DETAILED ACTION
This non-final rejection is responsive to communication filed May 20, 2024. Claims 1-20 are pending in this application.
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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on June 24, 2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 19 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. The preamble of claim 19 recites “the computer system of claim 17.” However, claim 17 is a method claim and not a system claim. Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1 recites: grouping layers of the target neural network to provide a plurality of layer groups based on the neural network representation, at least one layer group comprising multiple layers from the target neural network that can be executed by a single operation; and generating a latency prediction for executing the target neural network on a target hardware configuration based on the layer groups. The broadest reasonable interpretation of these steps is that the steps fall within the mental process groupings of abstract ideas because they cover concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (with the aid of pen and paper) group layers of a neural network representation and generate a latency prediction for executing the target neural network.
This judicial exception is not integrated into a practical application. The limitation “receiving a neural network representation for a target neural network having a plurality of layers” is mere data gathering recited at a high level of generality, and thus is insignificant extra-solution activity. The limitations: “one or more computer storage media storing computer-useable instructions that, when used by one or more computing devices”, “using a first machine learning model” and “using a second machine learning model” to perform recited tasks amount to no more than mere instructions to apply the exception using a generic computer. See MPEP 2106.05(f). Further, the limitations “using a first machine learning model” and “using a second machine learning model” also merely indicate a field of use or technological environment (neural networks) in which the judicial exception is performed and thus fail to add an inventive concept to the claims. See MPEP 2106.05(h). Even when viewed in combination, these additional elements do not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitation of “receiving” is recited at a high level of generality. This element amounts to receiving or transmitting data over a network and are well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. As discussed above, the recitations of “one or more computer storage media storing computer-useable instructions that, when used by one or more computing devices”, “using a first machine learning model” and “using a second machine learning model” to perform recited tasks amount to no more than mere instructions to apply the exception using a generic computer component. Further, the limitations “using a first machine learning model” and “using a second machine learning model” also merely indicate a field of use or technological environment (neural networks) in which the judicial exception is performed and thus fail to add an inventive concept to the claims. See MPEP 2106.05(h). Even when considered in combination, these additional elements represent mere instructions to implement an abstract idea or other exception on a computer and insignificant extra-solution activity, which do not provide an inventive concept.
Claim 2 recites: obtaining a graph representing the target neural network, the graph including nodes representing the layers of the target neural network and edges between the nodes based on connections between the layers of the target neural network; label each edge in the graph as fusible or not fusible to provide edge labels; and generating the layer groups by dividing the graph into sub-graphs based on the edge labels. The broadest reasonable interpretation of these steps is that the steps fall within the mental process groupings of abstract ideas because they cover concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (with the aid of pen and paper) visualize/obtain a graph representing a neural network, label edges and divide the graph into sub-graphs.
This judicial exception is not integrated into a practical application and the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the limitation “causing the first machine learning model to label each edge” amounts to no more than mere instructions to apply the exception using a generic computer component. Even when considered in combination, this additional element represents mere instructions to implement an abstract idea or other exception on a computer, which does not provide an inventive concept.
Claim 3 recites: generating node embeddings based on features of the layers of the target neural network; generating edge embeddings based on the node embeddings; and labeling the edges in the graph based on the edge embeddings. The broadest reasonable interpretation of these steps is that the steps fall within the mental process groupings of abstract ideas because they cover concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (with the aid of pen and paper) generate node embeddings, generate edge embeddings, and label edges.
This judicial exception is not integrated into a practical application and the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the limitations “using a graph attention (GAT) model of the first machine learning model”, “using a long short-term memory (LSTM) model of the first machine learning model” and “using a linear model of the first machine learning model” amount to no more than mere instructions to apply the exception using a generic computer component. Even when considered in combination, these additional elements represent mere instructions to implement an abstract idea or other exception on a computer, which does not provide an inventive concept.
Claim 4 recites: obtaining device features for the target hardware configuration; generating layer group latency predictions for the layer groups based on the device features; and combining the layer group latency predictions to generate the latency prediction. The broadest reasonable interpretation of these steps is that the steps fall within the mental process groupings of abstract ideas because they cover concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally obtain device features, generate layer group latency predictions, and combine the layer group latency predictions.
