spaces demo

LICENSE

@@ -0,0 +1,674 @@
+                    GNU GENERAL PUBLIC LICENSE
+                       Version 3, 29 June 2007
+
+ Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+                            Preamble
+
+  The GNU General Public License is a free, copyleft license for
+software and other kinds of works.
+
+  The licenses for most software and other practical works are designed
+to take away your freedom to share and change the works.  By contrast,
+the GNU General Public License is intended to guarantee your freedom to
+share and change all versions of a program--to make sure it remains free
+software for all its users.  We, the Free Software Foundation, use the
+GNU General Public License for most of our software; it applies also to
+any other work released this way by its authors.  You can apply it to
+your programs, too.
+
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+them if you wish), that you receive source code or can get it if you
+want it, that you can change the software or use pieces of it in new
+free programs, and that you know you can do these things.
+
+  To protect your rights, we need to prevent others from denying you
+these rights or asking you to surrender the rights.  Therefore, you have
+certain responsibilities if you distribute copies of the software, or if
+you modify it: responsibilities to respect the freedom of others.
+
+  For example, if you distribute copies of such a program, whether
+gratis or for a fee, you must pass on to the recipients the same
+freedoms that you received.  You must make sure that they, too, receive
+or can get the source code.  And you must show them these terms so they
+know their rights.
+
+  Developers that use the GNU GPL protect your rights with two steps:
+(1) assert copyright on the software, and (2) offer you this License
+giving you legal permission to copy, distribute and/or modify it.
+
+  For the developers' and authors' protection, the GPL clearly explains
+that there is no warranty for this free software.  For both users' and
+authors' sake, the GPL requires that modified versions be marked as
+changed, so that their problems will not be attributed erroneously to
+authors of previous versions.
+
+  Some devices are designed to deny users access to install or run
+modified versions of the software inside them, although the manufacturer
+can do so.  This is fundamentally incompatible with the aim of
+protecting users' freedom to change the software.  The systematic
+pattern of such abuse occurs in the area of products for individuals to
+use, which is precisely where it is most unacceptable.  Therefore, we
+have designed this version of the GPL to prohibit the practice for those
+products.  If such problems arise substantially in other domains, we
+stand ready to extend this provision to those domains in future versions
+of the GPL, as needed to protect the freedom of users.
+
+  Finally, every program is threatened constantly by software patents.
+States should not allow patents to restrict development and use of
+software on general-purpose computers, but in those that do, we wish to
+avoid the special danger that patents applied to a free program could
+make it effectively proprietary.  To prevent this, the GPL assures that
+patents cannot be used to render the program non-free.
+
+  The precise terms and conditions for copying, distribution and
+modification follow.
+
+                       TERMS AND CONDITIONS
+
+  0. Definitions.
+
+  "This License" refers to version 3 of the GNU General Public License.
+
+  "Copyright" also means copyright-like laws that apply to other kinds of
+works, such as semiconductor masks.
+
+  "The Program" refers to any copyrightable work licensed under this
+License.  Each licensee is addressed as "you".  "Licensees" and
+"recipients" may be individuals or organizations.
+
+  To "modify" a work means to copy from or adapt all or part of the work
+in a fashion requiring copyright permission, other than the making of an
+exact copy.  The resulting work is called a "modified version" of the
+earlier work or a work "based on" the earlier work.
+
+  A "covered work" means either the unmodified Program or a work based
+on the Program.
+
+  To "propagate" a work means to do anything with it that, without
+permission, would make you directly or secondarily liable for
+infringement under applicable copyright law, except executing it on a
+computer or modifying a private copy.  Propagation includes copying,
+distribution (with or without modification), making available to the
+public, and in some countries other activities as well.
+
+  To "convey" a work means any kind of propagation that enables other
+parties to make or receive copies.  Mere interaction with a user through
+a computer network, with no transfer of a copy, is not conveying.
+
+  An interactive user interface displays "Appropriate Legal Notices"
+to the extent that it includes a convenient and prominently visible
+feature that (1) displays an appropriate copyright notice, and (2)
+tells the user that there is no warranty for the work (except to the
+extent that warranties are provided), that licensees may convey the
+work under this License, and how to view a copy of this License.  If
+the interface presents a list of user commands or options, such as a
+menu, a prominent item in the list meets this criterion.
+
+  1. Source Code.
+
+  The "source code" for a work means the preferred form of the work
+for making modifications to it.  "Object code" means any non-source
+form of a work.
+
+  A "Standard Interface" means an interface that either is an official
+standard defined by a recognized standards body, or, in the case of
+interfaces specified for a particular programming language, one that
+is widely used among developers working in that language.
+
+  The "System Libraries" of an executable work include anything, other
+than the work as a whole, that (a) is included in the normal form of
+packaging a Major Component, but which is not part of that Major
+Component, and (b) serves only to enable use of the work with that
+Major Component, or to implement a Standard Interface for which an
+implementation is available to the public in source code form.  A
+"Major Component", in this context, means a major essential component
+(kernel, window system, and so on) of the specific operating system
+(if any) on which the executable work runs, or a compiler used to
+produce the work, or an object code interpreter used to run it.
+
+  The "Corresponding Source" for a work in object code form means all
+the source code needed to generate, install, and (for an executable
+work) run the object code and to modify the work, including scripts to
+control those activities.  However, it does not include the work's
+System Libraries, or general-purpose tools or generally available free
+programs which are used unmodified in performing those activities but
+which are not part of the work.  For example, Corresponding Source
+includes interface definition files associated with source files for
+the work, and the source code for shared libraries and dynamically
+linked subprograms that the work is specifically designed to require,
+such as by intimate data communication or control flow between those
+subprograms and other parts of the work.
+
+  The Corresponding Source need not include anything that users
+can regenerate automatically from other parts of the Corresponding
+Source.
+
+  The Corresponding Source for a work in source code form is that
+same work.
+
+  2. Basic Permissions.
+
+  All rights granted under this License are granted for the term of
+copyright on the Program, and are irrevocable provided the stated
+conditions are met.  This License explicitly affirms your unlimited
+permission to run the unmodified Program.  The output from running a
+covered work is covered by this License only if the output, given its
+content, constitutes a covered work.  This License acknowledges your
+rights of fair use or other equivalent, as provided by copyright law.
+
+  You may make, run and propagate covered works that you do not
+convey, without conditions so long as your license otherwise remains
+in force.  You may convey covered works to others for the sole purpose
+of having them make modifications exclusively for you, or provide you
+with facilities for running those works, provided that you comply with
+the terms of this License in conveying all material for which you do
+not control copyright.  Those thus making or running the covered works
+for you must do so exclusively on your behalf, under your direction
+and control, on terms that prohibit them from making any copies of
+your copyrighted material outside their relationship with you.
+
+  Conveying under any other circumstances is permitted solely under
+the conditions stated below.  Sublicensing is not allowed; section 10
+makes it unnecessary.
+
+  3. Protecting Users' Legal Rights From Anti-Circumvention Law.
+
+  No covered work shall be deemed part of an effective technological
+measure under any applicable law fulfilling obligations under article
+11 of the WIPO copyright treaty adopted on 20 December 1996, or
+similar laws prohibiting or restricting circumvention of such
+measures.
+
+  When you convey a covered work, you waive any legal power to forbid
+circumvention of technological measures to the extent such circumvention
+is effected by exercising rights under this License with respect to
+the covered work, and you disclaim any intention to limit operation or
+modification of the work as a means of enforcing, against the work's
+users, your or third parties' legal rights to forbid circumvention of
+technological measures.
+
+  4. Conveying Verbatim Copies.
+
+  You may convey verbatim copies of the Program's source code as you
+receive it, in any medium, provided that you conspicuously and
+appropriately publish on each copy an appropriate copyright notice;
+keep intact all notices stating that this License and any
+non-permissive terms added in accord with section 7 apply to the code;
+keep intact all notices of the absence of any warranty; and give all
+recipients a copy of this License along with the Program.
+
+  You may charge any price or no price for each copy that you convey,
+and you may offer support or warranty protection for a fee.
+
+  5. Conveying Modified Source Versions.
+
+  You may convey a work based on the Program, or the modifications to
+produce it from the Program, in the form of source code under the
+terms of section 4, provided that you also meet all of these conditions:
+
+    a) The work must carry prominent notices stating that you modified
+    it, and giving a relevant date.
+
+    b) The work must carry prominent notices stating that it is
+    released under this License and any conditions added under section
+    7.  This requirement modifies the requirement in section 4 to
+    "keep intact all notices".
+
+    c) You must license the entire work, as a whole, under this
+    License to anyone who comes into possession of a copy.  This
+    License will therefore apply, along with any applicable section 7
+    additional terms, to the whole of the work, and all its parts,
+    regardless of how they are packaged.  This License gives no
+    permission to license the work in any other way, but it does not
+    invalidate such permission if you have separately received it.
