DETAILED ACTION
This office action is in response to amendment filed on 7/26/2026.
Claims 1 – 4, 6, 9 – 12, 14, 15, 17 and 18 are amended.
Claims 1 – 20 are pending.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1 – 3, 6 – 9, 11, 14 and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Aladahalli et al (US 20220309315, hereinafter Aladahalli), in view of Che et al (US 20150347194, hereinafter Che), and further in view of Torres et al (US 20210209513, hereinafter Torres).
As per claim 1, Aladahalli discloses: A computer-implemented method, comprising:
receiving, by an orchestration platform, a request to perform a task with a fine-tuned model, the request comprising a model instance identifier; (Aladahalli [0047]: “if the electronic instructions identify one or more other computing tasks which are desired to be automated by the neural network 104, the extension component 114 can electronically insert into the neural network 104 one or more other sets of layers respectively corresponding to the one or more other computing tasks.” Examiner notes that the one or more other computing task is mapped to the claimed “model instance identifier”.)
retrieving, by the orchestration platform, the layer identified by the model instance identifier from a data store; (Aladahalli [0091]: “act 802 can comprise accessing, by a device (e.g., 112) operatively coupled to a processor, a neural network (e.g., 104), wherein the neural network includes a first set of layers (e.g., 106) trained to perform a first computing task.”; [0056]: “the first set of layers 106 can be trained to perform the first computing task. In various embodiments, the receiver component 112 can receive, retrieve, and/or otherwise access electronic instructions (not shown), which electronic instructions can identify a second computing task which is desired to be automated by the neural network 104.”)
loading, by the orchestration platform, the layer into the base model to generate the instance of the fine-tuned model, the base model pre-trained on a general dataset; and performing, by the orchestration platform, the task with the instance of the fine-tuned model. (Aladahalli [0092] – [0094]: “act 804 can include inserting, by the device (e.g., 114), a second set of layers (e.g., 302) into the neural network, wherein the second set of layers receive as input latent activations from the first set of layers… act 806 can include training, by the device (e.g., 116) and without changing the first set of layers, the second set of layers to perform a second computing task that is different from the first computing task… act 808 can include executing, by the device (e.g., 118), the neural network on an inputted data candidate (e.g., 202), wherein the first set of layers generate a first output (e.g., 204) corresponding to the first computing task, and wherein the second set of layers generate a second output (e.g., 402) corresponding to the second computing task.”)
Aladahalli did not explicitly disclose:
determining, by the orchestration platform, an instance of the fine-tuned model including a layer identified by the model instance identifier is not executing in an environment on the orchestration platform;
wherein the layer includes a set of fine-tuned parameters generated by a prior fine-tuning process and stored in the data store separately from the base model, the layer being pre-trained with data associated with the task;
initiating, by the orchestration platform, the environment with the instance of the fine-tuned model comprising the layer;
However, Che teaches:
determining, by the orchestration platform, an instance of the [VM] is not executing in an environment on the orchestration platform; (Che [0096]: “obtaining a configuration requirement with respect to a virtual machine as contained in a user need; and in response to the existence of a VM instance that satisfies the configuration requirement in a resource pool, providing the VM instance, wherein the resource pool is a resource pool according to a method described in the present invention.”; [0098]: “in response to no existence of a VM instance that satisfies the configuration requirement in the resource pool, searching in the resource pool for a VM instance that satisfies at least one part of the configuration requirement, installing to the VM instance application resources associated with other part of the configuration requirement, and providing the VM instance.”)
initiating, by the orchestration platform, the environment with the instance; (Che [0101]: “When the user requests a virtual machine where RHEL 6.4 and WAS 7.0 are installed, since the resource pool contains no VM instance satisfying this configuration requirement, the requested VM instance cannot be provided to the user directly. At this point, a VM instance where RHEL 6.4 is installed may be selected from the resource pool, and then WAS 7.0 is installed on this VM instance so as to satisfy the user need.”)
