin/s in use to match the version of FL Studio
247
+ you are using). After closing FL Studio I see instances of ''ilbridge (32 Bit)''
248
+ in the processes list - This is the FL Studio bridge. There is one instance per
249
+ bridged plugin. They should all disappear when FL Studio is
250
+
251
+ closed. If these remain AFTER FL Studio is closed you can kill them (its a bug).
252
+ Make sure you update to the latest version, [here](https://support.image-line.com/redirect/FLDemoDownload)
253
+ . If you already are, please report it to [Techsupport](https://forum.image-line.com/viewforum.php?f=200)
254
+ . I can''t run my old 32 Bit projects without CPU overload, skipping or underruns
255
+ - It''s likely that: 1. Plugins - FL Studio 64 Bit is loading your old project
256
+ and can''t find 64 Bit equivalents of the plugins. This means 32 Bit plugins are
257
+ bridged to 64 Bit mode and this uses more CPU than 32 Bit
258
+
259
+ plugin on FL Studio 32 Bit or 64 Bit plugins on FL Studio 64 Bit. It''s about 2% extra per plugin, so a couple won''t matter but 10+ definitely will.'
260
+ - source_sentence: How do I start recording from the transport panel, and what does
261
+ arming the record button do?
262
+ sentences:
263
+ - "Document_title: VFX Sequencer\nFile_name: plugins/VFX Sequencer_modified2.htm\n\
264
+ Heading_hierarchy: [VFX Sequencer -> Note Select]\nAnchor_id: [none]\n• N1 to\
265
+ \ N5 - These map to the first 5 notes in the input chord. From the lowest N1 to\
266
+ \ the highest N5 note. Up to 5 note chords may be sequenced, although when Chord\
267
+ \ is selected in Step Type , more than 5 notes can sound. If you play more than\
268
+ \ 5 note chords in 'Normal' mode, then only the\nlowest 5 notes will be used.\
269
+ \ • Root (Root) - By default, N1 sets the root note of the Chord played. The difference\
270
+ \ is N1 transposition can be changed under the [Arp\nEngine](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/VFX%20Sequencer_modified2.htm#vfx_sequencer_arpengine)\
271
+ \ settings. \nFor the (Root), regardless of what [Order algorithm](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/VFX%20Sequencer_modified2.htm#vfx_sequencer_orderalgo)\
272
+ \ is selected, it always returns the root note of the detected chord (if any).\
273
+ \ Also the (Root) note\nis not limited by the [Input range wrap]. • Scale (Fixed)\
274
+ \ - Uses a defined fixed note which can be scaled to the Scale / Key Signature\
275
+ \ set under the [Arp\nEngine](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/VFX%20Sequencer_modified2.htm#vfx_sequencer_arpengine)\
276
+ \ settings."
277
+ - "Document_title: PoiZone V2 \nFile_name: plugins/PoiZone.htm\nHeading_hierarchy:\
278
+ \ [PoiZone V2 -> Voicing]\nAnchor_id: [none]\n• 2 main oscillators for subtractive\
279
+ \ synthesis: SAW and PULSE shapes, pulse width adjustable. • 1 NOISE Oscillator.\
280
+ \ • Variable polyphony (1 to 32 voices)."
281
+ - 'Document_title: Toolbar Panels & Icons
282
+
283
+ File_name: toolbar_panels.htm
284
+
285
+ Heading_hierarchy: [Toolbar Panels & Icons -> Transport & Recording Controls]
286
+
287
+ Anchor_id: [none]
288
+
289
+ The Transport Panel contains controls for playing, recording and setting song
290
+ tempo: [Image: Green SONG mode button, transport controls | Ref: img_shot/panel_transport.png]
291
+ • Pat/Song mode ( L ) - Switch between pattern and song mode. In Pattern Mode
292
+ , only the current pattern is played.
293
+
294
+ In Song Mode the whole [Playlist](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/playlist.htm)
295
+ arrangement is played. Right-Clicking this control toggles the Playlist/Channel
296
+ Rack window. For more on working with Patterns and how they are used, see the
297
+ section in
298
+
299
+ the Channel Rack about [managing patterns](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/channelrack.htm#channelrack_patternselector)
300
+ and [Arranging
301
+
302
+ Workflow](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/basics_workflow.htm#Workflow_arranging)
303
+ . NOTE: You can ( Right-Click ) PAT to open the [Channel Rack](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/channelrack.htm)
304
+ and SO
305
+
306
+ NG to open the [Playlist](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/playlist.htm)
307
+ . [Image: Dark gray right-pointing triangle icon | Ref: img_glob/flicon_playpause.png]
308
+ Play/Pause ( Space / Ctrl + Space ) - Press the button once to play , again to
309
+ stop and rewind .