+
+    d) If the work has interactive user interfaces, each must display
+    Appropriate Legal Notices; however, if the Program has interactive
+    interfaces that do not display Appropriate Legal Notices, your
+    work need not make them do so.
+
+  A compilation of a covered work with other separate and independent
+works, which are not by their nature extensions of the covered work,
+and which are not combined with it such as to form a larger program,
+in or on a volume of a storage or distribution medium, is called an
+"aggregate" if the compilation and its resulting copyright are not
+used to limit the access or legal rights of the compilation's users
+beyond what the individual works permit.  Inclusion of a covered work
+in an aggregate does not cause this License to apply to the other
+parts of the aggregate.
+
+  6. Conveying Non-Source Forms.
+
+  You may convey a covered work in object code form under the terms
+of sections 4 and 5, provided that you also convey the
+machine-readable Corresponding Source under the terms of this License,
+in one of these ways:
+
+    a) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by the
+    Corresponding Source fixed on a durable physical medium
+    customarily used for software interchange.
+
+    b) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by a
+    written offer, valid for at least three years and valid for as
+    long as you offer spare parts or customer support for that product
+    model, to give anyone who possesses the object code either (1) a
+    copy of the Corresponding Source for all the software in the
+    product that is covered by this License, on a durable physical
+    medium customarily used for software interchange, for a price no
+    more than your reasonable cost of physically performing this
+    conveying of source, or (2) access to copy the
+    Corresponding Source from a network server at no charge.
+
+    c) Convey individual copies of the object code with a copy of the
+    written offer to provide the Corresponding Source.  This
+    alternative is allowed only occasionally and noncommercially, and
+    only if you received the object code with such an offer, in accord
+    with subsection 6b.
+
+    d) Convey the object code by offering access from a designated
+    place (gratis or for a charge), and offer equivalent access to the
+    Corresponding Source in the same way through the same place at no
+    further charge.  You need not require recipients to copy the
+    Corresponding Source along with the object code.  If the place to
+    copy the object code is a network server, the Corresponding Source
+    may be on a different server (operated by you or a third party)
+    that supports equivalent copying facilities, provided you maintain
+    clear directions next to the object code saying where to find the
+    Corresponding Source.  Regardless of what server hosts the
+    Corresponding Source, you remain obligated to ensure that it is
+    available for as long as needed to satisfy these requirements.
+
+    e) Convey the object code using peer-to-peer transmission, provided
+    you inform other peers where the object code and Corresponding
+    Source of the work are being offered to the general public at no
+    charge under subsection 6d.
+
+  A separable portion of the object code, whose source code is excluded
+from the Corresponding Source as a System Library, need not be
+included in conveying the object code work.
+
+  A "User Product" is either (1) a "consumer product", which means any
+tangible personal property which is normally used for personal, family,
+or household purposes, or (2) anything designed or sold for incorporation
+into a dwelling.  In determining whether a product is a consumer product,
+doubtful cases shall be resolved in favor of coverage.  For a particular
+product received by a particular user, "normally used" refers to a
+typical or common use of that class of product, regardless of the status
+of the particular user or of the way in which the particular user
+actually uses, or expects or is expected to use, the product.  A product
+is a consumer product regardless of whether the product has substantial
+commercial, industrial or non-consumer uses, unless such uses represent
+the only significant mode of use of the product.
+
+  "Installation Information" for a User Product means any methods,
+procedures, authorization keys, or other information required to install
+and execute modified versions of a covered work in that User Product from
+a modified version of its Corresponding Source.  The information must
+suffice to ensure that the continued functioning of the modified object
+code is in no case prevented or interfered with solely because
+modification has been made.
+
+  If you convey an object code work under this section in, or with, or
+specifically for use in, a User Product, and the conveying occurs as
+part of a transaction in which the right of possession and use of the
+User Product is transferred to the recipient in perpetuity or for a
+fixed term (regardless of how the transaction is characterized), the
+Corresponding Source conveyed under this section must be accompanied
+by the Installation Information.  But this requirement does not apply
+if neither you nor any third party retains the ability to install
+modified object code on the User Product (for example, the work has
+been installed in ROM).
+
+  The requirement to provide Installation Information does not include a
+requirement to continue to provide support service, warranty, or updates
+for a work that has been modified or installed by the recipient, or for
+the User Product in which it has been modified or installed.  Access to a
+network may be denied when the modification itself materially and
+adversely affects the operation of the network or violates the rules and
+protocols for communication across the network.
+
+  Corresponding Source conveyed, and Installation Information provided,
+in accord with this section must be in a format that is publicly
+documented (and with an implementation available to the public in
+source code form), and must require no special password or key for
+unpacking, reading or copying.
+
+  7. Additional Terms.
+
+  "Additional permissions" are terms that supplement the terms of this
+License by making exceptions from one or more of its conditions.
+Additional permissions that are applicable to the entire Program shall
+be treated as though they were included in this License, to the extent
+that they are valid under applicable law.  If additional permissions
+apply only to part of the Program, that part may be used separately
+under those permissions, but the entire Program remains governed by
+this License without regard to the additional permissions.
+
+  When you convey a copy of a covered work, you may at your option
+remove any additional permissions from that copy, or from any part of
+it.  (Additional permissions may be written to require their own
+removal in certain cases when you modify the work.)  You may place
+additional permissions on material, added by you to a covered work,
+for which you have or can give appropriate copyright permission.
+
+  Notwithstanding any other provision of this License, for material you
+add to a covered work, you may (if authorized by the copyright holders of
+that material) supplement the terms of this License with terms:
+
+    a) Disclaiming warranty or limiting liability differently from the
+    terms of sections 15 and 16 of this License; or
+
+    b) Requiring preservation of specified reasonable legal notices or
+    author attributions in that material or in the Appropriate Legal
+    Notices displayed by works containing it; or
+
+    c) Prohibiting misrepresentation of the origin of that material, or
+    requiring that modified versions of such material be marked in
+    reasonable ways as different from the original version; or
+
+    d) Limiting the use for publicity purposes of names of licensors or
+    authors of the material; or
+
+    e) Declining to grant rights under trademark law for use of some
+    trade names, trademarks, or service marks; or
+
+    f) Requiring indemnification of licensors and authors of that
+    material by anyone who conveys the material (or modified versions of
+    it) with contractual assumptions of liability to the recipient, for
+    any liability that these contractual assumptions directly impose on
+    those licensors and authors.
+
+  All other non-permissive additional terms are considered "further
+restrictions" within the meaning of section 10.  If the Program as you
+received it, or any part of it, contains a notice stating that it is
+governed by this License along with a term that is a further
+restriction, you may remove that term.  If a license document contains
+a further restriction but permits relicensing or conveying under this
+License, you may add to a covered work material governed by the terms
+of that license document, provided that the further restriction does
+not survive such relicensing or conveying.
+
+  If you add terms to a covered work in accord with this section, you
+must place, in the relevant source files, a statement of the
+additional terms that apply to those files, or a notice indicating
+where to find the applicable terms.
+
+  Additional terms, permissive or non-permissive, may be stated in the
+form of a separately written license, or stated as exceptions;
+the above requirements apply either way.
+
+  8. Termination.
+
+  You may not propagate or modify a covered work except as expressly
+provided under this License.  Any attempt otherwise to propagate or
+modify it is void, and will automatically terminate your rights under
+this License (including any patent licenses granted under the third
+paragraph of section 11).
+
+  However, if you cease all violation of this License, then your
+license from a particular copyright holder is reinstated (a)
+provisionally, unless and until the copyright holder explicitly and
+finally terminates your license, and (b) permanently, if the copyright
+holder fails to notify you of the violation by some reasonable means
+prior to 60 days after the cessation.
+
+  Moreover, your license from a particular copyright holder is
+reinstated permanently if the copyright holder notifies you of the
+violation by some reasonable means, this is the first time you have
+received notice of violation of this License (for any work) from that
+copyright holder, and you cure the violation prior to 30 days after
+your receipt of the notice.
+
+  Termination of your rights under this section does not terminate the
+licenses of parties who have received copies or rights from you under
+this License.  If your rights have been terminated and not permanently
+reinstated, you do not qualify to receive new licenses for the same
+material under section 10.
+
+  9. Acceptance Not Required for Having Copies.
+
+  You are not required to accept this License in order to receive or
+run a copy of the Program.  Ancillary propagation of a covered work
+occurring solely as a consequence of using peer-to-peer transmission
+to receive a copy likewise does not require acceptance.  However,
+nothing other than this License grants you permission to propagate or
+modify any covered work.  These actions infringe copyright if you do