This judicial exception is not integrated into a practical application and the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the limitation “using the second machine learning model” amounts to no more than mere instructions to apply the exception using a generic computer component. Even when considered in combination, this additional element represents mere instructions to implement an abstract idea or other exception on a computer, which does not provide an inventive concept.
Claim 5 recites the additional element: “wherein obtaining the device features for the target hardware configuration comprises receiving user-based input identifying one or more selected from the following: a hardware device identifier, a memory bus width, a memory clock rate, a number of cores, a number of stream-multiprocessors, and a compute clock rate”.
This judicial exception is not integrated into a practical application. This limitation is mere data gathering recited at a high level of generality, and thus is insignificant extra-solution activity. Even when viewed in combination, this additional element does not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitation of “receiving” is recited at a high level of generality. This element amounts to receiving or transmitting data over a network and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. Even when considered in combination, this additional element represents insignificant extra-solution activity, which does not provide an inventive concept.
Claim 6 recites: wherein each layer group is represented as an undirected graph to generate the layer group latency predictions. The broadest reasonable interpretation of this step is that the step falls within the mental process groupings of abstract ideas because is covers concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally represent layer groups as an undirected graph to generate layer group latency predictions.
This judicial exception is not integrated into a practical application and the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the limitation “when processed by the second machine learning model” amounts to no more than mere instructions to apply the exception using a generic computer component. Even when considered in combination, this additional element represents mere instructions to implement an abstract idea or other exception on a computer, which does not provide an inventive concept.
Claim 7 recites: determining a kernel for each layer group. The broadest reasonable interpretation of this step is that the step falls within the mental process groupings of abstract ideas because is covers concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally determine a kernel for each layer group.
This judicial exception is not integrated into a practical application because the limitation “using the second machine learning model” amounts to no more than mere instructions to apply the exception using a generic computer component. Further, the limitation “providing an indication of the kernel for each layer group for presentation” is mere data output recited at a high level of generality, and thus is insignificant extra-solution activity. Even when viewed in combination, these additional elements do not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitation of “providing” is recited at a high level of generality. This element amounts to receiving or transmitting data over a network or presenting offers and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. The limitation “using the second machine learning model” amounts to no more than mere instructions to apply the exception using a generic computer component. Even when considered in combination, these additional elements represent mere instructions to implement an abstract idea or other exception on a computer and insignificant extra-solution activity, which do not provide an inventive concept.
Claim 8 recites the additional element: “providing the latency prediction for presentation on a user device.” This judicial exception is not integrated into a practical application because the “providing an indication of the kernel for each layer group for presentation” is mere data output recited at a high level of generality, and thus is insignificant extra-solution activity. Even when viewed in combination, this additional element does not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitation of “providing” is recited at a high level of generality. This element amounts to receiving or transmitting data over a network or presenting offers and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. Even when considered in combination, this additional element represents insignificant extra-solution activity, which does not provide an inventive concept.
Claim 9 recites: wherein the operations further comprise: providing a recommendation for the target hardware configuration based on the latency prediction satisfying a latency threshold. The broadest reasonable interpretation of this step is that the step falls within the mental process groupings of abstract ideas because is covers concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (or manually with pen and paper) provide a recommendation.
This judicial exception is not integrated into a practical application and the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the limitation “one or more computer storage media” amounts to no more than mere instructions to apply the exception using a generic computer component. Even when considered in combination, this additional element represents mere instructions to implement an abstract idea or other exception on a computer, which does not provide an inventive concept.
Claims 10 and 11 recite: generating an optimized graph representing the target neural network, the optimized graph including nodes representing the layer groups and edges between the nodes based on connections between the layer groups, wherein each node of the optimized graph provides an indication of one or more layers from the target neural network and an indication of a kernel predicted by the second machine learning model. The broadest reasonable interpretation of this step is that the step falls within the mental process groupings of abstract ideas because is covers concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (or manually with pen and paper) generate an optimized graph representing the target neural network.
This judicial exception is not integrated into a practical application because the “providing a graphical representation of the optimized graph for presentation” is mere data output recited at a high level of generality, and thus is insignificant extra-solution activity. Even when viewed in combination, this additional element does not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitation of “providing” is recited at a high level of generality. This element amounts to receiving or transmitting data over a network or presenting offers and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. Even when considered in combination, this additional element represents insignificant extra-solution activity, which does not provide an inventive concept.
