ZKML

2024-01-02 Publish [ZKML release 1.0.1] (https://github.com/CortexFoundation/tachikoma/commits/zkml-v1.0.1/)

In this release, we implemented the general ZKML framework and passed the mnist model for test.

2024-01-16 Publish fixed commits for ZKML

Fixed some classification models' quantization bugs, including mobilenet, squeezenet, etc.

ZKCVM

The Cortex team has been closely following the latest zkEVM technology, which is one of the goals in the ZK trilogy (zkTX -> zkVM -> zkML).

2024-02-27 We modified some codes based on the ZK-Sync source to adapt CVM blockchain, project are organized at (https://github.com/CortexFoundation/zkcvm-mono/tree/main). We started to research and benchmark more ZK-EVM solutions

2024-03-26 Pass the ZKEVM cpu zk-rollup process.

2024-05-14 The GPU proof generation code is fixed and test pass through.

OPML

Due to the ZKP memory consumption and long-time proof generation, we transformed to research and develop the OPML technology for ZKML project. The HyperOracle's OPML project is compatible with ZKML, The last ZK-Proof step, which aims to exchange the neccessary layer2 data with layer one blockchain, can be replaced with Optimistic Machine Learning method.

Until now, we have dived into the details of OPML and plan to migrate the core method to CortexLabs' CVM-Runtime project.

2024-05-21 The early process of mips VM compilation has been uploaded to [the CVMRuntime official project] (https://github.com/CortexFoundation/cvm-runtime/commits/mips/).

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

MRT/CVM

1. MRT has added support for the Detection model, with quantization already implemented for models such as Yolo, SSD, etc.

2. Implemented automatic Spliter/Merger algorithm logic to eliminate the problem of duplicates in segmented subgraphs (increased reference counting).

3. Added constant pass fusion for operators oneslike and zeroslike.

4. Added adaptation and quantization logic for operators shape of, reverse, maximum, and minimum.

5. Adapted Circom for the detection model, added implementation for operators like slice_like, repeat, tile, etc.

6. Conducted research on OPML technology and its implementation, exploring more possibilities for on-chain AI inference.

ZkRollup

1. Implement a compression process for contract bytecode to reduce on-chain data consumption, and conduct compression testing.

2. Introduce StorageProcessor as an access layer to the PostgreSQL database, managing pools and transactions.

3. Implement the state's persistent storage layer based on the database (DB).

4. Utilize RocksDB to store some state-related data, serving as the storage for state during the VM execution process (enhanced speed).

5. Abstract StorageView, enabling the use of RocksDB, PostgreSQL, and in-memory storage as storage points, implementing ReadStorage and WriteStorage.

6. Fix the issue in the Metric exporter where metric data submission was not executed upon receiving a signal, and adjust the maximum retry attempts metric for the job processor, submitting the metric at the beginning of the job.

7. Add a simple implementation of Memory for the VM.

Cortex Full Node

1. build: make linter emit output;

2. p2p/discover: add liveness check in collectTableNodes (#28686);

3. discard work submit to full channel;

4. cuckoo consensus test & avoid using block override;

5. infer mem/gas/opcodes table added;

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Adjust the Trace API interface, encapsulate the transform pass, and add fuse, calibrate, quantize and export methods;

Trace API adds dataset, statistic binding, and accuracy eval interface;

circom adapts alexnet, dense, repairs padding shape, concat operator and other issues;

Fix dropout + tuplegetitem tvm running error, fix use too much shared data video memory error;

ZKML supports shufflenet, densenet, vgg, squeezenet, and the accuracy error of the original floating point model is 1%~2%;

ZkRollup

Implement data reading and modification in the memory and storage parts of the VM;

Perform different data processing according to different Operand types (register, stack, code area, etc.);

Implement the basic processing flow of a VM cycle: fetch, decode, and execute;

Implement the specific logic of basic arithmetic instructions (addition, subtraction, multiplication and division), binary operations, shift operations, vm context operations and pointer operations;

Cortex Full Node

cmd/evm: fix some issues with the evm run command (#28109);

trie: add getter for preimage store in trie.Database (#28155);

cmd/devp2p: use bootnodes as crawl input (#28139);

cmd/geth, internal/flags: print envvar config source and bad names (#28119);

metrics: add support for enabling metrics from env vars (#28118);

Community Updates👨‍👩‍👦‍👦

1. R&D on cross-chain bridge ecosystem;

2. Assist CVM in testing;

3. Collaboration with several NFT Artists;

4. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Machine-Learning(ML), CortexVirtual Machine(CVM)

1. ZKML optimizes ResNet model accuracy, improving accuracy by 14%

2. ZKML adjusts the Trace API and adds interfaces for data set binding, preprocessing, and postprocessing information.

3. ZKML completes the data set ImageNet, coco and other interfaces, and downloads more data sets

4. ZKML investigates ezkl, runs through the process and compares ZKML Circom benchmark performance

ZkRollup

1. Implementing the opcode of zkcvm

2. Try to implement the execution layer of zkcvm (modify consensus, block generation, smt, etc.)

3. Investigate and run through the new poseidon hash implementation and poseidon-hash circuit in halo2

4. Research and follow up on the latest zkevm development (cariovm, fuelvm, syncvm, midenvm)

5. OPCode that implements zkcvm

Cortex Full Node

1. core/rawdb: allocate database keys with explicit size to avoid slice

2. core: ensure txindex will be triggered at least once (#27847)