310
+
311
+ To Play/Stop use ( Space ). To pause use ( Ctrl + Space ). When the record button
312
+ is armed, pressing Play starts the recording process. macOS: To avoid a conflicts
313
+ with the macOS '' Spotlight '' command, ( command + Space ) change or disable
314
+ it from the System Preferences > Spotlight >
315
+
316
+ Keyboard Shortcuts section. NOTE: There are also play buttons on the top bar of
317
+ the [Piano roll](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/pianoroll.htm)
318
+ , [Channel Rack / Step
319
+
320
+ Sequencer](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/channelrack.htm)
321
+ and [Playlist](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/playlist.htm)
322
+ to aid workflow. [Image: Black square shape on gray background | Ref:
323
+
324
+ img_glob/flicon_stop.png] Stop ( Space ) - Stops playback or recording. Double-click
325
+ to stop all notes (panic mode). To pause playback use ( Ctrl + Space ). Right-Click
326
+ the Stop button to show the start location option : [Image: Orange PAT button,
327
+ tooltip shown | Ref:
328
+
329
+ img_shot/panel_transport_stop.png] ➤ Remember seek time - After pressing stop
330
+ or ( Space ), Play will start from the selected position (click on the Playlist
331
+ time-line to select). [Image: Circular knob control with pointer | Ref: img_glob/flicon_record.png]
332
+ Record ( R ) - Switches between record
333
+
334
+ and play mode. You can automate the button with Start/Stop recording [Time Markers](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/playlist.htm#Playlist_TimeMarkers)
335
+ . [Image: Red circular recording button with settings panel | Ref:
336
+
337
+ img_shot/panel_transport_recfilter.png] NOTE: ( Alt + Click ) the Record Button
338
+ to show the Recording Options popup: ''What do you want to record'', when it has
339
+ been disabled. Right-Click the Record button to show the Recording Filter options
340
+ (all items with a tick will be recorded, Left-click to
341
+
342
+ select items): ➤ Automation - [Record movements](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/recording_automation.htm)
343
+ of knobs, sliders and switches by the mouse or from [controllers](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/htm
344
+
345
+ l/basics_controllers.htm) . Turn this off if you want to adjust mixer track volumes,
346
+ generator settings, etc. while recording external audio and you don''t want those
347
+ movements recorded. ➤ Notes - [Record note
348
+
349
+ data](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/recording_scores.htm)
350
+ from [keyboards](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/basics_controllers.htm)
351
+ and drum-pads, etc. Turn this off if you want to accompany a vocalist,
352
+
353
+ for example, but don''t want to record your keyboard performance.'
354
+ - source_sentence: Can I automate the Gain image over time to create dynamic volume
355
+ changes?
356
+ sentences:
357
+ - "Document_title: PoiZone V2 \nFile_name: plugins/PoiZone.htm\nHeading_hierarchy:\
358
+ \ [PoiZone V2 -> Voicing]\nAnchor_id: [none]\n• 2 main oscillators for subtractive\
359
+ \ synthesis: SAW and PULSE shapes, pulse width adjustable. • 1 NOISE Oscillator.\
360
+ \ • Variable polyphony (1 to 32 voices)."
361
+ - "Document_title: Linking & Using External Hardware Controllers \nFile_name:\
362
+ \ automation_linking.htm\nHeading_hierarchy: [Linking & Using External Hardware\
363
+ \ Controllers -> Delete or backup global links]\nAnchor_id: [none]\nThe simplest\
364
+ \ method of correcting a few links is to reassign them. However, if you want to\
365
+ \ delete or edit global links , the link data files are stored in the FL Studio\
366
+ \ [User data folder](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/envsettings_files.htm#userdata)\
367
+ \ : ➤ .\n..\\FL Studio\\Settings\\Mapping\\Generic\\ You will find sub-folders\
368
+ \ for each controller used, (generic controller) being the default. If deleting,\
369
+ \ you have the option of deleting the individual ' BY TARGET.flmapping ' files\
370
+ \ (deleting recommended) OR editing out the links you don't want in a text editor.\n\
371
+ Each link is bound by the <link> and </link> identifiers."
372
+ - "Document_title: Harmor \nFile_name: plugins/Harmor.htm\nHeading_hierarchy: [Harmor\
373
+ \ -> About images and planes]\nAnchor_id: [none]\nThere are independent images\
374
+ \ that control the Pitch/Frequency and Gain of partials. Together these can create\
375
+ \ any sound, just as sampler can. In the image window the vertical dimension\n\
376
+ \ is frequency (each line of pixels is a single partial), while the horizontal\
377
+ \ dimension is time."