Claim 12 recites: generating a graph representation of a target neural network, the graph representation including nodes representing layers of the target neural network and edges between the nodes representing connections between the layers in the target neural network; generate edge labels identifying the edges of the graph representation as fusible or not fusible based on layer features associated with the nodes; partitioning the graph into a plurality of sub-graphs based on the edge labels; generate a latency prediction for each sub-graph based on the layer features associated with each node in each sub-graph and devices features of a target hardware configuration; and generating a total latency prediction for the target neural network by aggregating the latency predictions for the sub-graphs. The broadest reasonable interpretation of these steps is that the steps fall within the mental process groupings of abstract ideas because they cover concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (with the aid of pen and paper) generate a graph representation, generate edge labels, partition the graph, generate a latency prediction for each sub-graph and generate a total latency prediction.
This judicial exception is not integrated into a practical application. The limitations “causing a first graph neural network model” and “causing a second graph neural network model” to perform recited tasks amount to no more than mere instructions to apply the exception using a generic computer. See MPEP 2106.05(f). Further, the limitations “causing a first graph neural network model” and “causing a second graph neural network model” also merely indicate a field of use or technological environment (neural networks) in which the judicial exception is performed and thus fail to add an inventive concept to the claims. See MPEP 2106.05(h). Even when viewed in combination, these additional elements do not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitations of “causing a first graph neural network model” and “causing a second graph neural network model” to perform recited tasks amount to no more than mere instructions to apply the exception using a generic computer component. Further, the limitations “causing a first graph neural network model” and “causing a second graph neural network model” also merely indicate a field of use or technological environment (neural networks) in which the judicial exception is performed and thus fail to add an inventive concept to the claims. See MPEP 2106.05(h). Even when considered in combination, these additional elements represent mere instructions to implement an abstract idea or other exception on a computer, which do not provide an inventive concept.
Claim 13 recites: generates node embeddings based on the layer features associated with the nodes; generates edge embeddings based on the node embeddings; and a linear model that generates the edge labels based on the node embeddings. The broadest reasonable interpretation of these steps is that the steps fall within the mental process groupings of abstract ideas because they cover concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (with the aid of pen and paper) generate node embeddings, generate edge embeddings, and label edges.
This judicial exception is not integrated into a practical application and the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the limitations “a graph attention (GAT) model” and “a long short-term memory (LSTM) model” amount to no more than mere instructions to apply the exception using a generic computer component. Even when considered in combination, these additional elements represent mere instructions to implement an abstract idea or other exception on a computer, which does not provide an inventive concept.
Claim 14 recites the additional element: “receiving the device features for the target hardware configuration by receiving user-based input identifying one or more selected from the following: a hardware device identifier, a memory bus width, a memory clock rate, a number of cores, a number of stream-multiprocessors, and a compute clock rate”.
This judicial exception is not integrated into a practical application. This limitation is mere data gathering recited at a high level of generality, and thus is insignificant extra-solution activity. Even when viewed in combination, this additional element does not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitation of “receiving” is recited at a high level of generality. This element amounts to receiving or transmitting data over a network and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. Even when considered in combination, this additional element represents insignificant extra-solution activity, which does not provide an inventive concept.
Claim 15 recites: select a kernel for each sub-graph. The broadest reasonable interpretation of this step is that the step falls within the mental process groupings of abstract ideas because is covers concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally select a kernel for each layer group.
This judicial exception is not integrated into a practical application because the limitation “causing the second graph neural network” amounts to no more than mere instructions to apply the exception using a generic computer component. Further, the limitation “providing an indication of the kernel for each sub-graph for presentation” is mere data output recited at a high level of generality, and thus is insignificant extra-solution activity. Even when viewed in combination, these additional elements do not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitation of “providing” is recited at a high level of generality. This element amounts to receiving or transmitting data over a network or presenting offers and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. The limitation “causing the second graph neural network” amounts to no more than mere instructions to apply the exception using a generic computer component. Even when considered in combination, these additional elements represent mere instructions to implement an abstract idea or other exception on a computer and insignificant extra-solution activity, which do not provide an inventive concept.
Claim 16 recites: providing a recommendation for the target hardware configuration based on the latency prediction satisfying a latency threshold. The broadest reasonable interpretation of this step is that the step falls within the mental process groupings of abstract ideas because is covers concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (or manually with pen and paper) provide a recommendation.