3. ignore genesis block on importChain

4. trie: reduce allocs in recHash (#27770)

5. all: update golang/x/ext and fix slice sorting fallout (#27909)

Community Updates👨‍👩‍👦‍👦

1. R&D on cross-chain bridge ecosystem;

2. Compiling the current ZKML project R&D report;

3. Collaboration with several NFT Artists;

4. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM)

1. ZKML trains RNN language sequence prediction model;

2. ZKML subgraph interface storage is optimized, and parent and child traces share node pointers (hash);

3. ZKML investigates accuracy parameters and tests the impact of parameters on the quantitative model;

4. ZKML Make config class for quantization procedure(with statement);

ZkRollup

1. The test ran boojum's sha256 circuit. It succeeded (very quickly) when the preimage < 2kb, but would get stuck at 4kb and 8kb;

2. L1Client is divided into querying client and signing client to interact with L1 and fix errors in other modules caused by reconstruction;

3. add the generic support for RocksDB storage;

4. Some chore refactor: migrate workspace dependencies, use basic_type::web3 as a replacement for direct ethers-rs;

5. The circuit benchmark results are very poor (release mode), you need to consider switching to boojum (fri) or halo2;

Cortex Full Node

1. core/state: move UpdateContractCode before the trie hash is computedore/state: move UpdateContractCode before the trie hash is computed;

2. p2p: move ping handling into pingLoop goroutine (#27887);

3. all: remove trailing whitespace (#27741);

4. build: upgrade -dlgo version to Go 1.20.7 (#27835);

5. cmd/bootnode: fix timer leak (#27754);

Community Updates👨‍👩‍👦‍👦

1. R&D the current progress of ZKML;

2. R&D the technical progress of the current MPC wallet;

3. R&D the cross-chain bridge ecosystem;

4. Collaboration with several NFT Artists;

5. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Machine-Learning(ML), CortexVirtual Machine(CVM)

1. Complete the function of coco, voc, cifar10 dataset loading;

2. circom generator RightShift and MulScalar parameter alignment. Type annotations, Symbol type confusion fixes;

3. Generate graph optimization, lose invalid nodes after topological sorting. scalar is integrated into op as an attribute, the original parameter is still output to input.json, fix circom compilation repair;

4. Increase NLP model training, long sequence numerical prediction, convert fixed-point model and debug problems;

ZkRollup

1. Complete the reconstruction of zkx-contract into diamonds-proxy architecture, and update the deployment script and tool script;

2. Adjust L1 node and postgre docker startup (convenient for script startup and shutdown);

3. Golang implements the zkx-cli infrastructure tool: it is used for more convenient startup of development;

4. Update the contract and config crates to adapt to the new contract;

5. Fix the problem of sending too many tx requests and Nonce mismatch;

6. Overall reconstruction of zkx-contract from the original architecture (common proxy + contract split (including addtional)) to EIP-2535;

7. Investigate the new zksync-era system PLONK + FRI: boojum;

Cortex Full Node

1. crypto/secp256k1: define NDEBUG only if not defined #1600

2. update leveldb metrics collection and track state changes in state db

3. database put in sync mode

4. common/lru: add test case for BasicLRU.Peek

5. core: log message if DeriveFields returns error

6. send big tx by announce only

7. make signature of ContractCode hash-independent

8. add cortex robot

Community Updates👨‍👩‍👦‍👦

1. R&D the current progress of ZKML;

2. R&D the technical progress of the current MPC wallet;

3. Collaboration with several NFT Artists

4. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM)

1. The location and parameters of the logs interface in the design and test model quantification process

2. Fix bugs in zkml, and adjust and optimize the code

3. Implement the interface of image input_data to scale and round separately

4. After the model is converted to circom, modify input.json and set the input value

5. Resnet18 resnet34 resnet50 resnet101 resnet152 format converted successfully

ZkRollup

1. Implement the js sdk that interacts with zkx netwrok: zkx.js.

2. Use handlebars to re-render the solidity code of single verifier and aggregate verifier

3. Add health check api to monitor the availability of database and web3 connections

4. Modify the process of price fetcher getting data from third-party api (coingecko, coinmarketcap...), add a validator to verify whether the token can be used for zkx network transaction fees

Cortex Full Node

1. crypto/secp256k1: define NDEBUG only if not defined #1600

2. update leveldb metrics collection and track state changes in state db

3. database put in sync mode

4. common/lru: add test case for BasicLRU.Peek

5. core: log message if DeriveFields returns error

6. send big tx by announce only

7. make signature of ContractCode hash-independent

8. add cortex robot

Community Updates👨‍👩‍👦‍👦

1. R&D the current progress of ZKML;

2. R&D the technical progress of the current MPC wallet;

3. Collaboration with several NFT Artists

4. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM)

