DETAILED ACTION
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, 10 and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ravi et al (US 20250077842) in view of Kundu et al (US 20240320510), in further view of Zhou et al (NPL: LoRA-drop: Efficient LoRA Parameter Pruning based on Output Evaluation).
Regarding claim 1, Ravi discloses an apparatus for fine-tuning machine learning models (abstract selectively conditioning layers of a neural network), the apparatus comprising:
at least one memory (¶105 one or more processors and system memory); and
at least one processor coupled to the at least one memory and configured to (¶105 one or more processors and system memory);
determine a plurality of sensitivity scores based on a query to edit a first image (¶71 the user interface 802 also includes an edit text element for entering a text prompt 808. The selective layer conditioning system 102 can receive the image prompt 806 and the text prompt 808 for conditioning a neural network).
Ravi fails to teach where Kundu teaches wherein each respective sensitivity score of the plurality of sensitivity scores represents a sensitivity of a respective layer of a plurality of layers of a machine learning model (¶31 a sensitivity represents a quantification of how sensitive a particular element of a model is to various inputs. In other words, a sensitivity score represents a relative importance of an element (e.g., a layer, a node, a block, etc.) of a foundational model to an output of the foundational model);
apply an adapter to one or more layers of the plurality of layers that have a respective sensitivity score greater than a sensitivity threshold to adjust weights of the one or more layers (¶59 The example fine-tuning circuitry 250 then performs fine tuning of the trimmed model. (Block 350). This fine tuning produces a fine-tuned model. In examples disclosed herein, the fine-tuning is applied to layers of the model having a respective sensitivity score that meets a threshold sensitivity (e.g., as defined by the compute budget)).
Ravi further fails to teach where Zhou teaches fine-tune parameters of the one or more layers based on application of the adapter to the one or more layers (pg. 1 Introduction: LoRA freezes the pre-trained parameters and introduces auxiliary trainable parameters ΔW for each layer as shown in Figure 1 (e.g. Fig. 1 shows that LoRA adds ΔWx in each layer and thereby fine-tunes the frozen model)).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein each respective sensitivity score of the plurality of sensitivity scores represents a sensitivity of a respective layer of a plurality of layers of a machine learning model and apply an adapter to one or more layers of the plurality of layers that have a respective sensitivity score greater than a sensitivity threshold to adjust weights of the one or more layers from Kundu, and the teaching of fine-tune parameters of the one or more layers based on application of the adapter to the one or more layers from Zhou, into the apparatus for fine-tuning machine learning models as disclosed by Ravi. The motivation for doing this is to improve and create a fine-tuned model by providing resource-efficient fine-tuning, and further to reduce the number of trainable parameters and computational/storage requirements associated with fine-tuning while retaining model performance.
Regarding claim(s) 10 (drawn to a method):
The rejection/proposed combination of Ravi, Kundu, and Zhou, explained in the rejection of apparatus claim(s) 1, anticipates/renders obvious the steps of the method of claim(s) 10 because these steps occur in the operation of the proposed combination as discussed above. Thus, the arguments similar to that presented above for claim(s) 1 is/are equally applicable to claim(s) 10.
Regarding claim(s) 19 (drawn to a CRM):
The rejection/proposed combination of Ravi, Kundu, and Zhou, explained in the rejection of apparatus claim(s) 1, anticipates/renders obvious the steps of the computer readable medium of claim(s) 19 because these steps occur in the operation of the proposed combination as discussed above. Thus, the arguments similar to that presented above for claim(s) 1 is/are equally applicable to claim(s) 19. See further Ravi ¶91 e.g. one or more instructions stored on a computer-readable storage medium and executable by processors of one or more computing devices.
Claim(s) 6-8 and 15-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Ravi, Kundu, and Zhou as applied to claim 1 and 10 above, and further in view of Zhang et al (NPL: ADALORA: ADAPTIVE BUDGET ALLOCATION FOR PARAMETER-EFFICIENT FINE-TUNING).