+not accept this License.  Therefore, by modifying or propagating a
+covered work, you indicate your acceptance of this License to do so.
+
+  10. Automatic Licensing of Downstream Recipients.
+
+  Each time you convey a covered work, the recipient automatically
+receives a license from the original licensors, to run, modify and
+propagate that work, subject to this License.  You are not responsible
+for enforcing compliance by third parties with this License.
+
+  An "entity transaction" is a transaction transferring control of an
+organization, or substantially all assets of one, or subdividing an
+organization, or merging organizations.  If propagation of a covered
+work results from an entity transaction, each party to that
+transaction who receives a copy of the work also receives whatever
+licenses to the work the party's predecessor in interest had or could
+give under the previous paragraph, plus a right to possession of the
+Corresponding Source of the work from the predecessor in interest, if
+the predecessor has it or can get it with reasonable efforts.
+
+  You may not impose any further restrictions on the exercise of the
+rights granted or affirmed under this License.  For example, you may
+not impose a license fee, royalty, or other charge for exercise of
+rights granted under this License, and you may not initiate litigation
+(including a cross-claim or counterclaim in a lawsuit) alleging that
+any patent claim is infringed by making, using, selling, offering for
+sale, or importing the Program or any portion of it.
+
+  11. Patents.
+
+  A "contributor" is a copyright holder who authorizes use under this
+License of the Program or a work on which the Program is based.  The
+work thus licensed is called the contributor's "contributor version".
+
+  A contributor's "essential patent claims" are all patent claims
+owned or controlled by the contributor, whether already acquired or
+hereafter acquired, that would be infringed by some manner, permitted
+by this License, of making, using, or selling its contributor version,
+but do not include claims that would be infringed only as a
+consequence of further modification of the contributor version.  For
+purposes of this definition, "control" includes the right to grant
+patent sublicenses in a manner consistent with the requirements of
+this License.
+
+  Each contributor grants you a non-exclusive, worldwide, royalty-free
+patent license under the contributor's essential patent claims, to
+make, use, sell, offer for sale, import and otherwise run, modify and
+propagate the contents of its contributor version.
+
+  In the following three paragraphs, a "patent license" is any express
+agreement or commitment, however denominated, not to enforce a patent
+(such as an express permission to practice a patent or covenant not to
+sue for patent infringement).  To "grant" such a patent license to a
+party means to make such an agreement or commitment not to enforce a
+patent against the party.
+
+  If you convey a covered work, knowingly relying on a patent license,
+and the Corresponding Source of the work is not available for anyone
+to copy, free of charge and under the terms of this License, through a
+publicly available network server or other readily accessible means,
+then you must either (1) cause the Corresponding Source to be so
+available, or (2) arrange to deprive yourself of the benefit of the
+patent license for this particular work, or (3) arrange, in a manner
+consistent with the requirements of this License, to extend the patent
+license to downstream recipients.  "Knowingly relying" means you have
+actual knowledge that, but for the patent license, your conveying the
+covered work in a country, or your recipient's use of the covered work
+in a country, would infringe one or more identifiable patents in that
+country that you have reason to believe are valid.
+
+  If, pursuant to or in connection with a single transaction or
+arrangement, you convey, or propagate by procuring conveyance of, a
+covered work, and grant a patent license to some of the parties
+receiving the covered work authorizing them to use, propagate, modify
+or convey a specific copy of the covered work, then the patent license
+you grant is automatically extended to all recipients of the covered
+work and works based on it.
+
+  A patent license is "discriminatory" if it does not include within
+the scope of its coverage, prohibits the exercise of, or is
+conditioned on the non-exercise of one or more of the rights that are
+specifically granted under this License.  You may not convey a covered
+work if you are a party to an arrangement with a third party that is
+in the business of distributing software, under which you make payment
+to the third party based on the extent of your activity of conveying
+the work, and under which the third party grants, to any of the
+parties who would receive the covered work from you, a discriminatory
+patent license (a) in connection with copies of the covered work
+conveyed by you (or copies made from those copies), or (b) primarily
+for and in connection with specific products or compilations that
+contain the covered work, unless you entered into that arrangement,
+or that patent license was granted, prior to 28 March 2007.
+
+  Nothing in this License shall be construed as excluding or limiting
+any implied license or other defenses to infringement that may
+otherwise be available to you under applicable patent law.
+
+  12. No Surrender of Others' Freedom.
+
+  If conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot convey a
+covered work so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you may
+not convey it at all.  For example, if you agree to terms that obligate you
+to collect a royalty for further conveying from those to whom you convey
+the Program, the only way you could satisfy both those terms and this
+License would be to refrain entirely from conveying the Program.
+
+  13. Use with the GNU Affero General Public License.
+
+  Notwithstanding any other provision of this License, you have
+permission to link or combine any covered work with a work licensed
+under version 3 of the GNU Affero General Public License into a single
+combined work, and to convey the resulting work.  The terms of this
+License will continue to apply to the part which is the covered work,
+but the special requirements of the GNU Affero General Public License,
+section 13, concerning interaction through a network will apply to the
+combination as such.
+
+  14. Revised Versions of this License.
+
+  The Free Software Foundation may publish revised and/or new versions of
+the GNU General Public License from time to time.  Such new versions will
+be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+  Each version is given a distinguishing version number.  If the
+Program specifies that a certain numbered version of the GNU General
+Public License "or any later version" applies to it, you have the
+option of following the terms and conditions either of that numbered
+version or of any later version published by the Free Software
+Foundation.  If the Program does not specify a version number of the
+GNU General Public License, you may choose any version ever published
+by the Free Software Foundation.
+
+  If the Program specifies that a proxy can decide which future
+versions of the GNU General Public License can be used, that proxy's
+public statement of acceptance of a version permanently authorizes you
+to choose that version for the Program.
+
+  Later license versions may give you additional or different
+permissions.  However, no additional obligations are imposed on any
+author or copyright holder as a result of your choosing to follow a
+later version.
+
+  15. Disclaimer of Warranty.
+
+  THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
+APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
+HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
+OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
+THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
+IS WITH YOU.  SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
+ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
+
+  16. Limitation of Liability.
+
+  IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
+THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
+GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
+USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
+DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
+PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
+EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
+SUCH DAMAGES.
+
+  17. Interpretation of Sections 15 and 16.
+
+  If the disclaimer of warranty and limitation of liability provided
+above cannot be given local legal effect according to their terms,
+reviewing courts shall apply local law that most closely approximates
+an absolute waiver of all civil liability in connection with the
+Program, unless a warranty or assumption of liability accompanies a
+copy of the Program in return for a fee.
+
+                     END OF TERMS AND CONDITIONS
+
+            How to Apply These Terms to Your New Programs
+
+  If you develop a new program, and you want it to be of the greatest
+possible use to the public, the best way to achieve this is to make it
+free software which everyone can redistribute and change under these terms.
+
+  To do so, attach the following notices to the program.  It is safest
+to attach them to the start of each source file to most effectively
+state the exclusion of warranty; and each file should have at least
+the "copyright" line and a pointer to where the full notice is found.
+
+    <one line to give the program's name and a brief idea of what it does.>
+    Copyright (C) <year>  <name of author>
+
+    This program is free software: you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation, either version 3 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License
+    along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+Also add information on how to contact you by electronic and paper mail.
+
+  If the program does terminal interaction, make it output a short
+notice like this when it starts in an interactive mode:
+
+    <program>  Copyright (C) <year>  <name of author>
+    This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
+    This is free software, and you are welcome to redistribute it
+    under certain conditions; type `show c' for details.
+
+The hypothetical commands `show w' and `show c' should show the appropriate
+parts of the General Public License.  Of course, your program's commands
+might be different; for a GUI interface, you would use an "about box".
+
+  You should also get your employer (if you work as a programmer) or school,
+if any, to sign a "copyright disclaimer" for the program, if necessary.
+For more information on this, and how to apply and follow the GNU GPL, see
+<https://www.gnu.org/licenses/>.
+
+  The GNU General Public License does not permit incorporating your program
+into proprietary programs.  If your program is a subroutine library, you
+may consider it more useful to permit linking proprietary applications with
+the library.  If this is what you want to do, use the GNU Lesser General
+Public License instead of this License.  But first, please read
+<https://www.gnu.org/licenses/why-not-lgpl.html>.