1. ZKML adds TVM Runtime Inference reasoning simulation test quantized accuracy, and debugs the loss of accuracy;

2. ZKML adjusts the data set mount interface and output result analysis class;

3. ZKML adjusts the restore logic of the quantitative model output results;

4. Fix the problem of input_data not passing mrt transformer in model2circom;

5. ZKML adjusts the data set mount, and increases the data set input in the calibrated pass to improve the quantization accuracy;

6. Requantize bit digit adjustment, binary operators such as add/sub/mul/div are adjusted to 30/31, 16/16;

ZkRollup

1. Implement zkx-crypto (signature, verification, hash, etc.) related functions to wasm through Binaryen and wasm-pack;

2. Add batch tx, including processing it in mempool, statekeeper;

3. Realize batch fee calculation, add batch tx-related api interface and fix fee calculation problem;

4. The tx submitted in the Forced exit request is changed to batch tx to ensure success or failure at the same time;

5. Create the schema of the entire database and create the corresponding index (sqlx migration);

Cortex Full Node

1. hash less func added

2. narrow down the scope of the variable

3. consensus/clique: use slices package for sorting (#27488)

4. metrics: use slices package for sorting (#27493)

5. cmd/devp2p: use slices package for sorting (#27487)

6. accounts/keystore: use slices package for sorting (#27485)

Community Updates👨‍👩‍👦‍👦

1. NFT Gallery Powered by Cortex is now available

   https://t.co/VBJl9lCrPs

2. Collaboration with several NFT Artists

3. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

4. Working with Ledger integrations

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM)

1. MRT adds BN + Dense, and quantitative support for BN equivalent conversion;

2. MRT adds Dropout support and automatically fuse TupleGetItem;

3. Fix the symbol name problem: cause circom code compilation error (including '%', '.', etc. cause the variable name to be ambiguous during compilation);

4. MRT organizes existing work, codes, deletes invalid codes, writes documents, including TEST environment preparation, running scripts and reference links;

5. MRT already supports OP: nn.adaptive_avg_pool2d, TupleGetItem, nn.batch_norm, add, nn.bias_add, var, nn.dense, nn.conv2d, nn.max_pool2d, reshape, nn.relu, squeeze, nn.dropout;

ZkRollup

1. Generate all vk (exodus key, block key, recursive key) and generate the corresponding contract vk;

2. Zkx-contracts: plonk core, governance, verifier (finally divided into two separate verifiers, contract size limit);

3. Zkx-contracts: gatekeeper management contract upgrade, implement zkx main contract (main + additional part), add zkx NFT factory and forced exit contract, complete hardhat script to read conf and deploy contract;

Cortex Full Node

1. core/state/snapshot: check difflayer staleness early (#27255);

2. rpc: websocket should respect the HTTP_PROXY by default;

Community Updates👨‍👩‍👦‍👦

1. Updated MRT docs

2. Cortex NFT Gallery's(Thalamus) final internal testing is ongoing

3. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

4. Connecting several NFT artists for collabs

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM)

1. ZKML circom adapts conv2d padding: insert pad2d op before conv2d in symbol;

2. Circom modified conv2d inputs & shape to adapt to the calculation of upper and lower nodes in the topology map;

3. Circom fixes the problem that stride=1 is hard-coded to 1, adds stride support and derivation, and requires the stride of xy to be consistent;

4. ZKML adds support for cvm_right_shift, cvm_clip, etc., and converts bias_add to conv operator;

ZkRollup

1. Add l1 and l2 signature check services (for api submission check (double verification)), tx_sender is used for api submission tx;

2. Implement the fee ticker for the calculation of transaction fees;

3. Implement the api services of accounts, block, transactions, config, fee, token, status;

4. Realize the circuit of block state transition, emergency exit;

5. Add some tool functions for zk-prover and create the main entry function of zkx_proof and exit_proof;

6. Define and implement some api types (api_block, api_tx, api_account..), add a simple memory cache for api query token related information;

Cortex Full Node

1. Cortex full node release 1.10.41 stable version; cmd/devp2p: make crawler-route53-updater less verbose (#27116); p2p: access embedded fields of Server directly (#27078); core: fix comment to reflect function name (#27070); cmd/devp2p: fix erroneous log output in crawler (#27089);

Community Updates👨‍👩‍👦‍👦

1. By combining the CTXC public chain, we can explore more possibilities of using tools in the AI field for testing;

2. Conduct the final stage of testing and preparation for the release of the Cortex NFT Gallery(Thalamus);

3. Working on Cortex NFT community & system

4. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

5. Establish connections with several NFT artist

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT)

1. ZKML adjusts Trace API, fixes load/store issues, and adds force option;

2. ZKML adjusts the ShapeAdapter pass in ZKSection9, adds the function of eliminating Batch in CircomGenerator, and adapts to the circom deep learning library;

3. ZKML adjusted the circom deep learning library API and modified some calculation logic;

4. ZKML exports JSON data in CVM format, and adds operators such as conv and relu;

ZkRollup

1. Add parallel merkle tree, rescue hash, signature benchmark;

2. Account is used for test toolkit for other parts of the test (generate random accounts, account signatures and generate various transactions); Realize mint nft, withdraw nft, swap circuits, and realize the witness generation part for all types of tx;

3. The implementation of the STATE module is responsible for processing the execution of tx and priority_op, updating the L2 state and collecting the results of account update and fees.