378
+ pipeline_tag: sentence-similarity
379
+ library_name: sentence-transformers
380
+ metrics:
381
+ - cosine_accuracy@1
382
+ - cosine_accuracy@3
383
+ - cosine_accuracy@5
384
+ - cosine_accuracy@10
385
+ - cosine_precision@1
386
+ - cosine_precision@3
387
+ - cosine_precision@5
388
+ - cosine_precision@10
389
+ - cosine_recall@1
390
+ - cosine_recall@3
391
+ - cosine_recall@5
392
+ - cosine_recall@10
393
+ - cosine_ndcg@10
394
+ - cosine_mrr@10
395
+ - cosine_map@100
396
+ model-index:
397
+ - name: SentenceTransformer based on intfloat/multilingual-e5-small
398
+ results:
399
+ - task:
400
+ type: information-retrieval
401
+ name: Information Retrieval
402
+ dataset:
403
+ name: Unknown
404
+ type: unknown
405
+ metrics:
406
+ - type: cosine_accuracy@1
407
+ value: 0.6911409395973155
408
+ name: Cosine Accuracy@1
409
+ - type: cosine_accuracy@3
410
+ value: 0.8779865771812081
411
+ name: Cosine Accuracy@3
412
+ - type: cosine_accuracy@5
413
+ value: 0.925234899328859
414
+ name: Cosine Accuracy@5
415
+ - type: cosine_accuracy@10
416
+ value: 0.9642953020134228
417
+ name: Cosine Accuracy@10
418
+ - type: cosine_precision@1
419
+ value: 0.6911409395973155
420
+ name: Cosine Precision@1
421
+ - type: cosine_precision@3
422
+ value: 0.292662192393736
423
+ name: Cosine Precision@3
424
+ - type: cosine_precision@5
425
+ value: 0.1850469798657718
426
+ name: Cosine Precision@5
427
+ - type: cosine_precision@10
428
+ value: 0.09642953020134229
429
+ name: Cosine Precision@10
430
+ - type: cosine_recall@1
431
+ value: 0.6911409395973155
432
+ name: Cosine Recall@1
433
+ - type: cosine_recall@3
434
+ value: 0.8779865771812081
435
+ name: Cosine Recall@3
436
+ - type: cosine_recall@5
437
+ value: 0.925234899328859
438
+ name: Cosine Recall@5
439
+ - type: cosine_recall@10
440
+ value: 0.9642953020134228
441
+ name: Cosine Recall@10
442
+ - type: cosine_ndcg@10
443
+ value: 0.8339510205941362
444
+ name: Cosine Ndcg@10
445
+ - type: cosine_mrr@10
446
+ value: 0.7913869180781895
447
+ name: Cosine Mrr@10
448
+ - type: cosine_map@100
449
+ value: 0.7931169353960464
450
+ name: Cosine Map@100
451
+ ---
452
+
453
+ # SentenceTransformer based on intfloat/multilingual-e5-small
454
+
455
+ This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [intfloat/multilingual-e5-small](https://huggingface.co/intfloat/multilingual-e5-small). It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
456
+
457
+ ## Model Details
458
+
459
+ ### Model Description
460
+ - **Model Type:** Sentence Transformer
461
+ - **Base model:** [intfloat/multilingual-e5-small](https://huggingface.co/intfloat/multilingual-e5-small) <!-- at revision c007d7ef6fd86656326059b28395a7a03a7c5846 -->
462
+ - **Maximum Sequence Length:** 512 tokens
463
+ - **Output Dimensionality:** 384 dimensions
464
+ - **Similarity Function:** Cosine Similarity
465
+ <!-- - **Training Dataset:** Unknown -->
466
+ <!-- - **Language:** Unknown -->
467
+ <!-- - **License:** Unknown -->
468
+
469
+ ### Model Sources
470
+
471
+ - **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
472
+ - **Repository:** [Sentence Transformers on GitHub](https://github.com/huggingface/sentence-transformers)
473
+ - **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)
474
+
475
+ ### Full Model Architecture
476
+
477
+ ```
478
+ SentenceTransformer(
479
+ (0): Transformer({'max_seq_length': 512, 'do_lower_case': False, 'architecture': 'BertModel'})
480
+ (1): Pooling({'word_embedding_dimension': 384, '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})
481
+ (2): Normalize()
482
+ )
483
+ ```
484
+
485
+ ## Usage
486
+
487
+ ### Direct Usage (Sentence Transformers)
488
+
489
+ First install the Sentence Transformers library:
490
+
491
+ ```bash
492
+ pip install -U sentence-transformers
493
+ ```
494
+
495
+ Then you can load this model and run inference.
