Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&DMRT adjusts the implementation of Inference, uses TVM underlying Pass to perform sampling and output intermediate results, and especially judges some unrealized operators;MRT solves the problem of Circom Shape adaptation;MRT code refactoring, abstract parent class and virtual function;ZkRollupInvestigate the test run of cuda bellman;Investigate plonk, lookup and plonkup and test the impleme...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&DMRT adjusts the implementation of Inference, uses TVM underlying Pass to perform sampling and output intermediate results, and especially judges some unrealized operators;MRT solves the problem of Circom Shape adaptation;MRT code refactoring, abstract parent class and virtual function;ZkRollupInvestigate the test run of cuda bellman;Investigate plonk, lookup and plonkup and test the impleme...

IntroductionBig data has provided tremendous insights that enable machine learning and neural networks. However, artificial intelligence models need extra caution when applied to real-world scenarios as AI models could cause incorrect or misleading results in critical applications. Furthermore, AI model execution could also take up too many resources including memory, disk, and CPU/GPU. Not to mention that the results themselves cannot be reproduced for verification in most cases. All these s...

IntroductionBig data has provided tremendous insights that enable machine learning and neural networks. However, artificial intelligence models need extra caution when applied to real-world scenarios as AI models could cause incorrect or misleading results in critical applications. Furthermore, AI model execution could also take up too many resources including memory, disk, and CPU/GPU. Not to mention that the results themselves cannot be reproduced for verification in most cases. All these s...

Technical Updates 🤖*Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D ZkRollupZKMatrix adjusts the wallet plug-in logic and interfaces display;ZkNode optimizes RPC link logic to increase stability;The ZK Section 9 project participated in ETHSanFrancisco 2022, won 3 awards including Finalist and open-sourced the entry code: https://github.com/zk-ml/uchikoma/tree/ethsf-2022;ZKML research runs through PyTorch training, TVM IR analysis, Circom translator, and o...

Technical Updates 🤖*Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D ZkRollupZKMatrix adjusts the wallet plug-in logic and interfaces display;ZkNode optimizes RPC link logic to increase stability;The ZK Section 9 project participated in ETHSanFrancisco 2022, won 3 awards including Finalist and open-sourced the entry code: https://github.com/zk-ml/uchikoma/tree/ethsf-2022;ZKML research runs through PyTorch training, TVM IR analysis, Circom translator, and o...

Cortex Labs has officially launched ZkMatrix, a sidechain transaction acceleration solution. The project use ZkRollup technology (zero-knowledge proof) to consolidate batch transactions on the side chain to one transaction on the main chain, which can increase the TPS and throughput of the blockchain while significantly lowering the Endorphin Fee consumed by each transaction (same as Gas Fee for Ethereum). Furthermore, all data from this technology is saved on the first-layer chain, and the s...