app.py

@@ -0,0 +1,55 @@
+from model import PopMusicTransformer
+import os
+os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
+import tensorflow as tf
+tf.compat.v1.disable_eager_execution()
+import gradio as gr
+import requests
+import torchtext
+import zipfile
+
+torchtext.utils.download_from_url("https://drive.google.com/uc?id=1gxuTSkF51NP04JZgTE46Pg4KQsbHQKGo", root=".")
+torchtext.utils.download_from_url("https://drive.google.com/uc?id=1nAKjaeahlzpVAX0F9wjQEG_hL4UosSbo", root=".")
+
+with zipfile.ZipFile("REMI-tempo-checkpoint.zip","r") as zip_ref:
+    zip_ref.extractall(".")
+with zipfile.ZipFile("REMI-tempo-chord-checkpoint.zip","r") as zip_ref:
+    zip_ref.extractall(".")
+
+url = 'https://github.com/AK391/remi/blob/master/input.midi?raw=true'
+r = requests.get(url, allow_redirects=True)
+open("input.midi", 'wb').write(r.content)
+
+
+# declare model
+model = PopMusicTransformer(
+    checkpoint='REMI-tempo-checkpoint',
+    is_training=False)
+
+def inference(midi): 
+    # generate continuation
+    model.generate(
+        n_target_bar=4,
+        temperature=1.2,
+        topk=5,
+        output_path='./result/continuation.midi',
+        prompt=midi.name)
+    return './result/continuation.midi'
+        
+
+title = "Pop Music Transformer"
+description = "demo for Pop Music Transformer. To use it, simply upload your midi file, or click one of the examples to load them. Read more at the links below."
+article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2002.00212'>Pop Music Transformer: Beat-based Modeling and Generation of Expressive Pop Piano Compositions</a> | <a href='https://github.com/YatingMusic/remi'>Github Repo</a></p>"
+
+examples = [
+    ['input.midi']
+]
+gr.Interface(
+    inference, 
+    gr.inputs.File(label="Input Midi"), 
+    gr.outputs.File(label="Output Midi"),
+    title=title,
+    description=description,
+    article=article,
+    examples=examples
+    ).launch()