496
+ ```python
497
+ from sentence_transformers import SentenceTransformer
498
+
499
+ # Download from the 🤗 Hub
500
+ model = SentenceTransformer("sentence_transformers_model_id")
501
+ # Run inference
502
+ sentences = [
503
+ 'Can I automate the Gain image over time to create dynamic volume changes?',
504
+ 'Document_title: Harmor \nFile_name: plugins/Harmor.htm\nHeading_hierarchy: [Harmor -> About images and planes]\nAnchor_id: [none]\nThere are independent images that control the Pitch/Frequency and Gain of partials. Together these can create any sound, just as sampler can. In the image window the vertical dimension\n is frequency (each line of pixels is a single partial), while the horizontal dimension is time.',
505
+ 'Document_title: PoiZone V2 \nFile_name: plugins/PoiZone.htm\nHeading_hierarchy: [PoiZone V2 -> Voicing]\nAnchor_id: [none]\n• 2 main oscillators for subtractive synthesis: SAW and PULSE shapes, pulse width adjustable. • 1 NOISE Oscillator. • Variable polyphony (1 to 32 voices).',
506
+ ]
507
+ embeddings = model.encode(sentences)
508
+ print(embeddings.shape)
509
+ # [3, 384]
510
+
511
+ # Get the similarity scores for the embeddings
512
+ similarities = model.similarity(embeddings, embeddings)
513
+ print(similarities)
514
+ # tensor([[1.0000, 0.4301, 0.1394],
515
+ # [0.4301, 1.0000, 0.1051],
516
+ # [0.1394, 0.1051, 1.0000]])
517
+ ```
518
+
519
+ <!--
520
+ ### Direct Usage (Transformers)
521
+
522
+ <details><summary>Click to see the direct usage in Transformers</summary>
523
+
524
+ </details>
525
+ -->
526
+
527
+ <!--
528
+ ### Downstream Usage (Sentence Transformers)
529
+
530
+ You can finetune this model on your own dataset.
531
+
532
+ <details><summary>Click to expand</summary>
533
+
534
+ </details>
535
+ -->
536
+
537
+ <!--
538
+ ### Out-of-Scope Use
539
+
540
+ *List how the model may foreseeably be misused and address what users ought not to do with the model.*
541
+ -->
542
+
543
+ ## Evaluation
544
+
545
+ ### Metrics
546
+
547
+ #### Information Retrieval
548
+
549
+ * Evaluated with [<code>InformationRetrievalEvaluator</code>](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator)
550
+
551
+ | Metric | Value |
552
+ |:--------------------|:----------|
553
+ | cosine_accuracy@1 | 0.6911 |
554
+ | cosine_accuracy@3 | 0.878 |
555
+ | cosine_accuracy@5 | 0.9252 |
556
+ | cosine_accuracy@10 | 0.9643 |
557
+ | cosine_precision@1 | 0.6911 |
558
+ | cosine_precision@3 | 0.2927 |
559
+ | cosine_precision@5 | 0.185 |
560
+ | cosine_precision@10 | 0.0964 |
561
+ | cosine_recall@1 | 0.6911 |
562
+ | cosine_recall@3 | 0.878 |
563
+ | cosine_recall@5 | 0.9252 |
564
+ | cosine_recall@10 | 0.9643 |
565
+ | **cosine_ndcg@10** | **0.834** |
566
+ | cosine_mrr@10 | 0.7914 |
567
+ | cosine_map@100 | 0.7931 |
568
+
569
+ <!--
570
+ ## Bias, Risks and Limitations
571
+
572
+ *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*
573
+ -->
574
+
575
+ <!--
576
+ ### Recommendations
577
+
578
+ *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*
579
+ -->
580
+
581
+ ## Training Details
582
+
583
+ ### Training Dataset
584
+
585
+ #### Unnamed Dataset
586
+
587
+ * Size: 29,840 training samples
588
+ * Columns: <code>sentence_0</code> and <code>sentence_1</code>
589
+ * Approximate statistics based on the first 1000 samples:
590
+ | | sentence_0 | sentence_1 |
591
+ |:--------|:----------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------|
592
+ | type | string | string |
593
+ | details | <ul><li>min: 9 tokens</li><li>mean: 19.39 tokens</li><li>max: 34 tokens</li></ul> | <ul><li>min: 29 tokens</li><li>mean: 324.18 tokens</li><li>max: 512 tokens</li></ul> |
594
+ * Samples:
595
+ | sentence_0 | sentence_1 |
596
+ |:------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
597
