#OPML 1. Solved issues with Unicorn executing cvm: - Recompiled MIPS dynamic libraries and ELF executable. - Mapped file system to specific memory addresses.
2. Developed program test steps, rewrote compilation script, parsed C++ results from ELF file. 3. Deployed cvm-runtime to MLVM Unicorn for execution, obtaining checkpoint_final and investigating specific instructions.
#ZkRollup 1. Implemented input across all circuits. 2. Used grand products for queue equality proof. 3. Sorted & deduplicated code decommit requests; added code decompressor for memory unpacking. 4. Processed log queue into specific queues, sorting L1Message logs and storage requests. 5. Built circuit for main VM.
#CortexFullNode 1. Optimized prefetcher heap escape https://t.co/BeXBq0Iv5x 2. Reduced unnecessary JS bigint compilations https://t.co/LTOElTS3lj 3. Simplified MaxUint64 https://t.co/55MiFR2kOw 4. Fixed missing state when num is zero https://t.co/XW1MKm8XJm 5. Restructured dependencies on header
#OPML 1️⃣ Completed MIPS instruction set compilation for cvm-runtime & implemented the test model. 2️⃣ Fixed MIPS dynamic library loading & adapted to tensor/data types of MLVM. 3️⃣ Modified memory layout & adjusted model reading format to struct + weights; implemented data reading/writing logic for specific memory addresses. 4️⃣ Adapted cvm-runtime model inference to support step functions. 5️⃣ Added the Unicorn debug tool.
-ZKML v1.0.1 released, MNIST model test passed 🧠 -Fixed quantization bugs in MobileNet, SqueezeNet 🔧 -ZKCVM progress with CPU zk-rollup & GPU proof generation ✅ -Transition to OPML tech for better performance 🔄