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the teaching of Che into that of Aladahalli in order to receive a request to perform a task with a VM and determining if an instance of the VM is not executing in an environment on the orchestration platform, and initiating, by the orchestration platform, the environment with the instance. Che figure 6 and [0096] – [0100] teaches the commonly known concept in resource allocation by first checking if a specified VM required by the user is already exists first before allocate the VM to execute the requests, as it would improve the resource scheduling efficiency of the task allocation system by lowering the overhead of constantly provisioning new VM for each request and allowing reuse of computing resources. Furthermore, one of ordinary skill in the art can easily see that the VM taught by Che can easily be configured to host specific programs such as layers of a model. Thus, Applicants have merely claimed the combination of known parts in the field to achieve the predictable results of improving resource allocation efficiency and is therefore rejected under 35 USC 103.
Torres teaches:
wherein the layer includes a set of fine-tuned parameters generated by a prior fine-tuning process and stored in the data store separately from the base model, the layer being pre-trained with data associated with the task; (Torres [0010] – [0011]: task specific adapter layer; figure 3 and [0020].)
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the teaching of Torres into that of Aladahalli and Che in order to have the layer includes a set of fine-tuned parameters generated by a prior fine-tuning process and stored in the data store separately from the base model, the layer being pre-trained with data associated with the task. Torres [0012] teaches the benefit of doing so would reduce the number of trainable parameters, the volume of training data, and the amount of training time and resources, transfer learning with the adapter layers may improve machine learning based methods of NLP. The combination of references would enhance the overall appeals of all references by reducing the amount of data needed to train a model and is therefore rejected under 35 USC 103.
As per claim 2, the combination of Aladahalli, Che and Torres further teach:
The computer-implemented method of claim 1, further comprising returning, by the orchestration platform, a result of performing the task. (Aladahalli [0094]: “act 808 can include executing, by the device (e.g., 118), the neural network on an inputted data candidate (e.g., 202), wherein the first set of layers generate a first output (e.g., 204) corresponding to the first computing task, and wherein the second set of layers generate a second output (e.g., 402) corresponding to the second computing task.”)
As per claim 3, the combination of Aladahalli, Che and Torres further teach:
The computer-implemented method of claim 1, further comprising pre-training the layer with the base model. (Aladahalli [0085] – [0086]: “act 706 can include installing, by the device (e.g., 114), an i-th set of layers (e.g., 302 and/or 502) into the neural network, such that the i-th set of layers branch off from some other set of layers in the neural network… act 708 can include training, by the device (e.g., 116), the i-th set of layers to perform an i-th computing task, while freezing parameters of all other sets of layers in the neural network.”)
As per claim 6, the combination of Aladahalli, Che and Torres further teach:
The computer-implemented method of claim 1, comprising identifying, by the orchestration platform, the base model with a model identifier. (Aladahalli [0091]: “act 802 can comprise accessing, by a device (e.g., 112) operatively coupled to a processor, a neural network (e.g., 104), wherein the neural network includes a first set of layers (e.g., 106) trained to perform a first computing task.”.)
As per claim 7, the combination of Aladahalli, Che and Torres further teach:
The computer-implemented method of claim 6, wherein the request comprises a payload and metadata, and the metadata further comprises the model identifier and the model instance identifier. (Aladahalli [0056])
As per claim 8, the combination of Aladahalli, Che and Torres further teach:
The computer-implemented method of claim 7, wherein the payload comprises data, and the performing the task comprises determining an inference by processing the data with the fine-tuned model. (Aladahalli [0092] – [0094]: “act 804 can include inserting, by the device (e.g., 114), a second set of layers (e.g., 302) into the neural network, wherein the second set of layers receive as input latent activations from the first set of layers… act 806 can include training, by the device (e.g., 116) and without changing the first set of layers, the second set of layers to perform a second computing task that is different from the first computing task… act 808 can include executing, by the device (e.g., 118), the neural network on an inputted data candidate (e.g., 202), wherein the first set of layers generate a first output (e.g., 204) corresponding to the first computing task, and wherein the second set of layers generate a second output (e.g., 402) corresponding to the second computing task.”)