Cortex Labs has officially launched ZkMatrix, a sidechain transaction acceleration solution. The project use ZkRollup technology (zero-knowledge proof) to consolidate batch transactions on the side chain to one transaction on the main chain, which can increase the TPS and throughput of the blockchain while significantly lowering the Endorphin Fee consumed by each transaction (same as Gas Fee for Ethereum). Furthermore, all data from this technology is saved on the first-layer chain, and the s...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT/CVMMRT implements Calibrator, FuseBatchNorm, FuseAvgPool PassMRT implements Sampling, Discretor, and Quantizer Pass, using symmetric linear discrete method to quantify the modelImplemented FixPoint Pass to export the model to TVM Relay Expr, all the data are shapedRealized the disk import and export function of MRT internal Trace, and quickly imported the last running resultsZKRollUpTh...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT/CVMMRT implements Calibrator, FuseBatchNorm, FuseAvgPool PassMRT implements Sampling, Discretor, and Quantizer Pass, using symmetric linear discrete method to quantify the modelImplemented FixPoint Pass to export the model to TVM Relay Expr, all the data are shapedRealized the disk import and export function of MRT internal Trace, and quickly imported the last running resultsZKRollUpTh...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT/CVMMRT, TVM and Circom unified representation structure, where MRT uses built-in InferType and InferShapeAdd MRT operator to Circom's parser, and adapt to the shape adaptation of operators such as clipIncrease the export of Circom input parameters. The tested resnet data input size exceeds 100M. Consider integrating some parameters into the ZKP operator to reduce input consumption...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT/CVMMRT, TVM and Circom unified representation structure, where MRT uses built-in InferType and InferShapeAdd MRT operator to Circom's parser, and adapt to the shape adaptation of operators such as clipIncrease the export of Circom input parameters. The tested resnet data input size exceeds 100M. Consider integrating some parameters into the ZKP operator to reduce input consumption...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT & CVMMRT test ONNX backend model conversion and quantization;ZkRollupZkMatrix update the front-end operation logic, interactive functions and display interface, and adapt to the mobile terminalCortexLabs partners with ETHF, building ZkMatrix for themZkProver GPU version test is stable, and the long-running memory leak problem is solvedZkProver has improved compilation options to suppor...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT & CVMMRT test ONNX backend model conversion and quantization;ZkRollupZkMatrix update the front-end operation logic, interactive functions and display interface, and adapt to the mobile terminalCortexLabs partners with ETHF, building ZkMatrix for themZkProver GPU version test is stable, and the long-running memory leak problem is solvedZkProver has improved compilation options to suppor...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D ZkRollupZkMatrix live on Cortex: https://zkmatrix.cortexlabs.aiZkNode fixes multi-Prover parallel task problemZkNode adds the mainnet scan block option, quickly starts ZkRollup, and increases database index to speed up query performanceZkProver fixes GCC build issues and updates command line optionsZkWallet fixes display problems, front-end logic processingCortex Full NodeA new model stora...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D ZkRollupZkMatrix live on Cortex: https://zkmatrix.cortexlabs.aiZkNode fixes multi-Prover parallel task problemZkNode adds the mainnet scan block option, quickly starts ZkRollup, and increases database index to speed up query performanceZkProver fixes GCC build issues and updates command line optionsZkWallet fixes display problems, front-end logic processingCortex Full NodeA new model stora...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT/CVMMRT updated the accuracy test code, adjusted and tested the model output results of the sampling dataAdjust the parameter setting of Calibrate Pass: absmax, fix the problem of FuseBatchnormRefactored the Quantiza Pass code and fixed the Infer Precision issueFixed the FixPoint Infer Type problem and added subgraph verification functionThe model parameters are adjusted to Int8, which ...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT/CVMMRT updated the accuracy test code, adjusted and tested the model output results of the sampling dataAdjust the parameter setting of Calibrate Pass: absmax, fix the problem of FuseBatchnormRefactored the Quantiza Pass code and fixed the Infer Precision issueFixed the FixPoint Infer Type problem and added subgraph verification functionThe model parameters are adjusted to Int8, which ...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT/CVMTVM/MRT set CVM Python dependent library, and export Symbol format to CVM-Runtime;MRT repairs the precision test logic and formats the precision output style;MRT adds Symbol to Circom compilation adaptation code;CVM adjusts C API interface parameters;ZkRollupRealize the transaction verification of EIP-712;State separation (part1 is provided to the front-end for fast update without c...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT/CVMTVM/MRT set CVM Python dependent library, and export Symbol format to CVM-Runtime;MRT repairs the precision test logic and formats the precision output style;MRT adds Symbol to Circom compilation adaptation code;CVM adjusts C API interface parameters;ZkRollupRealize the transaction verification of EIP-712;State separation (part1 is provided to the front-end for fast update without c...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT/CVMMRT/CVM began to connect with TVM to realize automatic conversion and deployment, and improved the unified interface protocol to facilitate the subsequent ZKML design;MRT implements the Symbol representation corresponding to RelayExpr and supports mutual conversion;MRT implements the Calibrator at the operator level, and obtains the intermediate results of the model;Built a Python T...

Technical Updates 🤖Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D MRT/CVMMRT/CVM began to connect with TVM to realize automatic conversion and deployment, and improved the unified interface protocol to facilitate the subsequent ZKML design;MRT implements the Symbol representation corresponding to RelayExpr and supports mutual conversion;MRT implements the Calibrator at the operator level, and obtains the intermediate results of the model;Built a Python T...