This judicial exception is not integrated into a practical application and the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no additional elements in claim 16.
Claim 17 recites: generating an optimized graph representing the target neural network, the optimized graph including nodes representing the sub-graphs and edges between the nodes based on connections between the sub-graphs, wherein each node of the optimized graph provides an indication of one or more layers from the target neural network and an indication of a kernel. The broadest reasonable interpretation of this step is that the step falls within the mental process groupings of abstract ideas because is covers concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (or manually with pen and paper) generate an optimized graph representing the target neural network.
This judicial exception is not integrated into a practical application because the “providing a graphical representation of the optimized graph for presentation” is mere data output recited at a high level of generality, and thus is insignificant extra-solution activity. Even when viewed in combination, this additional element does not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitation of “providing” is recited at a high level of generality. This element amounts to receiving or transmitting data over a network or presenting offers and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. Even when considered in combination, this additional element represents insignificant extra-solution activity, which does not provide an inventive concept.
Claim 18 recites: obtaining a graph representation of a target neural network, the graph representation including nodes representing layers of the target neural network and edges between the nodes representing connections between the layers in the target neural network; labeling the edges of the graph representation as fusible or not fusible to provide edge labels by: generating node embeddings based on layer features associated with the nodes in the graph representation, generating edge embeddings based on the node embeddings, and labeling the edges in the graph representation based on the edge embeddings to provide the edge labels; partitioning the graph into a plurality of sub-graphs based on the edge labels; generate a latency prediction and a kernel prediction for each sub-graph based on the layer features associated with each node in each sub-graph and the devices features of the target hardware configuration; and generating a total latency prediction for the target neural network by aggregating the latency predictions for the sub-graphs. The broadest reasonable interpretation of these steps is that the steps fall within the mental process groupings of abstract ideas because they cover concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (with the aid of pen and paper) obtain a graph representation; label edges by generating node embeddings and edge embeddings and labeling the edges; partition the graph; generate a latency prediction for each sub-graph; and generate a total latency prediction.
This judicial exception is not integrated into a practical application. The limitations “one or more processors; and one or more computer storage media”, “by a first graph neural network model”, “using a graph attention (GAT) model of the first graph neural network model”, “using a long short-term memory (LSTM) model of the first graph neural network model”, “using a linear model of the first graph neural network model”, “causing a second graph neural network model” to perform recited tasks amount to no more than mere instructions to apply the exception using a generic computer. See MPEP 2106.05(f). Further, these limitations also merely indicate a field of use or technological environment (neural networks) in which the judicial exception is performed and thus fail to add an inventive concept to the claims. See MPEP 2106.05(h). Lastly, “receiving device features for a target hardware configuration” is mere data gatherings recited at high level of generality such that is amounts to insignificant extra-solution activity. Even when viewed in combination, these additional elements do not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitation of “receiving” is recited at a high level of generality. This element amounts to receiving or transmitting data over a network and are well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. As discussed above, the recitations of “one or more processors; and one or more computer storage media”, “by a first graph neural network model”, “using a graph attention (GAT) model of the first graph neural network model”, “using a long short-term memory (LSTM) model of the first graph neural network model”, “using a linear model of the first graph neural network model”, “causing a second graph neural network model” to perform recited tasks amount to no more than mere instructions to apply the exception using a generic computer component. Further, these limitations also merely indicate a field of use or technological environment (neural networks) in which the judicial exception is performed and thus fail to add an inventive concept to the claims. See MPEP 2106.05(h). Even when considered in combination, these additional elements represent mere instructions to implement an abstract idea or other exception on a computer and insignificant extra-solution activity, which do not provide an inventive concept.
Claim 19 recites: generating an optimized graph representing the target neural network, the optimized graph including nodes representing the sub-graphs and edges between the nodes based on connections between the sub-graphs, wherein each node of the optimized graph provides an indication of one or more layers from the target neural network and an indication of the predict kernel for the sub-graph represented by the node. The broadest reasonable interpretation of this step is that the step falls within the mental process groupings of abstract ideas because is covers concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (or manually with pen and paper) generate an optimized graph representing the target neural network.