As per claim 9, the combination of Aladahalli, Che and Torres further teach:
The computer-implemented method of claim 6, further comprising pre-training a plurality of base models including the base model with a different general dataset. (Aladahalli [0085] – [0086]: “act 706 can include installing, by the device (e.g., 114), an i-th set of layers (e.g., 302 and/or 502) into the neural network, such that the i-th set of layers branch off from some other set of layers in the neural network… act 708 can include training, by the device (e.g., 116), the i-th set of layers to perform an i-th computing task, while freezing parameters of all other sets of layers in the neural network.”)
As per claim 11, the combination of Aladahalli, Che and Torres further teach:
The computer-implemented method of claim 1, further comprising executing, by the orchestration platform, one or more of a plurality of base models including the base model in preparation to receive a plurality of layers. (Aladahalli [0091]: “act 802 can comprise accessing, by a device (e.g., 112) operatively coupled to a processor, a neural network (e.g., 104), wherein the neural network includes a first set of layers (e.g., 106) trained to perform a first computing task.”; [0085] – [0086]: “act 706 can include installing, by the device (e.g., 114), an i-th set of layers (e.g., 302 and/or 502) into the neural network, such that the i-th set of layers branch off from some other set of layers in the neural network… act 708 can include training, by the device (e.g., 116), the i-th set of layers to perform an i-th computing task, while freezing parameters of all other sets of layers in the neural network.”)
As per claim 14, it is the non-transitory computer-readable storage medium variant of claim 1 and is therefore rejected under the same rationale. (Aladahalli [0044])
As per claim 17, it is the non-transitory computer-readable storage medium variant of claim 6 and is therefore rejected under the same rationale.
Claim(s) 4, 5, 10, 15 and 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Aladahalli, Che and Torres, and further in view of Kale et al (US 20230333901, hereinafter Kale)..
As per claim 4, the combination of Aladahalli, Che and Torres did not teach:
The computer-implemented method of claim 1, further comprising storing, by the orchestration platform, a plurality of layers including the layer in the data store, wherein each of the plurality of layers is identified with a different model instance identifier.
However, Kale teaches:
The computer-implemented method of claim 1, further comprising storing, by the orchestration platform, a plurality of layers including the layer in the data store, wherein each of the plurality of layers is identified with a different model instance identifier. (Kale [0017] and [0023])
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the teaching of Kale into that of Aladahalli, Che and Torres in order to store a plurality of layers including the layer in the data store, wherein each of the plurality of layers is identified with a different model instance identifier. Aladahalli [0091] teaching accessing a neural network over internet, it would have been readily apparent and well known to one of ordinary skill in the art to see that the neural model can be stored in a repository such as the Model Store 130 of Kale to facilitate such access. Applicants have merely claimed the combination of known parts in the field to achieve predictable results of accessing stored ML model through network and is therefore rejected under 35 USC 103.
As per claim 5, the combination of Aladahalli, Che, Torres and Kale further teach:
The computer-implemented method of claim 4, wherein each of the plurality of layers is pre-trained with one of a plurality of base models. (Kale [0018])
As per claim 10, the combination of Aladahalli, Che and Torres did not teach:
The computer-implemented method of claim 1, further comprising storing, by the orchestration platform, a plurality of base models in the data store, wherein each of the plurality of base models is identified with a different model identifier.
However, Kale teaches:
The computer-implemented method of claim 1, further comprising storing, by the orchestration platform, a plurality of base models in the data store, wherein each of the plurality of base models is identified with a different model identifier. (Kale [0017] and [0023])
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the teaching of Kale into that of Aladahalli, Che and Torres in order to store a plurality of base models in the data store, wherein each of the plurality of base models is identified with a different model identifier. Aladahalli [0091] teaching accessing a neural network over internet, it would have been readily apparent and well known to one of ordinary skill in the art to see that the neural model can be stored in a repository such as the Model Store 130 of Kale to facilitate such access. Applicants have merely claimed the combination of known parts in the field to achieve predictable results of accessing stored ML model through network and is therefore rejected under 35 USC 103.
As per claim 15, it is the non-transitory computer-readable storage medium variant of claim 4 and is therefore rejected under the same rationale.