+ | <code>How do I load a *.SPEECH preset into the Sampler plugin?</code> | <code>Document_title: Speech Preset (*.SPEECH)<br>File_name: fformats_sample_speech.htm<br>Heading_hierarchy: [Speech Preset (*.SPEECH)]<br>Anchor_id: [none]<br>The Speech synthesizer processes text to create computerized or Vocoder-like vocals to your projects. The *.SPEECH presets are supported by all native FL Studio plugins that use custom samples for<br> synthesizing, i.e. [Sampler](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/chansettings_sampler.htm) , [Granulizer](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/Fruity%20Granulizer.htm) , [Fruity<br>Slicer](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/Fruity%20Slicer.htm) and [Fruity Scratcher](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/Fruity%20Scratcher.htm) .</code> |
598
+ | <code>Can I use Speech presets with the Fruity Granulizer to create vocal effects?</code> | <code>Document_title: Speech Preset (*.SPEECH)<br>File_name: fformats_sample_speech.htm<br>Heading_hierarchy: [Speech Preset (*.SPEECH)]<br>Anchor_id: [none]<br>The Speech synthesizer processes text to create computerized or Vocoder-like vocals to your projects. The *.SPEECH presets are supported by all native FL Studio plugins that use custom samples for<br> synthesizing, i.e. [Sampler](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/chansettings_sampler.htm) , [Granulizer](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/Fruity%20Granulizer.htm) , [Fruity<br>Slicer](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/Fruity%20Slicer.htm) and [Fruity Scratcher](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/Fruity%20Scratcher.htm) .</code> |
599
+ | <code>What kind of vocals can the Speech synthesizer create from text?</code> | <code>Document_title: Speech Preset (*.SPEECH)<br>File_name: fformats_sample_speech.htm<br>Heading_hierarchy: [Speech Preset (*.SPEECH)]<br>Anchor_id: [none]<br>The Speech synthesizer processes text to create computerized or Vocoder-like vocals to your projects. The *.SPEECH presets are supported by all native FL Studio plugins that use custom samples for<br> synthesizing, i.e. [Sampler](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/chansettings_sampler.htm) , [Granulizer](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/Fruity%20Granulizer.htm) , [Fruity<br>Slicer](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/Fruity%20Slicer.htm) and [Fruity Scratcher](https://www.image-line.com/fl-studio-learning/fl-studio-online-manual/html/plugins/Fruity%20Scratcher.htm) .</code> |
600
+ * Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
601
+ ```json
602
+ {
603
+ "scale": 20.0,
604
+ "similarity_fct": "cos_sim",
605
+ "gather_across_devices": false
606
+ }
607
+ ```
608
+
609
+ ### Training Hyperparameters
610
+ #### Non-Default Hyperparameters
611
+
612
+ - `eval_strategy`: steps
613
+ - `per_device_train_batch_size`: 10
614
+ - `per_device_eval_batch_size`: 10
615
+ - `num_train_epochs`: 2
616
+ - `multi_dataset_batch_sampler`: round_robin
617
+
618
+ #### All Hyperparameters
619
+ <details><summary>Click to expand</summary>
620
+
621
+ - `do_predict`: False
622
+ - `eval_strategy`: steps
623
+ - `prediction_loss_only`: True
624
+ - `per_device_train_batch_size`: 10
625
+ - `per_device_eval_batch_size`: 10
626
+ - `gradient_accumulation_steps`: 1
627
+ - `eval_accumulation_steps`: None
628
+ - `torch_empty_cache_steps`: None
629
+ - `learning_rate`: 5e-05
630
+ - `weight_decay`: 0.0
631
+ - `adam_beta1`: 0.9
632
+ - `adam_beta2`: 0.999
633
+ - `adam_epsilon`: 1e-08
634
+ - `max_grad_norm`: 1
635
+ - `num_train_epochs`: 2
636
+ - `max_steps`: -1
637
+ - `lr_scheduler_type`: linear
638
+ - `lr_scheduler_kwargs`: None
639
+ - `warmup_ratio`: None
640
+ - `warmup_steps`: 0
641
+ - `log_level`: passive
642
+ - `log_level_replica`: warning
643
+ - `log_on_each_node`: True
644
+ - `logging_nan_inf_filter`: True
645
+ - `enable_jit_checkpoint`: False
646