chord_recognition.py

@@ -0,0 +1,188 @@
+import miditoolkit
+import numpy as np
+
+class MIDIChord(object):
+    def __init__(self):
+        # define pitch classes
+        self.PITCH_CLASSES = ['C', 'C#', 'D', 'D#', 'E', 'F', 'F#', 'G', 'G#', 'A', 'A#', 'B']
+        # define chord maps (required)
+        self.CHORD_MAPS = {'maj': [0, 4],
+                           'min': [0, 3],
+                           'dim': [0, 3, 6],
+                           'aug': [0, 4, 8],
+                           'dom': [0, 4, 7, 10]}
+        # define chord insiders (+1)
+        self.CHORD_INSIDERS = {'maj': [7],
+                               'min': [7],
+                               'dim': [9],
+                               'aug': [],
+                               'dom': []}
+        # define chord outsiders (-1)
+        self.CHORD_OUTSIDERS_1 = {'maj': [2, 5, 9],
+                                  'min': [2, 5, 8],
+                                  'dim': [2, 5, 10],
+                                  'aug': [2, 5, 9],
+                                  'dom': [2, 5, 9]}
+        # define chord outsiders (-2)
+        self.CHORD_OUTSIDERS_2 = {'maj': [1, 3, 6, 8, 10],
+                                  'min': [1, 4, 6, 9, 11],
+                                  'dim': [1, 4, 7, 8, 11],
+                                  'aug': [1, 3, 6, 7, 10],
+                                  'dom': [1, 3, 6, 8, 11]}
+
+    def note2pianoroll(self, notes, max_tick, ticks_per_beat):
+        return miditoolkit.pianoroll.parser.notes2pianoroll(
+                note_stream_ori=notes,
+                max_tick=max_tick,
+                ticks_per_beat=ticks_per_beat)
+
+    def sequencing(self, chroma):
+        candidates = {}
+        for index in range(len(chroma)):
+            if chroma[index]:
+                root_note = index
+                _chroma = np.roll(chroma, -root_note)
+                sequence = np.where(_chroma == 1)[0]
+                candidates[root_note] = list(sequence)
+        return candidates
+
+    def scoring(self, candidates):
+        scores = {}
+        qualities = {}
+        for root_note, sequence in candidates.items():
+            if 3 not in sequence and 4 not in sequence:
+                scores[root_note] = -100
+                qualities[root_note] = 'None'
+            elif 3 in sequence and 4 in sequence:
+                scores[root_note] = -100
+                qualities[root_note] = 'None'
+            else:
+                # decide quality
+                if 3 in sequence:
+                    if 6 in sequence:
+                        quality = 'dim'
+                    else:
+                        quality = 'min'
+                elif 4 in sequence:
+                    if 8 in sequence:
+                        quality = 'aug'
+                    else:
+                        if 7 in sequence and 10 in sequence:
+                            quality = 'dom'
+                        else:
+                            quality = 'maj'
+                # decide score
+                maps = self.CHORD_MAPS.get(quality)
+                _notes = [n for n in sequence if n not in maps]
+                score = 0
+                for n in _notes:
+                    if n in self.CHORD_OUTSIDERS_1.get(quality):
+                        score -= 1
+                    elif n in self.CHORD_OUTSIDERS_2.get(quality):
+                        score -= 2
+                    elif n in self.CHORD_INSIDERS.get(quality):
+                        score += 1
+                scores[root_note] = score
+                qualities[root_note] = quality
+        return scores, qualities
+
+    def find_chord(self, pianoroll):
+        chroma = miditoolkit.pianoroll.utils.tochroma(pianoroll=pianoroll)
+        chroma = np.sum(chroma, axis=0)
+        chroma = np.array([1 if c else 0 for c in chroma])
+        if np.sum(chroma) == 0:
+            return 'N', 'N', 'N', 0
+        else:
+            candidates = self.sequencing(chroma=chroma)
+            scores, qualities = self.scoring(candidates=candidates)
+            # bass note
+            sorted_notes = []
+            for i, v in enumerate(np.sum(pianoroll, axis=0)):
+                if v > 0:
+                    sorted_notes.append(int(i%12))
+            bass_note = sorted_notes[0]
+            # root note
+            __root_note = []
+            _max = max(scores.values())
+            for _root_note, score in scores.items():
+                if score == _max:
+                    __root_note.append(_root_note)
+            if len(__root_note) == 1:
+                root_note = __root_note[0]
+            else:
+                #TODO: what should i do
+                for n in sorted_notes:
+                    if n in __root_note:
+                        root_note = n
+                        break
+            # quality
+            quality = qualities.get(root_note)
+            sequence = candidates.get(root_note)
+            # score
+            score = scores.get(root_note)
+            return self.PITCH_CLASSES[root_note], quality, self.PITCH_CLASSES[bass_note], score
+
+    def greedy(self, candidates, max_tick, min_length):
+        chords = []
+        # start from 0
+        start_tick = 0
+        while start_tick < max_tick:
+            _candidates = candidates.get(start_tick)
+            _candidates = sorted(_candidates.items(), key=lambda x: (x[1][-1], x[0]))
+            # choose
+            end_tick, (root_note, quality, bass_note, _) = _candidates[-1]
+            if root_note == bass_note:
+                chord = '{}:{}'.format(root_note, quality)
+            else:
+                chord = '{}:{}/{}'.format(root_note, quality, bass_note)
+            chords.append([start_tick, end_tick, chord])
+            start_tick = end_tick
+        # remove :None
+        temp = chords
+        while ':None' in temp[0][-1]:
+            try:
+                temp[1][0] = temp[0][0]
+                del temp[0]
+            except:
+                print('NO CHORD')
+                return []
+        temp2 = []
+        for chord in temp:
+            if ':None' not in chord[-1]:
+                temp2.append(chord)
+            else:
+                temp2[-1][1] = chord[1]
+        return temp2
+
+    def extract(self, notes):
+        # read
+        max_tick = max([n.end for n in notes])
+        ticks_per_beat = 480
+        pianoroll = self.note2pianoroll(
+            notes=notes, 
+            max_tick=max_tick, 
+            ticks_per_beat=ticks_per_beat)
+        # get lots of candidates
+        candidates = {}
+        # the shortest: 2 beat, longest: 4 beat
+        for interval in [4, 2]:
+            for start_tick in range(0, max_tick, ticks_per_beat):
+                # set target pianoroll
+                end_tick = int(ticks_per_beat * interval + start_tick)
+                if end_tick > max_tick:
+                    end_tick = max_tick
+                _pianoroll = pianoroll[start_tick:end_tick, :]
+                # find chord
+                root_note, quality, bass_note, score = self.find_chord(pianoroll=_pianoroll)
+                # save
+                if start_tick not in candidates:
+                    candidates[start_tick] = {}
+                    candidates[start_tick][end_tick] = (root_note, quality, bass_note, score)
+                else:
+                    if end_tick not in candidates[start_tick]:
+                        candidates[start_tick][end_tick] = (root_note, quality, bass_note, score)
+        # greedy
+        chords = self.greedy(candidates=candidates, 
+                             max_tick=max_tick, 
+                             min_length=ticks_per_beat)
+        return chords

finetune.py

@@ -0,0 +1,45 @@
+from model import PopMusicTransformer
+from glob import glob
+import os
+os.environ['CUDA_VISIBLE_DEVICES'] = '0'
+
+def main():
+    # declare model
+    model = PopMusicTransformer(
+        checkpoint='REMI-tempo-checkpoint',
+        is_training=True)
+    # prepare data
+    midi_paths = glob('YOUR PERSOANL FOLDER/*.midi') # you need to revise it
+    training_data = model.prepare_data(midi_paths=midi_paths)
+
+    # check output checkpoint folder
+    ####################################
+    # if you use "REMI-tempo-chord-checkpoint" for the pre-trained checkpoint
+    # please name your output folder as something with "chord"
+    # for example: my-love-chord, cute-doggy-chord, ...
+    # if use "REMI-tempo-checkpoint"
+    # for example: my-love, cute-doggy, ...
+    ####################################
+    output_checkpoint_folder = 'REMI-finetune' # your decision
+    if not os.path.exists(output_checkpoint_folder):
+        os.mkdir(output_checkpoint_folder)
+    
+    # finetune
+    model.finetune(
+        training_data=training_data,
+        output_checkpoint_folder=output_checkpoint_folder)
+
+    ####################################
+    # after finetuning, please choose which checkpoint you want to try
+    # and change the checkpoint names you choose into "model"
+    # and copy the "dictionary.pkl" into the your output_checkpoint_folder
+    # ***** the same as the content format in "REMI-tempo-checkpoint" *****
+    # and then, you can use "main.py" to generate your own music!
+    # (do not forget to revise the checkpoint path to your own in "main.py")
+    ####################################
+
+    # close
+    model.close()
+
+if __name__ == '__main__':
+    main()