As per claim 16, it is the non-transitory computer-readable storage medium variant of claim 5 and is therefore rejected under the same rationale.
Claim(s) 18 – 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Aladahalli, in view of Torres.
As per claim 18, Aladahalli discloses: A computing apparatus comprising: a processor; and a memory storing instructions that, when executed by the processor, cause the processor to perform operations comprising:
executing a base model in an environment in a cluster system; loading one or more layers into a cache of the cluster system; (Aladahalli [0091]: “act 802 can comprise accessing, by a device (e.g., 112) operatively coupled to a processor, a neural network (e.g., 104), wherein the neural network includes a first set of layers (e.g., 106) trained to perform a first computing task.”; [0105]: “The illustrated embodiments of the embodiments herein can be also practiced in distributed computing environments where certain tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules can be located in both local and remote memory storage devices.”.)
receiving a request to perform a task, the request comprising a model instance identifier associated with a specific layer; (Aladahalli [0047]: “if the electronic instructions identify one or more other computing tasks which are desired to be automated by the neural network 104, the extension component 114 can electronically insert into the neural network 104 one or more other sets of layers respectively corresponding to the one or more other computing tasks.” Examiner notes that the one or more other computing task is mapped to the claimed “model instance identifier”.)
identifying the specific layer from the one or more layers loaded into the cache of the cluster system; loading the specific layer into the base model to generate a fine-tuned model to process the task; and processing the task to determine a result including one or more inferences. (Aladahalli [0092] – [0094]: “act 804 can include inserting, by the device (e.g., 114), a second set of layers (e.g., 302) into the neural network, wherein the second set of layers receive as input latent activations from the first set of layers… act 806 can include training, by the device (e.g., 116) and without changing the first set of layers, the second set of layers to perform a second computing task that is different from the first computing task… act 808 can include executing, by the device (e.g., 118), the neural network on an inputted data candidate (e.g., 202), wherein the first set of layers generate a first output (e.g., 204) corresponding to the first computing task, and wherein the second set of layers generate a second output (e.g., 402) corresponding to the second computing task.”)
Adadahalli did not explicitly disclose:
wherein the identifying the specific layer is based on the model instance identifier;
However, Torres teaches:
wherein the identifying the specific layer is based on the model instance identifier; (Torres [0030] and [0034]: model artifact.)
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the teaching of Torres into that of Aladahalli and Che in order to have the layer includes a set of fine-tuned parameters generated by a prior fine-tuning process and stored in the data store separately from the base model, the layer being pre-trained with data associated with the task. Torres [0012] teaches the benefit of doing so would reduce the number of trainable parameters, the volume of training data, and the amount of training time and resources, transfer learning with the adapter layers may improve machine learning based methods of ML. The combination of references would enhance the overall appeals of all references by reducing the amount of data needed to train a model and is therefore rejected under 35 USC 103.
As per claim 19, the combination of Aladahalli and Torres further teach:
The computing apparatus of claim 18, wherein the base model is trained on a general dataset, and the specific layer is trained with the base model with a specific dataset. (Aladahalli [0085] – [0086]: “act 706 can include installing, by the device (e.g., 114), an i-th set of layers (e.g., 302 and/or 502) into the neural network, such that the i-th set of layers branch off from some other set of layers in the neural network… act 708 can include training, by the device (e.g., 116), the i-th set of layers to perform an i-th computing task, while freezing parameters of all other sets of layers in the neural network.”)
As per claim 20, the combination of Aladahalli and Torres further teach:
The computing apparatus of claim 18, wherein the cache is host-level cache or cluster level cache. (Aladahalli [0105]: “The illustrated embodiments of the embodiments herein can be also practiced in distributed computing environments where certain tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules can be located in both local and remote memory storage devices.”.)
Response to Arguments
Applicant’s arguments with respect to claim(s) 1 – 20 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CHARLES M SWIFT whose telephone number is (571)270-7756. The examiner can normally be reached Monday - Friday: 9:30 AM - 7PM.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, April Blair can be reached at 5712701014. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/CHARLES M SWIFT/Primary Examiner, Art Unit 2196
Prosecution Insights