+ - `save_on_each_node`: False
647
+ - `save_only_model`: False
648
+ - `restore_callback_states_from_checkpoint`: False
649
+ - `use_cpu`: False
650
+ - `seed`: 42
651
+ - `data_seed`: None
652
+ - `bf16`: False
653
+ - `fp16`: False
654
+ - `bf16_full_eval`: False
655
+ - `fp16_full_eval`: False
656
+ - `tf32`: None
657
+ - `local_rank`: -1
658
+ - `ddp_backend`: None
659
+ - `debug`: []
660
+ - `dataloader_drop_last`: False
661
+ - `dataloader_num_workers`: 0
662
+ - `dataloader_prefetch_factor`: None
663
+ - `disable_tqdm`: False
664
+ - `remove_unused_columns`: True
665
+ - `label_names`: None
666
+ - `load_best_model_at_end`: False
667
+ - `ignore_data_skip`: False
668
+ - `fsdp`: []
669
+ - `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
670
+ - `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
671
+ - `parallelism_config`: None
672
+ - `deepspeed`: None
673
+ - `label_smoothing_factor`: 0.0
674
+ - `optim`: adamw_torch_fused
675
+ - `optim_args`: None
676
+ - `group_by_length`: False
677
+ - `length_column_name`: length
678
+ - `project`: huggingface
679
+ - `trackio_space_id`: trackio
680
+ - `ddp_find_unused_parameters`: None
681
+ - `ddp_bucket_cap_mb`: None
682
+ - `ddp_broadcast_buffers`: False
683
+ - `dataloader_pin_memory`: True
684
+ - `dataloader_persistent_workers`: False
685
+ - `skip_memory_metrics`: True
686
+ - `push_to_hub`: False
687
+ - `resume_from_checkpoint`: None
688
+ - `hub_model_id`: None
689
+ - `hub_strategy`: every_save
690
+ - `hub_private_repo`: None
691
+ - `hub_always_push`: False
692
+ - `hub_revision`: None
693
+ - `gradient_checkpointing`: False
694
+ - `gradient_checkpointing_kwargs`: None
695
+ - `include_for_metrics`: []
696
+ - `eval_do_concat_batches`: True
697
+ - `auto_find_batch_size`: False
698
+ - `full_determinism`: False
699
+ - `ddp_timeout`: 1800
700
+ - `torch_compile`: False
701
+ - `torch_compile_backend`: None
702
+ - `torch_compile_mode`: None
703
+ - `include_num_input_tokens_seen`: no
704
+ - `neftune_noise_alpha`: None
705
+ - `optim_target_modules`: None
706
+ - `batch_eval_metrics`: False
707
+ - `eval_on_start`: False
708
+ - `use_liger_kernel`: False
709
+ - `liger_kernel_config`: None
710
+ - `eval_use_gather_object`: False
711
+ - `average_tokens_across_devices`: True
712
+ - `use_cache`: False
713
+ - `prompts`: None
714
+ - `batch_sampler`: batch_sampler
715
+ - `multi_dataset_batch_sampler`: round_robin
716
+ - `router_mapping`: {}
717
+ - `learning_rate_mapping`: {}
718
+
719
+ </details>
720
+
721
+ ### Training Logs
722
+ <details><summary>Click to expand</summary>
723
+
724
+ | Epoch | Step | Training Loss | cosine_ndcg@10 |
725
+ |:------:|:----:|:-------------:|:--------------:|
726
+ | 0.0168 | 50 | - | 0.7103 |
727
+ | 0.0335 | 100 | - | 0.7267 |
728
+ | 0.0503 | 150 | - | 0.7500 |
729
+ | 0.0670 | 200 | - | 0.7715 |
730
+ | 0.0838 | 250 | - | 0.7892 |
731
+ | 0.1005 | 300 | - | 0.7921 |
732
+ | 0.1173 | 350 | - | 0.7940 |
733
+ | 0.1340 | 400 | - | 0.7958 |
734
+ | 0.1508 | 450 | - | 0.7889 |
735
+ | 0.1676 | 500 | 0.3978 | 0.7999 |
736
+ | 0.1843 | 550 | - | 0.7861 |
737
+ | 0.2011 | 600 | - | 0.7848 |
738
+ | 0.2178 | 650 | - | 0.7780 |
739
+ | 0.2346 | 700 | - | 0.7885 |
740
+ | 0.2513 | 750 | - | 0.7926 |
741
+ | 0.2681 | 800 | - | 0.7914 |
742
+ | 0.2849 | 850 | - | 0.8043 |
743
+ | 0.3016 | 900 | - | 0.7939 |
744
+ | 0.3184 | 950 | - | 0.8057 |
745
+ | 0.3351 | 1000 | 0.1115 | 0.8093 |
746
+ | 0.3519 | 1050 | - | 0.8056 |
747
+ | 0.3686 | 1100 | - | 0.7941 |
748
+ | 0.3854 | 1150 | - | 0.8042 |
749
+ | 0.4021 | 1200 | - | 0.8007 |
750
+ | 0.4189 | 1250 | - | 0.8071 |
751
+ | 0.4357 | 1300 | - | 0.8121 |
752
+ | 0.4524 | 1350 | - | 0.8037 |
753
+ | 0.4692 | 1400 | - | 0.7958 |
754
+ | 0.4859 | 1450 | - | 0.8052 |
755
+ | 0.5027 | 1500 | 0.0989 | 0.8028 |
756
+ | 0.5194 | 1550 | - | 0.7989 |
757
+ | 0.5362 | 1600 | - | 0.8078 |
758