main.py

@@ -0,0 +1,31 @@
+from model import PopMusicTransformer
+import os
+os.environ['CUDA_VISIBLE_DEVICES'] = '0'
+
+def main():
+    # declare model
+    model = PopMusicTransformer(
+        checkpoint='REMI-tempo-checkpoint',
+        is_training=False)
+    
+    # generate from scratch
+    model.generate(
+        n_target_bar=16,
+        temperature=1.2,
+        topk=5,
+        output_path='./result/from_scratch.midi',
+        prompt=None)
+    
+    # generate continuation
+    model.generate(
+        n_target_bar=16,
+        temperature=1.2,
+        topk=5
+        output_path='./result/continuation.midi',
+        prompt='./data/evaluation/000.midi')
+    
+    # close model
+    model.close()
+
+if __name__ == '__main__':
+    main()

model.py

@@ -0,0 +1,294 @@
+import tensorflow as tf
+import numpy as np
+import miditoolkit
+import modules
+import pickle
+import utils
+import time
+
+class PopMusicTransformer(object):
+    ########################################
+    # initialize
+    ########################################
+    def __init__(self, checkpoint, is_training=False):
+        # load dictionary
+        self.dictionary_path = '{}/dictionary.pkl'.format(checkpoint)
+        self.event2word, self.word2event = pickle.load(open(self.dictionary_path, 'rb'))
+        # model settings
+        self.x_len = 512
+        self.mem_len = 512
+        self.n_layer = 12
+        self.d_embed = 512
+        self.d_model = 512
+        self.dropout = 0.1
+        self.n_head = 8
+        self.d_head = self.d_model // self.n_head
+        self.d_ff = 2048
+        self.n_token = len(self.event2word)
+        self.learning_rate = 0.0002
+        # load model
+        self.is_training = is_training
+        if self.is_training:
+            self.batch_size = 4
+        else:
+            self.batch_size = 1
+        self.checkpoint_path = '{}/model'.format(checkpoint)
+        self.load_model()
+
+    ########################################
+    # load model
+    ########################################
+    def load_model(self):
+        # placeholders
+        self.x = tf.compat.v1.placeholder(tf.int32, shape=[self.batch_size, None])
+        self.y = tf.compat.v1.placeholder(tf.int32, shape=[self.batch_size, None])
+        self.mems_i = [tf.compat.v1.placeholder(tf.float32, [self.mem_len, self.batch_size, self.d_model]) for _ in range(self.n_layer)]
+        # model
+        self.global_step = tf.compat.v1.train.get_or_create_global_step()
+        initializer = tf.compat.v1.initializers.random_normal(stddev=0.02, seed=None)
+        proj_initializer = tf.compat.v1.initializers.random_normal(stddev=0.01, seed=None)
+        with tf.compat.v1.variable_scope(tf.compat.v1.get_variable_scope()):
+            xx = tf.transpose(self.x, [1, 0])
+            yy = tf.transpose(self.y, [1, 0])
+            loss, self.logits, self.new_mem = modules.transformer(
+                dec_inp=xx,
+                target=yy,
+                mems=self.mems_i,
+                n_token=self.n_token,
+                n_layer=self.n_layer,
+                d_model=self.d_model,
+                d_embed=self.d_embed,
+                n_head=self.n_head,
+                d_head=self.d_head,
+                d_inner=self.d_ff,
+                dropout=self.dropout,
+                dropatt=self.dropout,
+                initializer=initializer,
+                proj_initializer=proj_initializer,
+                is_training=self.is_training,
+                mem_len=self.mem_len,
+                cutoffs=[],
+                div_val=-1,
+                tie_projs=[],
+                same_length=False,
+                clamp_len=-1,
+                input_perms=None,
+                target_perms=None,
+                head_target=None,
+                untie_r=False,
+                proj_same_dim=True)
+        self.avg_loss = tf.reduce_mean(loss)
+        # vars
+        all_vars = tf.compat.v1.trainable_variables()
+        grads = tf.gradients(self.avg_loss, all_vars)
+        grads_and_vars = list(zip(grads, all_vars))
+        all_trainable_vars = tf.reduce_sum([tf.reduce_prod(v.shape) for v in tf.compat.v1.trainable_variables()])
+        # optimizer
+        decay_lr = tf.compat.v1.train.cosine_decay(
+            self.learning_rate,
+            global_step=self.global_step,
+            decay_steps=400000,
+            alpha=0.004)
+        optimizer = tf.compat.v1.train.AdamOptimizer(learning_rate=decay_lr)
+        self.train_op = optimizer.apply_gradients(grads_and_vars, self.global_step)
+        # saver
+        self.saver = tf.compat.v1.train.Saver()
+        config = tf.compat.v1.ConfigProto(allow_soft_placement=True)
+        config.gpu_options.allow_growth = True
+        self.sess = tf.compat.v1.Session(config=config)
+        self.saver.restore(self.sess, self.checkpoint_path)
+
+    ########################################
+    # temperature sampling
+    ########################################
+    def temperature_sampling(self, logits, temperature, topk):
+        probs = np.exp(logits / temperature) / np.sum(np.exp(logits / temperature))
+        if topk == 1:
+            prediction = np.argmax(probs)
+        else:
+            sorted_index = np.argsort(probs)[::-1]
+            candi_index = sorted_index[:topk]
+            candi_probs = [probs[i] for i in candi_index]
+            # normalize probs
+            candi_probs /= sum(candi_probs)
+            # choose by predicted probs
+            prediction = np.random.choice(candi_index, size=1, p=candi_probs)[0]
+        return prediction
+
+    ########################################
+    # extract events for prompt continuation
+    ########################################
+    def extract_events(self, input_path):
+        note_items, tempo_items = utils.read_items(input_path)
+        note_items = utils.quantize_items(note_items)
+        max_time = note_items[-1].end
+        if 'chord' in self.checkpoint_path:
+            chord_items = utils.extract_chords(note_items)
+            items = chord_items + tempo_items + note_items
+        else:
+            items = tempo_items + note_items
+        groups = utils.group_items(items, max_time)
+        events = utils.item2event(groups)
+        return events
+
+    ########################################
+    # generate
+    ########################################
+    def generate(self, n_target_bar, temperature, topk, output_path, prompt=None):
+        # if prompt, load it. Or, random start
+        if prompt:
+            events = self.extract_events(prompt)
+            words = [[self.event2word['{}_{}'.format(e.name, e.value)] for e in events]]
+            words[0].append(self.event2word['Bar_None'])
+        else:
+            words = []
+            for _ in range(self.batch_size):
+                ws = [self.event2word['Bar_None']]
+                if 'chord' in self.checkpoint_path:
+                    tempo_classes = [v for k, v in self.event2word.items() if 'Tempo Class' in k]
+                    tempo_values = [v for k, v in self.event2word.items() if 'Tempo Value' in k]
+                    chords = [v for k, v in self.event2word.items() if 'Chord' in k]
+                    ws.append(self.event2word['Position_1/16'])
+                    ws.append(np.random.choice(chords))
+                    ws.append(self.event2word['Position_1/16'])
+                    ws.append(np.random.choice(tempo_classes))
+                    ws.append(np.random.choice(tempo_values))
+                else:
+                    tempo_classes = [v for k, v in self.event2word.items() if 'Tempo Class' in k]
+                    tempo_values = [v for k, v in self.event2word.items() if 'Tempo Value' in k]
+                    ws.append(self.event2word['Position_1/16'])
+                    ws.append(np.random.choice(tempo_classes))
+                    ws.append(np.random.choice(tempo_values))
+                words.append(ws)
+        # initialize mem
+        batch_m = [np.zeros((self.mem_len, self.batch_size, self.d_model), dtype=np.float32) for _ in range(self.n_layer)]
+        # generate
+        original_length = len(words[0])
+        initial_flag = 1
+        current_generated_bar = 0
+        while current_generated_bar < n_target_bar:
+            # input
+            if initial_flag:
+                temp_x = np.zeros((self.batch_size, original_length))
+                for b in range(self.batch_size):
+                    for z, t in enumerate(words[b]):
+                        temp_x[b][z] = t
+                initial_flag = 0
+            else:
+                temp_x = np.zeros((self.batch_size, 1))
+                for b in range(self.batch_size):
+                    temp_x[b][0] = words[b][-1]
+            # prepare feed dict
+            feed_dict = {self.x: temp_x}
+            for m, m_np in zip(self.mems_i, batch_m):
+                feed_dict[m] = m_np
+            # model (prediction)
+            _logits, _new_mem = self.sess.run([self.logits, self.new_mem], feed_dict=feed_dict)
+            # sampling
+            _logit = _logits[-1, 0]
+            word = self.temperature_sampling(
+                logits=_logit, 
+                temperature=temperature,
+                topk=topk)
+            words[0].append(word)
+            # if bar event (only work for batch_size=1)
+            if word == self.event2word['Bar_None']:
+                current_generated_bar += 1
+            # re-new mem
+            batch_m = _new_mem
+        # write
+        if prompt:
+            utils.write_midi(
+                words=words[0][original_length:],
+                word2event=self.word2event,
+                output_path=output_path,
+                prompt_path=prompt)
+        else:
+            utils.write_midi(
+                words=words[0],
+                word2event=self.word2event,
+                output_path=output_path,
+                prompt_path=None)
+
+    ########################################
+    # prepare training data
+    ########################################
+    def prepare_data(self, midi_paths):
+        # extract events
+        all_events = []
+        for path in midi_paths:
+            events = self.extract_events(path)
+            all_events.append(events)
+        # event to word
+        all_words = []
+        for events in all_events:
+            words = []
+            for event in events:
+                e = '{}_{}'.format(event.name, event.value)
+                if e in self.event2word:
+                    words.append(self.event2word[e])
+                else:
+                    # OOV
+                    if event.name == 'Note Velocity':
+                        # replace with max velocity based on our training data
+                        words.append(self.event2word['Note Velocity_21'])
+                    else:
+                        # something is wrong
+                        # you should handle it for your own purpose
+                        print('something is wrong! {}'.format(e))
+            all_words.append(words)
+        # to training data
+        self.group_size = 5
+        segments = []
+        for words in all_words:
+            pairs = []
+            for i in range(0, len(words)-self.x_len-1, self.x_len):
+                x = words[i:i+self.x_len]
+                y = words[i+1:i+self.x_len+1]
+                pairs.append([x, y])
+            pairs = np.array(pairs)
+            # abandon the last
+            for i in np.arange(0, len(pairs)-self.group_size, self.group_size*2):
+                data = pairs[i:i+self.group_size]
+                if len(data) == self.group_size:
+                    segments.append(data)
+        segments = np.array(segments)
+        return segments
+
+    ########################################
+    # finetune
+    ########################################
+    def finetune(self, training_data, output_checkpoint_folder):
+        # shuffle
+        index = np.arange(len(training_data))
+        np.random.shuffle(index)
+        training_data = training_data[index]
+        num_batches = len(training_data) // self.batch_size
+        st = time.time()
+        for e in range(200):
+            total_loss = []
+            for i in range(num_batches):
+                segments = training_data[self.batch_size*i:self.batch_size*(i+1)]
+                batch_m = [np.zeros((self.mem_len, self.batch_size, self.d_model), dtype=np.float32) for _ in range(self.n_layer)]
+                for j in range(self.group_size):
+                    batch_x = segments[:, j, 0, :]
+                    batch_y = segments[:, j, 1, :]
+                    # prepare feed dict
+                    feed_dict = {self.x: batch_x, self.y: batch_y}
+                    for m, m_np in zip(self.mems_i, batch_m):
+                        feed_dict[m] = m_np
+                    # run
+                    _, gs_, loss_, new_mem_ = self.sess.run([self.train_op, self.global_step, self.avg_loss, self.new_mem], feed_dict=feed_dict)
+                    batch_m = new_mem_
+                    total_loss.append(loss_)
+                    print('>>> Epoch: {}, Step: {}, Loss: {:.5f}, Time: {:.2f}'.format(e, gs_, loss_, time.time()-st))
+            self.saver.save(self.sess, '{}/model-{:03d}-{:.3f}'.format(output_checkpoint_folder, e, np.mean(total_loss)))
+            # stop
+            if np.mean(total_loss) <= 0.1:
+                break
+
+    ########################################
+    # close
+    ########################################
+    def close(self):
+        self.sess.close()

