Datasets:
Expand committed telemetry dataset: mining, HFT, and sync node data
#3
by rmems - opened
Current State
The committed dataset is small:
fresh_sync_data.jsonl: 8 records (Kaspa + Monero telemetry)dataset/research/qubic_ticks.jsonl: 27,430 records (Qubic epoch 205)hybrid_training_results.json: Training metrics
The local data/ directory has more files but they are untracked.
What Needs Adding
Mining Telemetry
- Dynex, Quai, Qubic, Kaspa, Monero, Ocean, Verus hashrate/power/temp data
- Block acceptance events, epoch transitions
- GPU NVML snapshots (temperature, power, VRAM, clock)
HFT Market Data
- Order flow, price ticks, latency, spread
- Trading logs from Ghost Market simulation (31,573 records referenced in old README)
- Multi-asset: DNX, RENDER, BTC, NEAR, PEPE, ASI, SOL
Sync Node Data
- Block height, sync progress, epoch transitions
- Node health metrics
Hardware Telemetry
- GPU thermal/power readings during training
- FPGA supervisor readings during online learning
Goal
Have enough committed data for reproducible training runs. The dataset should be large enough that a clean training run produces diverse, non-monotonic weights.
Mimo Code agent: MiMo-V2.5-Pro