4. Each tx type has passed the corresponding test (correct and wrong transaction parameters), and all transaction execution processes have been checked benchmark test;

Cortex Full Node

1. Use the newly typed atomics in the miner package common: fix json marshaller MixedcaseAddress (#26998)

2. Fix tx pool test and other bugs

Community Updates👨‍👩‍👦‍👦

1. Conduct the final stage of testing and preparation for the release of the Cortex NFT Gallery(Thalamus);

2. Planning Cortex NFT community & system

3. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

4. Establish connections with several NFT artist

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT)

ZKML adds new adaptation to CVM, using codegen to generate operator json data ZKML develops circomlib adaptation code ZkRollup Implement zkx (zk for cortex in plonk) type-related libraries (block, tx, op, aggregate, gas, etc.), and pass related single tests. (Compared with zion, the new types: mint_nft, withdraw_nft, swap and change pubkey can be authorized by onchain functions, and support eip-1271 sig) The test investigated halo2 related codes (fibonacci circuit), pse version ipa -> kzg (on-chain verification) Refactor the recursive circuit part, pass the basic process test and benchmark. (setup and vk create: 886s, proof for aggregate(proof num:2, depth:3): 1374s)

ZkRollup

core/txpool: allow future local transactions fix ws test timeout and other bugs consensus: improve consensus engine definition nit: int->uint64 in GetTxReceipt

Cortex investigates Ledger hard wallet signatures and develops adaptations

Community Updates👨‍👩‍👦‍👦

1. Cortex connects with NFT artists for joint creation and other matters;

2. Cortex NFT Gallery tutorial edit:

3. Cortex NFT community planning;

4. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D

1. MRT adjusts the implementation of Inference, uses TVM underlying Pass to perform sampling and output intermediate results, and especially judges some unrealized operators;

2. MRT solves the problem of Circom Shape adaptation;

3. MRT code refactoring, abstract parent class and virtual function;

ZkRollup

1. Investigate the test run of cuda bellman;

2. Investigate plonk, lookup and plonkup and test the implementation of recursive aggregation circuits;

3. Implement the basic transaction circuit (transfer, deposit, etc.) of the R1CS version;

Cortex Full Node

1. Expose chain freezer constructor without internals;

2. Fix storage engine and other bugs;

3. Fix dns test, cast address to pointer and reset gas poo;

Community Updates👨‍👩‍👦‍👦

1. Cortex NFT Gallery internal test

2. Cortex NFT Gallery Guide writing

3. Research on P2E

4. Connecting several NFT artists for collaboration

5. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D

MRT/CVM

1. MRT, TVM and Circom unified representation structure, where MRT uses built-in InferType and InferShape

2. Add MRT operator to Circom's parser, and adapt to the shape adaptation of operators such as clip

3. Increase 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

4. Adjust the type check in MRT FixPoint, reduce TVM adaptation, and increase shaping check

5. Increase the type annotation of MRT Symbol, adjust the representation structure, and enhance the readability of the code

ZkRollup

1. The WitnessGenerator module adjusts some storage-related interfaces and adds merkle_state to generate proofs

2. Refactored the execution flow of StateKeeper: PendingBlock synchronously initializes StateKeeper and WitnessGenerator

3. Implement the L1Watcher module to monitor the data on the chain

4. Mempool handles unconfirmed events

5. ZKSync - era research (supporting L2 smart contracts), newly open-sourced part of the code of the GPU prover

Cortex Full Node

1. pepple MemTableStopWritesThreshold fix

2. fix off-by-one in ipc endpoint length check

3. some funcs added and fix about pee

Community Updates👨‍👩‍👦‍👦

1. BSC Station Partnership

2. Cortex NFT Gallery internal test

3. Latest Cortex explorer test

   https://ansible.cortexlabs.ai/

4. Cortex NFT Gallery Guide

5. Research on P2E

6. Continue to monitor popular market ventures and examine project logic and Web3, NFT & Defi Tokenomic models

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D

MRT/CVM

1. TVM/MRT set CVM Python dependent library, and export Symbol format to CVM-Runtime;

2. MRT repairs the precision test logic and formats the precision output style;

3. MRT adds Symbol to Circom compilation adaptation code;

4. CVM adjusts C API interface parameters;

ZkRollup

1. Realize the transaction verification of EIP-712;

2. State separation (part1 is provided to the front-end for fast update without calculating the hash, part2 calculates witness and root hash), stores it after the pending block mini-batch, updates executed-tx, account state in db..., and avoids the lag of front-end acquisition of data;