modules.py

@@ -0,0 +1,233 @@
+import tensorflow as tf
+
+def embedding_lookup(lookup_table, x):
+    return tf.compat.v1.nn.embedding_lookup(lookup_table, x)
+
+
+def normal_embedding_lookup(x, n_token, d_embed, d_proj, initializer,
+                            proj_initializer, scope='normal_embed', **kwargs):
+    emb_scale = d_proj ** 0.5
+    with tf.compat.v1.variable_scope(scope):
+        lookup_table = tf.compat.v1.get_variable('lookup_table', [n_token, d_embed], initializer=initializer)
+        y = embedding_lookup(lookup_table, x)
+        if d_proj != d_embed:
+            proj_W = tf.compat.v1.get_variable('proj_W', [d_embed, d_proj], initializer=proj_initializer)
+            y = tf.einsum('ibe,ed->ibd', y, proj_W)
+        else:
+            proj_W = None
+        ret_params = [lookup_table, proj_W]
+    y *= emb_scale
+    return y, ret_params
+
+
+def normal_softmax(hidden, target, n_token, params, scope='normal_softmax', **kwargs):
+    def _logit(x, W, b, proj):
+        y = x
+        if proj is not None:
+            y = tf.einsum('ibd,ed->ibe', y, proj)
+        return tf.einsum('ibd,nd->ibn', y, W) + b
+
+    params_W, params_projs = params[0], params[1]
+
+    with tf.compat.v1.variable_scope(scope):
+        softmax_b = tf.compat.v1.get_variable('bias', [n_token], initializer=tf.zeros_initializer())
+        output = _logit(hidden, params_W, softmax_b, params_projs)
+        nll = tf.nn.sparse_softmax_cross_entropy_with_logits(labels=target, logits=output)
+    return nll, output
+
+
+def positional_embedding(pos_seq, inv_freq, bsz=None):
+    sinusoid_inp = tf.einsum('i,j->ij', pos_seq, inv_freq)
+    pos_emb = tf.concat([tf.sin(sinusoid_inp), tf.cos(sinusoid_inp)], -1)
+    if bsz is not None:
+        return tf.tile(pos_emb[:, None, :], [1, bsz, 1])
+    else:
+        return pos_emb[:, None, :]
+
+
+def positionwise_FF(inp, d_model, d_inner, dropout, kernel_initializer,
+                    scope='ff', is_training=True):
+    output = inp
+    with tf.compat.v1.variable_scope(scope):
+        output = tf.keras.layers.Dense(d_inner, activation=tf.nn.relu, 
+                                       kernel_initializer=kernel_initializer, name='layer_1')(inp)
+        output = tf.keras.layers.Dropout(dropout, name='drop_1')(output, training=is_training)
+        output = tf.keras.layers.Dense(d_model, activation=tf.nn.relu, 
+                                       kernel_initializer=kernel_initializer, name='layer_2')(output)
+        output = tf.keras.layers.Dropout(dropout, name='drop_2')(output, training=is_training)
+        output = tf.keras.layers.LayerNormalization(axis=-1)(output + inp)
+    return output
+
+
+def _create_mask(qlen, mlen, same_length=False):
+    attn_mask = tf.ones([qlen, qlen])
+    mask_u = tf.linalg.band_part(attn_mask, 0, -1)
+    mask_dia = tf.linalg.band_part(attn_mask, 0, 0)
+    attn_mask_pad = tf.zeros([qlen, mlen])
+    ret = tf.concat([attn_mask_pad, mask_u - mask_dia], 1)
+    if same_length:
+        mask_l = tf.matrix_band_part(attn_mask, -1, 0)
+        ret = tf.concat([ret[:, :qlen] + mask_l - mask_dia, ret[:, qlen:]], 1)
+    return ret
+
+
+def _cache_mem(curr_out, prev_mem, mem_len=None):
+    if mem_len is None or prev_mem is None:
+        new_mem = curr_out
+    elif mem_len == 0:
+        return prev_mem
+    else:
+        new_mem = tf.concat([prev_mem, curr_out], 0)[-mem_len:]
+    return tf.stop_gradient(new_mem)
+
+
+def rel_shift(x):
+    x_size = tf.shape(x)
+    x = tf.pad(x, [[0, 0], [1, 0], [0, 0], [0, 0]])
+    x = tf.reshape(x, [x_size[1] + 1, x_size[0], x_size[2], x_size[3]])
+    x = tf.slice(x, [1, 0, 0, 0], [-1, -1, -1, -1])
+    x = tf.reshape(x, x_size)
+    return x
+
+
+def rel_multihead_attn(w, r, r_w_bias, r_r_bias, attn_mask, mems, d_model,
+                       n_head, d_head, dropout, dropatt, is_training,
+                       kernel_initializer, scope='rel_attn'):
+    scale = 1 / (d_head ** 0.5)
+    with tf.compat.v1.variable_scope(scope):
+        qlen = tf.shape(w)[0]
+        rlen = tf.shape(r)[0]
+        bsz = tf.shape(w)[1]
+
+        cat = tf.concat([mems, w], 0) if mems is not None and mems.shape.ndims > 1 else w
+
+        w_heads = tf.keras.layers.Dense(3 * n_head * d_head, use_bias=False, 
+                                        kernel_initializer=kernel_initializer, name='qkv')(cat)
+        r_head_k = tf.keras.layers.Dense(n_head * d_head, use_bias=False,
+                                         kernel_initializer=kernel_initializer, name='r')(r)
+        
+        w_head_q, w_head_k, w_head_v = tf.split(w_heads, 3, -1)
+        w_head_q = w_head_q[-qlen:]
+
+        klen = tf.shape(w_head_k)[0]
+
+        w_head_q = tf.reshape(w_head_q, [qlen, bsz, n_head, d_head])
+        w_head_k = tf.reshape(w_head_k, [klen, bsz, n_head, d_head])
+        w_head_v = tf.reshape(w_head_v, [klen, bsz, n_head, d_head])
+
+        r_head_k = tf.reshape(r_head_k, [rlen, n_head, d_head])
+
+        rw_head_q = w_head_q + r_w_bias
+        rr_head_q = w_head_q + r_r_bias
+
+        AC = tf.einsum('ibnd,jbnd->ijbn', rw_head_q, w_head_k)
+        BD = tf.einsum('ibnd,jnd->ijbn', rr_head_q, r_head_k)
+        BD = rel_shift(BD)
+
+        attn_score = (AC + BD) * scale
+        attn_mask_t = attn_mask[:, :, None, None]
+        attn_score = attn_score * (1 - attn_mask_t) - 1e30 * attn_mask_t
+
+        attn_prob = tf.nn.softmax(attn_score, 1)
+        attn_prob = tf.keras.layers.Dropout(dropatt)(attn_prob, training=is_training)
+
+        attn_vec = tf.einsum('ijbn,jbnd->ibnd', attn_prob, w_head_v)
+        size_t = tf.shape(attn_vec)
+        attn_vec = tf.reshape(attn_vec, [size_t[0], size_t[1], n_head * d_head])
+
+        attn_out = tf.keras.layers.Dense(d_model, use_bias=False, 
+                                         kernel_initializer=kernel_initializer, name='o')(attn_vec)
+        attn_out = tf.keras.layers.Dropout(dropout)(attn_out, training=is_training)
+        output = tf.keras.layers.LayerNormalization(axis=-1)(attn_out + w)
+        return output
+
+
+def transformer(dec_inp, target, mems, n_token, n_layer, d_model, d_embed,
+                n_head, d_head, d_inner, dropout, dropatt,
+                initializer, is_training, proj_initializer=None,
+                mem_len=None, cutoffs=[], div_val=1, tie_projs=[],
+                same_length=False, clamp_len=-1,
+                input_perms=None, target_perms=None, head_target=None,
+                untie_r=False, proj_same_dim=True,
+                scope='transformer'):
+    """
+    cutoffs: a list of python int. Cutoffs for adaptive softmax.
+    tie_projs: a list of python bools. Whether to tie the projections.
+    perms: a list of tensors. Each tensor should of size [len, bsz, bin_size].
+        Only used in the adaptive setting.
+    """
+    new_mems = []
+    with tf.compat.v1.variable_scope(scope):
+        if untie_r:
+            r_w_bias = tf.compat.v1.get_variable('r_w_bias', [n_layer, n_head, d_head], initializer=initializer)
+            r_r_bias = tf.compat.v1.get_variable('r_r_bias', [n_layer, n_head, d_head], initializer=initializer)
+        else:
+            r_w_bias = tf.compat.v1.get_variable('r_w_bias', [n_head, d_head], initializer=initializer)
+            r_r_bias = tf.compat.v1.get_variable('r_r_bias', [n_head, d_head], initializer=initializer)
+
+        qlen = tf.shape(dec_inp)[0]
+        mlen = tf.shape(mems[0])[0] if mems is not None else 0
+        klen = qlen + mlen
+
+        if proj_initializer is None:
+            proj_initializer = initializer
+
+        embeddings, shared_params = normal_embedding_lookup(
+            x=dec_inp,
+            n_token=n_token,
+            d_embed=d_embed,
+            d_proj=d_model,
+            initializer=initializer,
+            proj_initializer=proj_initializer)
+        
+        attn_mask = _create_mask(qlen, mlen, same_length)
+        
+        pos_seq = tf.range(klen - 1, -1, -1.0)
+        if clamp_len > 0:
+            pos_seq = tf.minimum(pos_seq, clamp_len)
+        inv_freq = 1 / (10000 ** (tf.range(0, d_model, 2.0) / d_model))
+        pos_emb = positional_embedding(pos_seq, inv_freq)
+
+        output = tf.keras.layers.Dropout(rate=dropout)(embeddings, training=is_training)
+        pos_emb = tf.keras.layers.Dropout(rate=dropout)(pos_emb, training=is_training)
+
+        if mems is None:
+            mems = [None] * n_layer
+
+        for i in range(n_layer):
+            # cache new mems
+            new_mems.append(_cache_mem(output, mems[i], mem_len))
+
+            with tf.compat.v1.variable_scope('layer_{}'.format(i)):
+                output = rel_multihead_attn(
+                    w=output,
+                    r=pos_emb,
+                    r_w_bias=r_w_bias if not untie_r else r_w_bias[i],
+                    r_r_bias=r_r_bias if not untie_r else r_r_bias[i],
+                    attn_mask=attn_mask,
+                    mems=mems[i],
+                    d_model=d_model,
+                    n_head=n_head,
+                    d_head=d_head,
+                    dropout=dropout,
+                    dropatt=dropatt,
+                    is_training=is_training,
+                    kernel_initializer=initializer)
+
+                output = positionwise_FF(
+                    inp=output,
+                    d_model=d_model,
+                    d_inner=d_inner,
+                    dropout=dropout,
+                    kernel_initializer=initializer,
+                    is_training=is_training)
+
+        output = tf.keras.layers.Dropout(dropout)(output, training=is_training)
+
+        loss, logits = normal_softmax(
+            hidden=output,
+            target=target,
+            n_token=n_token,
+            params=shared_params)
+
+        return loss, logits, new_mems