This judicial exception is not integrated into a practical application because the “providing a graphical representation of the optimized graph for presentation” is mere data output recited at a high level of generality, and thus is insignificant extra-solution activity. Even when viewed in combination, this additional element does not integrate the recited judicial exception into a practical application, and the claim is directed to the judicial exception.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, the recitation of “providing” is recited at a high level of generality. This element amounts to receiving or transmitting data over a network or presenting offers and is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. Even when considered in combination, this additional element represents insignificant extra-solution activity, which does not provide an inventive concept.
Claim 20 recites: providing a recommendation for the target hardware configuration based on the latency prediction satisfying a latency threshold. The broadest reasonable interpretation of this step is that the step falls within the mental process groupings of abstract ideas because is covers concepts performed in the human mind, including observation, evaluation, judgment, and opinion. See MPEP 2106.04(a)(2), subsection III. For example, a user can mentally (or manually with pen and paper) provide a recommendation.
This judicial exception is not integrated into a practical application and the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no additional elements in claim 20.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1 and 9 are rejected under 35 U.S.C. 102(a)(1) and (a)(2) as being anticipated by Hsu et al. (US 20240119283 A1) (‘Hsu’).
With respect to claim 1, Hsu teaches one or more computer storage media storing computer-useable instructions (paragraphs 5 and 32) that, when used by one or more computing devices, cause the one or more computing devices to perform operations, the operations comprising:
receiving a neural network representation for a target neural network having a plurality of layers (Fig. 2; paragraph 18);
grouping, using a first machine learning model, layers of the target neural network to provide a plurality of layer groups based on the neural network representation, at least one layer group comprising multiple layers from the target neural network that can be executed by a single operation (paragraphs 17-18); and
generating, using a second machine learning model, a latency prediction for executing the target neural network on a target hardware configuration based on the layer groups (paragraphs 14 and 22).
With respect to claim 9, Hsu teaches wherein the operations further comprise: providing a recommendation for the target hardware configuration based on the latency prediction satisfying a latency threshold (i.e. optimal configuration) (paragraphs 21, 23, and 27).
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.
Claims 2, 4, 6, 8, 10-12, 16 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Hsu et al. (US 20240119283 A1) (‘Hsu’) in view of Apparao et al. (US 2019/0266504 A1) (‘Apparao’).
With respect to claim 2, Hsu teaches wherein grouping the layers of the target neural network using the first machine learning model comprises: obtaining a graph representing the target neural network, the graph including nodes representing the layers of the target neural network and edges between the nodes based on connections between the layers of the target neural network (Fig. 2, paragraph 18).
Hsu does not explicitly teach causing the first machine learning model to label each edge in the graph as fusible or not fusible to provide edge labels; and generating the layer groups by dividing the graph into sub-graphs based on the edge labels.
Apparao teaches obtaining a graph representing the target neural network, the graph including nodes representing the layers of the target neural network and edges between the nodes based on connections between the layers of the target neural network (Fig. 3B; paragraph 19);
causing the first machine learning model to label each edge in the graph as fusible or not fusible to provide edge labels (paragraphs 28, 30 and 87); and
generating the layer groups by dividing the graph into sub-graphs based on the edge labels (paragraphs 30, 33, 35-36).
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the invention to have modified Hsu to label edges as taught by Apparao to aid in determination of computational costs, which are used to determine an optimal distribution of the computational workload involved in running the one or more operations among multiple hardware executors (Apparao, abstract and paragraph 28).
With respect to claim 4, Hsu teaches wherein generating the latency prediction using the second machine learning model comprises:
obtaining configuration (paragraphs 14-15); and
generating, using the second machine learning model, layer group latency predictions for the layer groups based on the configuration (paragraphs 14 and 23).
Hsu does not explicitly teach obtaining device features for the target hardware configuration; generating, using the second machine learning model, layer group latency predictions for the layer groups based on the device features; or combining the layer group latency predictions to generate the latency prediction.
Apparao teaches obtaining device features for the target hardware configuration (paragraphs 14-16 and 33);
generating, using the second machine learning model, layer group latency predictions for the layer groups based on the device features (paragraphs 30-31); and
combining the layer group latency predictions to generate the latency prediction (paragraphs 31-33).
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the invention to have modified Hsu to use device features to predict latency as taught by Apparao to aid in determination of computational costs, which are used to determine an optimal distribution of the computational workload involved in running the one or more operations among multiple hardware executors (Apparao, abstract and paragraph 30).