Regarding claim 6, the combination Ravi, Kundu, and Zhou teach the apparatus of claim 1, but fail to teach where Zhang teaches wherein the sensitivity threshold is variable based on the plurality of sensitivity scores (pg. 1 Abstract: we propose AdaLoRA, which adaptively allocates the parameter budget among weight matrices according to their importance score. Such a novel approach allows us to effectively prune the singular values of unimportant updates, which is essentially to reduce their parameter budget but circumvent intensive exact SVD computations; pg. 2 Introduction: diff pruning can increase the parameter efficiency substantially by adaptively retaining important updates and pruning unimportant ones).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the sensitivity threshold is variable based on the plurality of sensitivity scores from Zhang into the apparatus as disclosed by the combination of Ravi, Kundu, and Zhou. The motivation for doing this is to optimize fine-tuning performance.
Regarding claim 7, the combination of Ravi, Kundu, and Zhou teach the apparatus of claim 1. Zhang teaches in pg. 1 Abstract: we propose AdaLoRA, which adaptively allocates the parameter budget among weight matrices according to their importance score.
Although Zhang does not expressly disclose selecting the threshold as a value corresponding to the preset percentage of the computed scores, it would have been an obvious to one of ordinary skill in the art at the time the invention was made to implement Zhang’s pruning threshold based on a predetermined percentage of the ranked importance scores because the threshold determines the number of retained components and therefore the adaptation budget. The threshold value continues a result-effective variable, and selecting a percentage-based cutoff would have been a predictable design choice yielding the expected results of retaining a desired proportion of higher-scoring components while pruning lower-scoring components. Furthermore, the combination would optimize fine-tuning performance.
Regarding claim 8, the combination of Ravi, Kundu, and Zhou teach the apparatus of claim 1, but fail to teach where Zhang teaches wherein the adapter is a low-ranking adaptation (LoRA) adapter (pg. 1 Abstract: we propose AdaLoRA, which adaptively allocates the parameter budget among weight matrices according to their importance score).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the adapter is a low-ranking adaptation (LoRA) adapter from Zhang into the apparatus as disclosed by the combination of Ravi, Kundu, and Zhou. The motivation for doing this is to optimize fine-tuning performance.
Regarding claim(s) 15-17 (drawn to a method):
The rejection/proposed combination of Ravi, Kundu, Zhou and Zhang, explained in the rejection of apparatus claim(s) 6-8, anticipates/renders obvious the steps of the method of claim(s) 15-17 because these steps occur in the operation of the proposed combination as discussed above. Thus, the arguments similar to that presented above for claim(s) 6-8 is/are equally applicable to claim(s) 15-17.
Allowable Subject Matter
Claims 2-5, 9, 11-14, 18 and 20 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
The following is a statement of reasons for the indication of allowable subject matter:
Regarding claim 2, and similarly regarding claims 11 and 20, the prior art of record, alone or in combination, fails to teach at least “wherein the at least one processor is configured to: generate the first image using the machine learning model including a first text caption as input; add noise to the first image to reconstruct the first image; and determine the plurality of sensitivity scores based on a gradient associated with a loss function representing differences between: a noise prediction used to reconstruct the first image from the first text caption; and a noise prediction used to generate a second image from an augmented version of the first text caption.”. Claims 3-5 and 12-14 depend off of claims 2 and 11, respectively, and would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Regarding claim 9, and similarly regarding claim 18, the prior art of record, alone or in combination, fails to teach at least “wherein the LoRA adapter is applied head-wise to the one or more layers that have the respective sensitivity score greater than the sensitivity threshold”.
Response to Arguments
Applicant’s arguments with respect to claim(s) 1, 6-8, 10, 15-17, 19 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 KEVIN KY whose telephone number is (571)272-7648. The examiner can normally be reached Monday-Friday 9-5PM.
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, Vincent Rudolph can be reached at 571-272-8243. 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.
/KEVIN KY/ Primary Examiner, Art Unit 2671