3. Adjust the committer logic in Mempool and core. Mempool checks the nonce of commit and the continuity of tx on l1 during execution (id increments);

4. Add State init process (used to initialize state keeper and witness generator);

5. Add RwLock to Merkle Tree implementation, change the performance of most methods, and realize smt thread safety and internal variability;

Cortex Full Node

1. terminate downloader at last;

2. Report the import stats before returning the various results;

3. all: use HTTP package to replace HTTP method names (#26535)

4. miner address validation improvement;

5. Update 1.10.38 stable version;

Community Updates👨‍👩‍👦‍👦

1. Cortex NFT Gallery Beta test

2. Cortex new explorer feature test

3. Cortex NFT Gallery guide writing

4. NFT Community planning

5. Continue to monitor popular market ventures and examine project logic and Web3, NFT & GameFi Tokenomic models

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D

MRT/CVM

1. MRT updated the accuracy test code, adjusted and tested the model output results of the sampling data

2. Adjust the parameter setting of Calibrate Pass: absmax, fix the problem of FuseBatchnorm

3. Refactored the Quantiza Pass code and fixed the Infer Precision issue

4. Fixed the FixPoint Infer Type problem and added subgraph verification function

5. The model parameters are adjusted to Int8, which reduces the model size

6. CVM encapsulates the operator interface code and completes the test code

ZkRollup

1. Fixed the error that ZK for ETHF could not submit proofs to the chain, Chain ID Error

2. Zion reference implements the Rust version to speed up transaction execution

3. Implemented Poseidon hash, babyjubjub, benchmark and pubkey compression algorithms

Cortex Full Node

1. core: reset txpool on sethead

2. implement resettable freezer

3. fix tracer crash due to balance nil

4. return proper error from debug_TraceTransaction

Community Updates👨‍👩‍👦‍👦

1. Cortex Wallet (NFT Gallery) test;

2. Cortex Christmas NFT Giveaway;

3. NFT community point system planning & research;

4. Continue to monitor popular market ventures and examine project logic and Web3, NFT & GameFi Tokenomic models

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D

MRT/CVM

1. MRT implements Calibrator, FuseBatchNorm, FuseAvgPool Pass

2. MRT implements Sampling, Discretor, and Quantizer Pass, using symmetric linear discrete method to quantify the model

3. Implemented FixPoint Pass to export the model to TVM Relay Expr, all the data are shaped

4. Realized the disk import and export function of MRT internal Trace, and quickly imported the last running results

ZKRollUp

1. The ZkProver GPU code is merged and updated, and it takes an average of 6s to solve a 64-size block, which is 5 times faster

2. ZkMatrix front-end packaging optimization, asynchronous acceleration, CDN acceleration, etc.

Cortex Full Node

1. The model on the Cortex chain is archived to improve the data backup process, construct the NFT contract, and use stable diffusion to randomly generate pictures, and the NFT has been sent to the model and data address. The NFT contract address is: 0x60fA83Ea5dEC41f1D1a2724E06A40BE86aF325da

2. Cortex has built an IPFS storage platform for model data: https://cortex.infura-ipfs.io/ipfs/

   • Cortex Theseus: Pass block into collectLogs

   • Cortex Theseus: improve HTTP write timeout handling

3. Community Updates👨‍👩‍👦‍👦

   1. Cortex new browser for basic functional testing;

   2. Christmas NFT event planning;

   3. NFT community game planning;

   4. Continue to collect, translate, and original in-depth related content, NFT and DeFi and other fields, and try to develop a new community and media content operation mode;

   5. Continue to pay attention to popular projects in the market and analyze project logic and economic models.

   ---

   About Cortex 😇

   Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

   Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

   TestNet

   | Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

   Social Media

   | Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

   Telegram Groups

   | English | Korean | Chinese | Russian|Turkish

Technical Updates 🤖

Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D

MRT/CVM

1. MRT/CVM began to connect with TVM to realize automatic conversion and deployment, and improved the unified interface protocol to facilitate the subsequent ZKML design;

2. MRT implements the Symbol representation corresponding to RelayExpr and supports mutual conversion;

3. MRT implements the Calibrator at the operator level, and obtains the intermediate results of the model;

4. Built a Python Transformer framework based on Symbol representation, and the test passed FuseBatchnorm and other passes;

5. Researched and tested TVM from_mxnet, from_pytorch and other front-end APIs, and ran through TVM model/dataset import operations;

6. Correctness of CVM check operator, including CPU, FORMAL, and CUDA versions;

ZKRollUp

1. Front-end data compression to optimize loading speed;

2. ZkProver GPU adjusts the code, optimizes the data structure and tests;

Cortex Full Node

1. In the stage of model archiving and construction on the Cortex chain, the model data information is obtained from the cerebro api and tested;

2. TorrentFS solves the problem of unstable downloading and repackages the downloading function;

3. The Cortex official Remix editor is updated to version 0.8.15, link reference: https://cerebro.cortexlabs.ai/remix/#optimize=false;