With respect to claim 6, Hsu in view of Apparao teaches wherein each layer group is represented as an undirected graph when processed by the second machine learning model to generate the layer group latency predictions (Hsu, paragraphs 14 and 22; Apparao, paragraphs 19, 22-23 and 33).
With respect to claim 8, Hsu teaches providing the latency prediction (paragraphs 14 and 22).
Hsu does not explicitly teach providing the latency prediction for presentation on a user device.
Apparao teaches providing the latency prediction for presentation on a user device (paragraphs 22 and 60).
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the invention to have modified Hsu to display results as taught by Apparao to enable incorporation of user involvement (Apparao, paragraph 19) and an improved user experience such as faster responses and presentation of results (Apparao, paragraphs 17 and 30).
With respect to claim 10, Hsu teaches wherein the operations further comprise: generating an optimized graph representing the target neural network, the optimized graph including nodes representing the layer groups and edges between the nodes based on connections between the layer groups (Fig. 2; paragraph 18).
Hsu does not explicitly teach providing a graphical representation of the optimized graph for presentation.
Apparao teaches generating an optimized graph representing the target neural network, the optimized graph including nodes representing the layer groups and edges between the nodes based on connections between the layer groups (Fig. 3B; paragraphs 19 and 29); and
providing a graphical representation of the optimized graph for presentation (paragraphs 22 and 60).
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the invention to have modified Hsu to display the representation as taught by Apparao to enable incorporation of user involvement (Apparao, paragraph 19) and an improved user experience such as faster responses and presentation of results (Apparao, paragraphs 17 and 30).
With respect to claim 11, Hsu in view of Apparao teaches wherein each node of the optimized graph provides an indication of one or more layers from the target neural network and an indication of a kernel predicted by the second machine learning model (Hsu, Fig. 2; paragraph 18; Apparao, Figs. 3A-3B, paragraphs 22 and 60).
With respect to claim 12, Hsu teaches a computer-implemented method comprising:
generating a graph representation of a target neural network, the graph representation including nodes representing layers of the target neural network and edges between the nodes representing connections between the layers in the target neural network (Fig. 2, paragraph 18); and
causing a second graph neural network model to generate a latency prediction for each sub-graph based on the layer features associated with each node in each sub-graph (paragraphs 14 and 22).
Hsu does not explicitly teach causing a first graph neural network model to generate edge labels identifying the edges of the graph representation as fusible or not fusible based on layer features associated with the nodes; partitioning the graph into a plurality of sub-graphs based on the edge labels; generating latency predictions based on devices features of a target hardware configuration; or generating a total latency prediction for the target neural network by aggregating the latency predictions for the sub-graphs.
Apparao teaches a graph representing the target neural network, the graph including nodes representing the layers of the target neural network and edges between the nodes based on connections between the layers of the target neural network (Fig. 3B; paragraph 19);
causing a first graph neural network model to generate edge labels identifying the edges of the graph representation as fusible or not fusible based on layer features associated with the nodes (paragraphs 28, 39, and 87);
partitioning the graph into a plurality of sub-graphs based on the edge labels (paragraphs 30, 33, 35-56);
generating latency predictions (paragraphs 30-31) based on devices features of a target hardware configuration (paragraphs 14-16 and 33); and
generating a total latency prediction for the target neural network by aggregating the latency predictions for the sub-graphs (paragraphs 31-33).
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the invention to have modified Hsu to generate label edges and total latency prediction as taught by Apparao to aid in determination of computational costs, which are used to determine an optimal distribution of the computational workload involved in running the one or more operations among multiple hardware executors (Apparao, abstract and paragraph 28).
With respect to claim 16, Hsu in view of Apparao teaches wherein the operations further comprise: providing a recommendation for the target hardware configuration based on the latency prediction satisfying a latency threshold (i.e. optimal configuration) (paragraphs 21, 23, and 27; Apparao – requirements in paragraph 14).
With respect to claim 17, Hsu in view of Apparao teaches generating an optimized graph representing the target neural network, the optimized graph including nodes representing the sub-graphs and edges between the nodes based on connections between the sub-graphs, wherein each node of the optimized graph provides an indication of one or more layers from the target neural network and an indication of a kernel (Hsu, Fig. 2, paragraph 18; Apparao, Figs. 3A-3B, paragraphs 19, 22, 29 and 60); and
providing a graphical representation of the optimized graph for presentation (Apparao, paragraphs 22 and 60).