requirements.txt

@@ -0,0 +1,4 @@
+miditoolkit
+tensorflow-gpu==1.14.0
+gradio
+torchtext

utils.py

@@ -0,0 +1,348 @@
+import chord_recognition
+import numpy as np
+import miditoolkit
+import copy
+
+# parameters for input
+DEFAULT_VELOCITY_BINS = np.linspace(0, 128, 32+1, dtype=np.int)
+DEFAULT_FRACTION = 16
+DEFAULT_DURATION_BINS = np.arange(60, 3841, 60, dtype=int)
+DEFAULT_TEMPO_INTERVALS = [range(30, 90), range(90, 150), range(150, 210)]
+
+# parameters for output
+DEFAULT_RESOLUTION = 480
+
+# define "Item" for general storage
+class Item(object):
+    def __init__(self, name, start, end, velocity, pitch):
+        self.name = name
+        self.start = start
+        self.end = end
+        self.velocity = velocity
+        self.pitch = pitch
+
+    def __repr__(self):
+        return 'Item(name={}, start={}, end={}, velocity={}, pitch={})'.format(
+            self.name, self.start, self.end, self.velocity, self.pitch)
+
+# read notes and tempo changes from midi (assume there is only one track)
+def read_items(file_path):
+    midi_obj = miditoolkit.midi.parser.MidiFile(file_path)
+    # note
+    note_items = []
+    notes = midi_obj.instruments[0].notes
+    notes.sort(key=lambda x: (x.start, x.pitch))
+    for note in notes:
+        note_items.append(Item(
+            name='Note', 
+            start=note.start, 
+            end=note.end, 
+            velocity=note.velocity, 
+            pitch=note.pitch))
+    note_items.sort(key=lambda x: x.start)
+    # tempo
+    tempo_items = []
+    for tempo in midi_obj.tempo_changes:
+        tempo_items.append(Item(
+            name='Tempo',
+            start=tempo.time,
+            end=None,
+            velocity=None,
+            pitch=int(tempo.tempo)))
+    tempo_items.sort(key=lambda x: x.start)
+    # expand to all beat
+    max_tick = tempo_items[-1].start
+    existing_ticks = {item.start: item.pitch for item in tempo_items}
+    wanted_ticks = np.arange(0, max_tick+1, DEFAULT_RESOLUTION)
+    output = []
+    for tick in wanted_ticks:
+        if tick in existing_ticks:
+            output.append(Item(
+                name='Tempo',
+                start=tick,
+                end=None,
+                velocity=None,
+                pitch=existing_ticks[tick]))
+        else:
+            output.append(Item(
+                name='Tempo',
+                start=tick,
+                end=None,
+                velocity=None,
+                pitch=output[-1].pitch))
+    tempo_items = output
+    return note_items, tempo_items
+
+# quantize items
+def quantize_items(items, ticks=120):
+    # grid
+    grids = np.arange(0, items[-1].start, ticks, dtype=int)
+    # process
+    for item in items:
+        index = np.argmin(abs(grids - item.start))
+        shift = grids[index] - item.start
+        item.start += shift
+        item.end += shift
+    return items      
+
+# extract chord
+def extract_chords(items):
+    method = chord_recognition.MIDIChord()
+    chords = method.extract(notes=items)
+    output = []
+    for chord in chords:
+        output.append(Item(
+            name='Chord',
+            start=chord[0],
+            end=chord[1],
+            velocity=None,
+            pitch=chord[2].split('/')[0]))
+    return output
+
+# group items
+def group_items(items, max_time, ticks_per_bar=DEFAULT_RESOLUTION*4):
+    items.sort(key=lambda x: x.start)
+    downbeats = np.arange(0, max_time+ticks_per_bar, ticks_per_bar)
+    groups = []
+    for db1, db2 in zip(downbeats[:-1], downbeats[1:]):
+        insiders = []
+        for item in items:
+            if (item.start >= db1) and (item.start < db2):
+                insiders.append(item)
+        overall = [db1] + insiders + [db2]
+        groups.append(overall)
+    return groups
+
+# define "Event" for event storage
+class Event(object):
+    def __init__(self, name, time, value, text):
+        self.name = name
+        self.time = time
+        self.value = value
+        self.text = text
+
+    def __repr__(self):
+        return 'Event(name={}, time={}, value={}, text={})'.format(
+            self.name, self.time, self.value, self.text)
+
+# item to event
+def item2event(groups):
+    events = []
+    n_downbeat = 0
+    for i in range(len(groups)):
+        if 'Note' not in [item.name for item in groups[i][1:-1]]:
+            continue
+        bar_st, bar_et = groups[i][0], groups[i][-1]
+        n_downbeat += 1
+        events.append(Event(
+            name='Bar',
+            time=None, 
+            value=None,
+            text='{}'.format(n_downbeat)))
+        for item in groups[i][1:-1]:
+            # position
+            flags = np.linspace(bar_st, bar_et, DEFAULT_FRACTION, endpoint=False)
+            index = np.argmin(abs(flags-item.start))
+            events.append(Event(
+                name='Position', 
+                time=item.start,
+                value='{}/{}'.format(index+1, DEFAULT_FRACTION),
+                text='{}'.format(item.start)))
+            if item.name == 'Note':
+                # velocity
+                velocity_index = np.searchsorted(
+                    DEFAULT_VELOCITY_BINS, 
+                    item.velocity, 
+                    side='right') - 1
+                events.append(Event(
+                    name='Note Velocity',
+                    time=item.start, 
+                    value=velocity_index,
+                    text='{}/{}'.format(item.velocity, DEFAULT_VELOCITY_BINS[velocity_index])))
+                # pitch
+                events.append(Event(
+                    name='Note On',
+                    time=item.start, 
+                    value=item.pitch,
+                    text='{}'.format(item.pitch)))
+                # duration
+                duration = item.end - item.start
+                index = np.argmin(abs(DEFAULT_DURATION_BINS-duration))
+                events.append(Event(
+                    name='Note Duration',
+                    time=item.start,
+                    value=index,
+                    text='{}/{}'.format(duration, DEFAULT_DURATION_BINS[index])))
+            elif item.name == 'Chord':
+                events.append(Event(
+                    name='Chord', 
+                    time=item.start,
+                    value=item.pitch,
+                    text='{}'.format(item.pitch)))
+            elif item.name == 'Tempo':
+                tempo = item.pitch
+                if tempo in DEFAULT_TEMPO_INTERVALS[0]:
+                    tempo_style = Event('Tempo Class', item.start, 'slow', None)
+                    tempo_value = Event('Tempo Value', item.start, 
+                        tempo-DEFAULT_TEMPO_INTERVALS[0].start, None)
+                elif tempo in DEFAULT_TEMPO_INTERVALS[1]:
+                    tempo_style = Event('Tempo Class', item.start, 'mid', None)
+                    tempo_value = Event('Tempo Value', item.start, 
+                        tempo-DEFAULT_TEMPO_INTERVALS[1].start, None)
+                elif tempo in DEFAULT_TEMPO_INTERVALS[2]:
+                    tempo_style = Event('Tempo Class', item.start, 'fast', None)
+                    tempo_value = Event('Tempo Value', item.start, 
+                        tempo-DEFAULT_TEMPO_INTERVALS[2].start, None)
+                elif tempo < DEFAULT_TEMPO_INTERVALS[0].start:
+                    tempo_style = Event('Tempo Class', item.start, 'slow', None)
+                    tempo_value = Event('Tempo Value', item.start, 0, None)
+                elif tempo > DEFAULT_TEMPO_INTERVALS[2].stop:
+                    tempo_style = Event('Tempo Class', item.start, 'fast', None)
+                    tempo_value = Event('Tempo Value', item.start, 59, None)
+                events.append(tempo_style)
+                events.append(tempo_value)     
+    return events
+
+#############################################################################################
+# WRITE MIDI
+#############################################################################################
+def word_to_event(words, word2event):
+    events = []
+    for word in words:
+        event_name, event_value = word2event.get(word).split('_')
+        events.append(Event(event_name, None, event_value, None))
+    return events
+
+def write_midi(words, word2event, output_path, prompt_path=None):
+    events = word_to_event(words, word2event)
+    # get downbeat and note (no time)
+    temp_notes = []
+    temp_chords = []
+    temp_tempos = []
+    for i in range(len(events)-3):
+        if events[i].name == 'Bar' and i > 0:
+            temp_notes.append('Bar')
+            temp_chords.append('Bar')
+            temp_tempos.append('Bar')
+        elif events[i].name == 'Position' and \
+            events[i+1].name == 'Note Velocity' and \
+            events[i+2].name == 'Note On' and \
+            events[i+3].name == 'Note Duration':
+            # start time and end time from position
+            position = int(events[i].value.split('/')[0]) - 1
+            # velocity
+            index = int(events[i+1].value)
+            velocity = int(DEFAULT_VELOCITY_BINS[index])
+            # pitch
+            pitch = int(events[i+2].value)
+            # duration
+            index = int(events[i+3].value)
+            duration = DEFAULT_DURATION_BINS[index]
+            # adding
+            temp_notes.append([position, velocity, pitch, duration])
+        elif events[i].name == 'Position' and events[i+1].name == 'Chord':
+            position = int(events[i].value.split('/')[0]) - 1
+            temp_chords.append([position, events[i+1].value])
+        elif events[i].name == 'Position' and \
+            events[i+1].name == 'Tempo Class' and \
+            events[i+2].name == 'Tempo Value':
+            position = int(events[i].value.split('/')[0]) - 1
+            if events[i+1].value == 'slow':
+                tempo = DEFAULT_TEMPO_INTERVALS[0].start + int(events[i+2].value)
+            elif events[i+1].value == 'mid':
+                tempo = DEFAULT_TEMPO_INTERVALS[1].start + int(events[i+2].value)
+            elif events[i+1].value == 'fast':
+                tempo = DEFAULT_TEMPO_INTERVALS[2].start + int(events[i+2].value)
+            temp_tempos.append([position, tempo])
+    # get specific time for notes
+    ticks_per_beat = DEFAULT_RESOLUTION
+    ticks_per_bar = DEFAULT_RESOLUTION * 4 # assume 4/4
+    notes = []
+    current_bar = 0
+    for note in temp_notes:
+        if note == 'Bar':
+            current_bar += 1
+        else:
+            position, velocity, pitch, duration = note
+            # position (start time)
+            current_bar_st = current_bar * ticks_per_bar
+            current_bar_et = (current_bar + 1) * ticks_per_bar
+            flags = np.linspace(current_bar_st, current_bar_et, DEFAULT_FRACTION, endpoint=False, dtype=int)
+            st = flags[position]
+            # duration (end time)
+            et = st + duration
+            notes.append(miditoolkit.Note(velocity, pitch, st, et))
+    # get specific time for chords
+    if len(temp_chords) > 0:
+        chords = []
+        current_bar = 0
+        for chord in temp_chords:
+            if chord == 'Bar':
+                current_bar += 1
+            else:
+                position, value = chord
+                # position (start time)
+                current_bar_st = current_bar * ticks_per_bar
+                current_bar_et = (current_bar + 1) * ticks_per_bar
+                flags = np.linspace(current_bar_st, current_bar_et, DEFAULT_FRACTION, endpoint=False, dtype=int)
+                st = flags[position]
+                chords.append([st, value])
+    # get specific time for tempos
+    tempos = []
+    current_bar = 0
+    for tempo in temp_tempos:
+        if tempo == 'Bar':
+            current_bar += 1
+        else:
+            position, value = tempo
+            # position (start time)
+            current_bar_st = current_bar * ticks_per_bar
+            current_bar_et = (current_bar + 1) * ticks_per_bar
+            flags = np.linspace(current_bar_st, current_bar_et, DEFAULT_FRACTION, endpoint=False, dtype=int)
+            st = flags[position]
+            tempos.append([int(st), value])
+    # write
+    if prompt_path:
+        midi = miditoolkit.midi.parser.MidiFile(prompt_path)
+        #
+        last_time = DEFAULT_RESOLUTION * 4 * 4
+        # note shift
+        for note in notes:
+            note.start += last_time
+            note.end += last_time
+        midi.instruments[0].notes.extend(notes)
+        # tempo changes
+        temp_tempos = []
+        for tempo in midi.tempo_changes:
+            if tempo.time < DEFAULT_RESOLUTION*4*4:
+                temp_tempos.append(tempo)
+            else:
+                break
+        for st, bpm in tempos:
+            st += last_time
+            temp_tempos.append(miditoolkit.midi.containers.TempoChange(bpm, st))
+        midi.tempo_changes = temp_tempos
+        # write chord into marker
+        if len(temp_chords) > 0:
+            for c in chords:
+                midi.markers.append(
+                    miditoolkit.midi.containers.Marker(text=c[1], time=c[0]+last_time))
+    else:
+        midi = miditoolkit.midi.parser.MidiFile()
+        midi.ticks_per_beat = DEFAULT_RESOLUTION
+        # write instrument
+        inst = miditoolkit.midi.containers.Instrument(0, is_drum=False)
+        inst.notes = notes
+        midi.instruments.append(inst)
+        # write tempo
+        tempo_changes = []
+        for st, bpm in tempos:
+            tempo_changes.append(miditoolkit.midi.containers.TempoChange(bpm, st))
+        midi.tempo_changes = tempo_changes
+        # write chord into marker
+        if len(temp_chords) > 0:
+            for c in chords:
+                midi.markers.append(
+                    miditoolkit.midi.containers.Marker(text=c[1], time=c[0]))
+    # write
+    midi.dump(output_path)