Community Updates 👨‍👩‍👦‍👦

1. Zkmatrix Gitbook released

   https://cortex-blockchain.gitbook.io/zkmatrix/

2. The Cortex Labs Foundation participated in the ETHSANFRANCISCO 2022 hackathon and was shortlisted for the Top 10;

   https://twitter.com/CTXCBlockchain/status/1589571259595386880

3. "ZK Section 9 — ETHSANFRANCISCO 2022" technical document released;

4. Test the Cortex Wallet function;

5. Cortex new version browser for basic functional testing;

6. Organize Christmas NFT activities;

7. Browser multi-dimensional test research;

8. NFT community integral gameplay planning research;

9. Continue to collect, translate, and original in-depth related content, NFT and DeFi and other fields, and try to develop a new community and media content operation mode;

10. Continue to pay attention to popular projects in the market and analyze project logic and economic models.

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish

Big 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 shortages lead developers to find a generalized solution known as model quantization. After the quantization process, an AI model can reproduce execution for inference. With zero-knowledge proof (ZKP) system, AI models can Even be verified easily and quickly in constant time.

Currently, the proving time in zk-SANRKS is too slow to be applied to real-world AI applications and many projects have been working to solve this issue. vCNN[1] proposed a new efficient verifiable convolutional neural network framework to increase proving performance in 2020. The following ZEN[2] team proposed a SIMD compiler optimization that encodes multiple 8bit integers in large finite field elements without overwhelming extraction cost. In the meanwhile, zkCNN[3] created a ZKP scheme that allowed the owner of the CNN model to prove to others that the prediction of a data sample could be indeed calculated by the model, without leaking any information about the model itself. Mystique[4] presented an improved ZK protocol for matrix multiplication that yielded a 7* improvement compared to the state-of-the-art. And in 2021, zk-ml[5] published a demo that implement a zk-SNARK circuit where the proof verified that a private model could have a certain accuracy under a public dataset. In 2022, the famous zk-SNARK language, known as circom, created a library for machine learning in circomlib-ml[6] and zk-mnist[7] for hand-drawn digit recognition based on blockchain DApp.

AIGC Design

AIGC, also known as AI-Generated Content, has come into fashion recently. This service provides AI-powered and deep-learning-based text-to-image generation capabilities. Projects like DALL-E 2, Stable Diffusion, and Midjourney are all working on AIGC with slightly different focuses. The CortexLabs believes that It would be beneficial if we can bring these AI fashions into the world of Web3.

The three key stakeholders (AI creators, ML engineers, ZKP developers) don’t talk to each other at the moment. Yet, AI + ZKP + Blockchain is significant.

AI art is gaining traction, AI films have won major international awards, but AI creators lack guidance for Web3. Meanwhile, machine learning models are hard to train and generate astonishing results. Also, ML engineers don’t understand the methods and requirements to convert an algorithm to a blockchain-friendly format. Lastly, ZKP developers lack visibility into the capabilities of AIGC models and the performance/ecosystem/adoption gap.

In ETHSANFRANCISCO 2022 competition, we proposed the ZK Section 9 for on-chain artificial intelligence content-generating(AIGC) with Succinct ZK Proofs. As of this writing, we won the following prizes:

• ETHSanFrancisco Finalist

• ENS’s Integration Bounty

• Optimism’s Top 10 Deployed

With ZK Section 9, one can generate an image with an AI model, prove it with ZKP, and submit the output content on a blockchain contract. The blockchain can easily verify that the image is generated by the composer. For CortexLabs, ZK Section 9 is the main cornerstone to step into zkCVM: a complete zkRollup solution for on-chain AI models.

Here is the presentation: https://docs.google.com/presentation/d/1CIxGfM_oySgWkgdNZqh-hoeH8C8zJDxnEpHx_VK6-KQ/.

Machine learning models are huge and divergent, huge for complex models e.g. FastBERT[1]: ~1800 MFLOPS (million floating point operations), divergent for model types e.g. NLP, CNN, GAN. ZKP is hard to generate or operate: ceremony, contribution, beacon, etc. And do not forget the speed. Transforming huge and divergent AI models into a ZKP-friendly format is a challenging job. Translator and the following on-chain AI model proof brings several problems to solve, we have started to overcome the difficulties, though it is not perfect until now. More solution details will be introduced in the next section.

Focus

As mentioned at the beginning, the translation from the AI model into ZKP circuits has been challenging for developers. There are lots of key points to talk about.

Model Quantization

AI quantization has been developed and optimized over the years. Int8 quantization has become a popular approach for limited memory, computing resources, and power. The quantization trend is not only for machine learning frameworks such as TensorFlow, PyTorch, and TVM but also for hardware toolchains. For example, the USB Accelerator TPU from Coral, Inference devices with int8 precision including the Intel acceleration DL Boost instructions, NVIDIA TensorRT, Xilinx DNNDK, and many other int edge devices for machine learning.

These requirements can be considerable, and much research has been investigated. But there is no common scheme for all-in-one quantization frameworks. Many target devices and frameworks have been applied to model inference, but each executor has a different focus on quantization goals. Most quantizations are even not full-integers for accelerated purposes.