With respect to claim 19, Hsu in view of Apparao teaches generating an optimized graph representing the target neural network, the optimized graph including nodes representing the sub-graphs and edges between the nodes based on connections between the sub-graphs, wherein each node of the optimized graph provides an indication of one or more layers from the target neural network and an indication of the predict kernel for the sub-graph represented by the node (Hsu, Fig. 2, paragraph 18; Apparao, Figs. 3A-3B, paragraphs 19, 22, 29 and 60); and
providing a graphical representation of the optimized graph for presentation (Apparao, paragraphs 22 and 60).
Claims 5 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Hsu et al. (US 20240119283 A1) (‘Hsu’) in view of Apparao et al. (US 2019/0266504 A1) (‘Apparao’) as applied to claim 4 above, and further in view of Rao et al. (US 2023/0091667 A1) (‘Rao’).
With respect to claim 5, Hsu in view Apparao teaches receiving user input (Apparao, paragraph 19).
Hsu in view of Apparao does not explicitly teach wherein obtaining the device features for the target hardware configuration comprises receiving user-based input identifying one or more selected from the following: a hardware device identifier, a memory bus width, a memory clock rate, a number of cores, a number of stream-multiprocessors, and a compute clock rate.
Rao teaches wherein obtaining the device features for the target hardware configuration comprises receiving user-based input identifying one or more selected from the following: a hardware device identifier, a memory bus width, a memory clock rate, a number of cores, a number of stream-multiprocessors, and a compute clock rate (paragraph 50).
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the invention to have further modified Hsu in view of Apparao to receive user input as taught by Rao to enable user specification of performance indicators, and thus provide a more customized operation.
With respect to claim 14, Hsu in view Apparao teaches receiving user input (Apparao, paragraph 19).
Hsu in view of Apparao does not explicitly teach receiving the device features for the target hardware configuration by receiving user-based input identifying one or more selected from the following: a hardware device identifier, a memory bus width, a memory clock rate, a number of cores, a number of stream-multiprocessors, and a compute clock rate.
Rao teaches receiving the device features for the target hardware configuration by receiving user-based input identifying one or more selected from the following: a hardware device identifier, a memory bus width, a memory clock rate, a number of cores, a number of stream-multiprocessors, and a compute clock rate (paragraph 50).
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the invention to have further modified Hsu in view of Apparao to receive user input as taught by Rao to enable user specification of performance indicators, and thus provide a more customized operation.
Claims 7 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Hsu et al. (US 20240119283 A1) (‘Hsu’) in view of Apparao et al. (US 2019/0266504 A1) (‘Apparao’) as applied to claim 4 above, and further in view of Hong et al. (US 11301762 B1) (‘Hong’).
With respect to claim 7, Hsu in view of Apparao teaches providing an indication of the kernel for each layer group for presentation (Apparao, paragraph 60).
Hsu in view of Apparao does not explicitly teach wherein the operations further comprise: determining, using the second machine learning model, a kernel for each layer group.
Hong teaches wherein the operations further comprise: determining, using the second machine learning model, a kernel for each layer group (paragraph 71, 77, 85, 92, and 98).
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the invention to have further modified Hsu to determine a kernel as taught by Hong to enable selection of the most efficient kernel, thereby improving performance and maximizing the resource use (Hong, paragraph 57).
With respect to claim 15, Hsu in view of Apparao teaches providing an indication of the kernel for each sub-graph for presentation (Apparao, paragraph 60).
Hsu in view of Apparao does not explicitly teach causing the second graph neural network to select a kernel for each sub-graph.
Hong teaches causing the second graph neural network to select a kernel for each sub-graph (paragraph 71, 77, 85, 92, and 98).
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the invention to have further modified Hsu to select a kernel as taught by Hong to enable selection of the most efficient kernel, thereby improving performance and maximizing the resource use (Hong, paragraph 57).
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALICIA M WILLOUGHBY whose telephone number is (571)272-5599. The examiner can normally be reached 9-5:30, EST, M-F.
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, Ajay Bhatia can be reached at 571-272-3906. 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.
/ALICIA M WILLOUGHBY/Primary Examiner, Art Unit 2156 August 14, 2026
Prosecution Insights