+ | 0.5529 | 1650 | - | 0.8117 |
759
+ | 0.5697 | 1700 | - | 0.8108 |
760
+ | 0.5865 | 1750 | - | 0.8101 |
761
+ | 0.6032 | 1800 | - | 0.8102 |
762
+ | 0.6200 | 1850 | - | 0.8080 |
763
+ | 0.6367 | 1900 | - | 0.8150 |
764
+ | 0.6535 | 1950 | - | 0.8156 |
765
+ | 0.6702 | 2000 | 0.0901 | 0.8138 |
766
+ | 0.6870 | 2050 | - | 0.8127 |
767
+ | 0.7038 | 2100 | - | 0.8123 |
768
+ | 0.7205 | 2150 | - | 0.8128 |
769
+ | 0.7373 | 2200 | - | 0.8141 |
770
+ | 0.7540 | 2250 | - | 0.8108 |
771
+ | 0.7708 | 2300 | - | 0.8108 |
772
+ | 0.7875 | 2350 | - | 0.8164 |
773
+ | 0.8043 | 2400 | - | 0.8159 |
774
+ | 0.8210 | 2450 | - | 0.8175 |
775
+ | 0.8378 | 2500 | 0.0908 | 0.8206 |
776
+ | 0.8546 | 2550 | - | 0.8223 |
777
+ | 0.8713 | 2600 | - | 0.8238 |
778
+ | 0.8881 | 2650 | - | 0.8264 |
779
+ | 0.9048 | 2700 | - | 0.8212 |
780
+ | 0.9216 | 2750 | - | 0.8204 |
781
+ | 0.9383 | 2800 | - | 0.8236 |
782
+ | 0.9551 | 2850 | - | 0.8170 |
783
+ | 0.9718 | 2900 | - | 0.8217 |
784
+ | 0.9886 | 2950 | - | 0.8246 |
785
+ | 1.0 | 2984 | - | 0.8222 |
786
+ | 1.0054 | 3000 | 0.0868 | 0.8207 |
787
+ | 1.0221 | 3050 | - | 0.8173 |
788
+ | 1.0389 | 3100 | - | 0.8165 |
789
+ | 1.0556 | 3150 | - | 0.8211 |
790
+ | 1.0724 | 3200 | - | 0.8236 |
791
+ | 1.0891 | 3250 | - | 0.8207 |
792
+ | 1.1059 | 3300 | - | 0.8173 |
793
+ | 1.1227 | 3350 | - | 0.8197 |
794
+ | 1.1394 | 3400 | - | 0.8164 |
795
+ | 1.1562 | 3450 | - | 0.8212 |
796
+ | 1.1729 | 3500 | 0.0611 | 0.8225 |
797
+ | 1.1897 | 3550 | - | 0.8250 |
798
+ | 1.2064 | 3600 | - | 0.8256 |
799
+ | 1.2232 | 3650 | - | 0.8253 |
800
+ | 1.2399 | 3700 | - | 0.8254 |
801
+ | 1.2567 | 3750 | - | 0.8254 |
802
+ | 1.2735 | 3800 | - | 0.8284 |
803
+ | 1.2902 | 3850 | - | 0.8324 |
804
+ | 1.3070 | 3900 | - | 0.8311 |
805
+ | 1.3237 | 3950 | - | 0.8272 |
806
+ | 1.3405 | 4000 | 0.0581 | 0.8245 |
807
+ | 1.3572 | 4050 | - | 0.8227 |
808
+ | 1.3740 | 4100 | - | 0.8235 |
809
+ | 1.3908 | 4150 | - | 0.8211 |
810
+ | 1.4075 | 4200 | - | 0.8199 |
811
+ | 1.4243 | 4250 | - | 0.8230 |
812
+ | 1.4410 | 4300 | - | 0.8248 |
813
+ | 1.4578 | 4350 | - | 0.8266 |
814
+ | 1.4745 | 4400 | - | 0.8268 |
815
+ | 1.4913 | 4450 | - | 0.8273 |
816
+ | 1.5080 | 4500 | 0.0499 | 0.8305 |
817
+ | 1.5248 | 4550 | - | 0.8293 |
818
+ | 1.5416 | 4600 | - | 0.8291 |
819
+ | 1.5583 | 4650 | - | 0.8287 |
820
+ | 1.5751 | 4700 | - | 0.8285 |
821
+ | 1.5918 | 4750 | - | 0.8286 |
822
+ | 1.6086 | 4800 | - | 0.8289 |
823
+ | 1.6253 | 4850 | - | 0.8277 |
824
+ | 1.6421 | 4900 | - | 0.8283 |
825
+ | 1.6588 | 4950 | - | 0.8287 |
826
+ | 1.6756 | 5000 | 0.0595 | 0.8285 |
827
+ | 1.6924 | 5050 | - | 0.8289 |
828
+ | 1.7091 | 5100 | - | 0.8274 |
829
+ | 1.7259 | 5150 | - | 0.8277 |
830
+ | 1.7426 | 5200 | - | 0.8296 |
831
+ | 1.7594 | 5250 | - | 0.8326 |
832
+ | 1.7761 | 5300 | - | 0.8323 |
833
+ | 1.7929 | 5350 | - | 0.8308 |
834
+ | 1.8097 | 5400 | - | 0.8312 |
835
+ | 1.8264 | 5450 | - | 0.8314 |
836
+ | 1.8432 | 5500 | 0.0544 | 0.8328 |
837
+ | 1.8599 | 5550 | - | 0.8331 |
838
+ | 1.8767 | 5600 | - | 0.8327 |
839
+ | 1.8934 | 5650 | - | 0.8335 |
840
+ | 1.9102 | 5700 | - | 0.8340 |
841
+
842
+ </details>
843
+
844
+ ### Framework Versions
845
+ - Python: 3.12.12
846
+ - Sentence Transformers: 5.2.3
847
+ - Transformers: 5.0.0
848
+ - PyTorch: 2.10.0+cu128
849
+ - Accelerate: 1.12.0
850
+ - Datasets: 4.0.0
851
+ - Tokenizers: 0.22.2
852
+
853
+ ## Citation
854
+
855
+ ### BibTeX
856
+
857
+ #### Sentence Transformers
858
+ ```bibtex
859
+ @inproceedings{reimers-2019-sentence-bert,
860
+ title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
861
+ author = "Reimers, Nils and Gurevych, Iryna",
862
+ booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
863
+ month = "11",
864
+ year = "2019",
865
+ publisher = "Association for Computational Linguistics",
866
+ url = "https://arxiv.org/abs/1908.10084",
867
+ }
868
+ ```
869
+
870
+ #### MultipleNegativesRankingLoss
871