TVM concentrates on the target device deployment including quantization. The internal relay representation of TVM is a place to start for quantized models. And its front end can connect with many ML frameworks including ONNX and PyTorch.

Hence, we use PyTorch to train the AIGC model and transform it into the TVM IR representation. Then, we quantize the model and dump it into the disk.

Translator

The pre-posted ZKML papers or projects are almost model-specified schemes, which are various in zk-SNARKS implementation and can not be adapted to arbitrary AI models. So we designed a uniform model to circom translator procedure:

• parse from the TVM intermediate relay IR, which is generated from another model framework such as PyTorch, TensorFlow, etc.

• construct a temporary model representation, which can dump to or load from disk.

• generate consistent circom code for zk-SNARK.

When diving into the translator, you will find it’s not that simple. The parser should accept arbitrary IR representation, which is dumped from TVM QAT models. It’s hard to leverage between the model universality and the tolerance for error. As for the self-constructed internal symbol, we refer to the MxNet symbol graph representation, this representation is easy and scalable to apply the model transformation.

The most workload is located at the circom generator. First, parse all existing circom ML operators and store the necessary information about the operators such as dependent file path, operator arguments, and input/output. Then it’s time to map the last internal symbol to correct circom operator generators. Symbol and generator are not one-to-one mappings, we had to write valid map rules for all supported operators. And the last is to inject appropriate circom code into the template.

Circom ML Library

In this competition case, the quantization integer-only target is the ML circom operators which are like the Intel DL Boost instructions. We refer to the cvm-runtime to write ML-complete circom operators. The ZKP circom constraints are slightly different but mostly the same as the CVM operators’ design.

Besides, we implement an incomplete inference engine for the circom ML operators. It’s necessary for the circom procedure correctness verification.

ZKP Optimization

ZKP constraints are too huge to generate proof in a limited time. Previous papers have designed lots of methods like SIMD. But these optimizations were not considered due to the game time limit. Our attempts were made in the model translator to reduce the constraints in generated circom.

Contract Optimization

For the last step, there is one trouble needing to be solved. That is the contract Gas limit on the blockchain. The Ethereum block has a maximum of 8000K Gas limit to avoid Turing halting and block syncing delay problems. But it’s far from enough for a normal AIGC model to generate an image size of 32 * 32. There are several solutions: one is using the SIMD like the ZEN to embed many output elements into one field. Another is truncating model outputs into different parts and deploying contracts separately in many blocks. We tried the latter method.

Future Work

We have made some key innovations in the AIGC design, but it’s not enough for the final target. Some points are implemented and some are ignored or not perfect enough, due to the time limit. Here we list a few scalable works that can extend in the future.

ZK-Friendly Quantization

In the ZK Section 9 project, quantization is applied with the TVM and PyTorch’s internal quant aware training (QAT) method. The QAT target is designed for specific devices such as ML accelerators, this design can not be well adapted to zero-knowledge proof. We have to make a temporary symbol representation in our translator for applying many fusion passes to ZK-compatible quantization. These passes may include batch size adjustment, shape adaptor for circom operators, more compact constant fusion, etc.

Those existing ML frameworks have different backend representations and different implementations. For our united quantization goal, we hope a consolidated model intermediate representation (IR) comes true. If not, we will choose one that has the best universality, sufficient stability, and enough capacity. Then the ONNX is placed in front of us. We had chosen NNVM before because it is the oldest and stablest machine learning backend. However, most modern AI frameworks discard it and accept the ONNX-compatible scheme as the representation, for flexibility, and capacity purposes.

As for the quantization procedure, many papers have been researched on this topic. Quantization is mainly divided into quantization-aware training (QAT) and post-training quantization (PTQ) according to the occasion and cost of quantizing. The CortexLabs have made the PTQ method in MRT and gotten a decent accuracy loss for an int32 maximum quantization. In the future, transforming the PTQ from MxNet into ONNX format is necessary, also for the QAT implementations.

Inference Engine

In this project, we have made a rough inference engine in python. The engine is written in NumPy format and a few MxNet ndarray reference codes. This implementation is not sufficient for cross-validation, for the lack of code detail examination and test data. Luckily, the CortexLabs have ever published a well-tested integer-only inference engine: cvm-runtime, which is written in C++ and implemented with CPU, GPU, and formalization version. In the future, more test cases and operators will be applied to this inference engine for security and scalability.

Optimizations

ZKP-generated constraints are naturally too huge, this is a big problem for ZKP applications. The more constraints number ZKP generated, the more time and computing resources the ZKP proof procedure will cost. A simple linear layer in the AI model will lead to ~10M constraints, the subsequent snarkJS prove process almost consumes 200G memory and one-hour proving time in a 96 core 2.7 GHz CPU, 1.5T Memory machine! Moreover, a naive ResNet has 18 layers, and the relationships between layer constraints are multiplication instead of addition. So optimizations are imperative and fortunately, previous researchers have put hard work on these.

SIMD[2]

Embed multi-input elements into one big int256 field. This will make use of stranded encodings for matrix multiplication.

sign-bit grouping[2]

Use unsigned int quantization instead of an integer, this quantization is more R1CS friendly. ZKP uses elliptic curve cryptography (ECC) to generate a proof, all operation data are the 256 unsigned bit integers. Signed operators can be applied for ZKP, but the constraints will be far larger than unsigned ones.

remainder-based verification[2]

Sometimes division operators cannot be avoided in AI models, which are more complex than addition, subtraction, or multiplication. Use an extra matrix R to store the division remainder, and utilize this remainder to avoid the division.

Project Architecture

In this section, we briefly describe the project architecture, the GitHub directory structure is as follows:

├── circuits/ │ ├── Arithmetic.circom │ ├── circomlib/ │ ├── circomlib-matrix/ │ ├── operators │ └── util.circom ├── contracts/ ├── frontend/ ├── integer_only_gan/ ├── main.py ├── python/ └── README.md

Model training and quantization to TVM are located in interger_only_gan. Model translator and inference engine are located in python and the cmd is the main.py, print usage with command -h. Circom ML operators are located circuits/operators. The frontend demo is located in frontend. Blockchain contract debug and deploy code is located in contracts.

Future Applications

Our goal for this competition is AIGC, but the application future is much more than that. Like the infinite rooms of AI models, our project has a profound influence on expanding the capabilities of blockchains:

• (Shown) Vision models -> AIGC

• Language models -> chatbot, writing assistant

• Linear models and decision trees -> fraud detection, Sybil attack prevention

• Multi-modal models -> recommender systems

And some scenarios are considered for reference:

• Converting machine learning models to zero-knowledge proofs, enabling people in underdeveloped countries to generate income with on-chain AIGC or ZK-ML-enhanced trustless freelancing.

• Governance tech in consensus is tricky. Our tools can enhance existing voting and Sybil resistance techniques by allowing an ML-based approach: imagine a self-evolving DAO smart contract powered by a neural network.

• Wash-trading detectors in DEX.

• Provable biometric IDs. Like in Worldcoin.

• AI Oracles which can verify off-chain world data. Like steps, health data, environmental data, etc.

• AI competitions. People can commit their model weights’ hash first and the inputs can be revealed later.

• GameFi NPCs. We can have AI characters instead of plain old scripted NPCs.

• Dynamic tokenomics. With AI-controlled tokenomics, we can have potentially better algo-stablecoins.

• AI DAOs. AI can participate in DAO’s decision makings.

• Automated traders. Although on-chain AI will most likely fail in competition with real-world Hedge funds, it can be a showcase.

• Anti-fraud in DeFi. Like in lending protocols and insurance protocols.

• NFTs. Like what we’ve built in this project.

• Self-evolving Blockchains. Ultimately, ZKML can be used to determine crucial parameters of blockchain, such as block interval, block size, and block rewards, based on collected data.

More possibilities are in your imagination.

References

1. vCNN

2. ZEN with GitHub

3. zkCNN with GitHub

4. Mystique

5. zk-ml(2021)

6. CircomLib-ML(2022)

7. zk-mnist(2022) with Blog

8. proto-neural-zkp(2022)

9. 2PAI(2022)

Technical Updates 🤖

*Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D

ZkRollup

1. ZKMatrix adjusts the wallet plug-in logic and interfaces display;

2. ZkNode optimizes RPC link logic to increase stability;

3. 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;

4. ZKML research runs through PyTorch training, TVM IR analysis, Circom translator, and other processes.

Cortex Full Node

1. Cortex archives the model on the chain set the FTP link of the model, and cooperates with the model NFT to work;

2. Update P2P discover: improve discv5 NODES response packing;

3. Update P2P node: implement a per-source timeout in FairMix;

4. Torrent FS connections adjust;

5. Update boot nodes and reset max peers;

6. Fix native tracer, tracker and dependencies.

Community Updates 👨‍👩‍👦‍👦

1. Continue to monitor popular market ventures and examine project logic and Web3, NFT & GameFi Tokenomic models;

2. Tested the ZKMatrix wallet plug-in function;

3. NFT material arrangement;

4. Cortex explorer multi-dimensional testing & research

5. NFT community points gameplay planning & research

About Cortex 😇

Cortex’s main mission is to provide state-of-the-art machine-learning models on the blockchain in which users can infer using smart contracts on the Cortex blockchain. One of Cortex’s goals also includes implementing a machine-learning platform that allows users to post tasks on the platform, and submit AI DApps (Artificial Intelligence Decentralized Applications).

Cortex is the only public blockchain that allows the execution of nontrivial AI algorithms on the blockchain. MainNet has launched. Go build!

TestNet

| Block Explorer — Cerebro| Mining Pool | Remix Editor | Software | ZkMatrix

Social Media

| Website | GitHub | Twitter | Facebook | Reddit | Kakao | Mail | Discord

Telegram Groups

| English | Korean | Chinese | Russian|Turkish