+ ```bibtex
872
+ @misc{henderson2017efficient,
873
+ title={Efficient Natural Language Response Suggestion for Smart Reply},
874
+ author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
875
+ year={2017},
876
+ eprint={1705.00652},
877
+ archivePrefix={arXiv},
878
+ primaryClass={cs.CL}
879
+ }
880
+ ```
881
+
882
+ <!--
883
+ ## Glossary
884
+
885
+ *Clearly define terms in order to be accessible across audiences.*
886
+ -->
887
+
888
+ <!--
889
+ ## Model Card Authors
890
+
891
+ *Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*
892
+ -->
893
+
894
+ <!--
895
+ ## Model Card Contact
896
+
897
+ *Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*
898
+ -->
config.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_cross_attention": false,
3
+ "architectures": [
4
+ "BertModel"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.1,
7
+ "bos_token_id": null,
8
+ "classifier_dropout": null,
9
+ "dtype": "float32",
10
+ "eos_token_id": null,
11
+ "hidden_act": "gelu",
12
+ "hidden_dropout_prob": 0.1,
13
+ "hidden_size": 384,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 1536,
16
+ "is_decoder": false,
17
+ "layer_norm_eps": 1e-12,
18
+ "max_position_embeddings": 512,
19
+ "model_type": "bert",
20
+ "num_attention_heads": 12,
21
+ "num_hidden_layers": 12,
22
+ "pad_token_id": 0,
23
+ "position_embedding_type": "absolute",
24
+ "tie_word_embeddings": true,
25
+ "tokenizer_class": "XLMRobertaTokenizer",
26
+ "transformers_version": "5.0.0",
27
+ "type_vocab_size": 2,
28
+ "use_cache": true,
29
+ "vocab_size": 250037
30
+ }
config_sentence_transformers.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "SentenceTransformer",
3
+ "__version__": {
4
+ "sentence_transformers": "5.2.3",
5
+ "transformers": "5.0.0",
6
+ "pytorch": "2.10.0+cu128"
7
+ },
8
+ "prompts": {
9
+ "query": "",
10
+ "document": ""
11
+ },
12
+ "default_prompt_name": null,
13
+ "similarity_fn_name": "cosine"
14
+ }
eval/Information-Retrieval_evaluation_results.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ epoch,steps,cosine-Accuracy@1,cosine-Accuracy@3,cosine-Accuracy@5,cosine-Accuracy@10,cosine-Precision@1,cosine-Recall@1,cosine-Precision@3,cosine-Recall@3,cosine-Precision@5,cosine-Recall@5,cosine-Precision@10,cosine-Recall@10,cosine-MRR@10,cosine-NDCG@10,cosine-MAP@100
2
+ 1.0,2984,0.6734228187919463,0.8653691275167785,0.914496644295302,0.9585234899328859,0.6734228187919463,0.6734228187919463,0.2884563758389262,0.8653691275167785,0.18289932885906038,0.914496644295302,0.09585234899328861,0.9585234899328859,0.7777948758921878,0.8222036005852663,0.779801669956141
3
+ 2.0,5968,0.690738255033557,0.8787919463087248,0.9259060402684564,0.9633557046979866,0.690738255033557,0.690738255033557,0.2929306487695749,0.8787919463087248,0.18518120805369126,0.9259060402684564,0.09633557046979867,0.9633557046979866,0.7912152977522063,0.8336339862772744,0.7930272778163093
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:275315338ccd49f13ccbb653ed9fde3b2bcb97a2cc48cad4e9e443fce8721c1e
3
+ size 470637392
modules.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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_Normalize",
18
+ "type": "sentence_transformers.models.Normalize"
19
+ }
20
+ ]
sentence_bert_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "max_seq_length": 512,
3
+ "do_lower_case": false
4
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef04f2b385d1514f500e779207ace0f53e30895ce37563179e29f4022d28ca38
3
+ size 17083053
tokenizer_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "bos_token": "<s>",
4
+ "clean_up_tokenization_spaces": true,
5
+ "cls_token": "<s>",
6
+ "eos_token": "</s>",
7
+ "is_local": false,
8
+ "mask_token": "<mask>",
9
+ "model_max_length": 512,
10
+ "pad_token": "<pad>",
11
+ "sep_token": "</s>",
12
+ "sp_model_kwargs": {},
13
+ "tokenizer_class": "TokenizersBackend",
14
+ "unk_token": "<unk>"
15
+ }