{
  "language": [
    "python",
    "rust",
    "julia"
  ],
  "license": "mit",
  "multilinguality": false,
  "size_categories": [
    "100K<n<1M"
  ],
  "task_categories": [
    "reinforcement-learning",
    "time-series-forecasting"
  ],
  "task_ids": [
    "time-series-forecasting"
  ],
  "pretty_name": "Spikenaut SNN Telemetry - Action-Proposal Trajectories + GPU/Mining/HFT Telemetry",
  "description": "v3 restructures this corpus into an action-proposal trajectory dataset for the Spikenaut Supervisor control stack (learned policy proposes, deterministic safety filter is final authority): state_telemetry, outcomes, action_proposals, safety_filter_log, encoding_params, plus clean gpu_telemetry_v3/qubic_signals derivations. All v2 configs (993,298 records of GPU, mining, and HFT paper-trading telemetry, captured March 2026 on an NVIDIA RTX 5080) remain published unchanged. See the Changelog in README.md.",
  "total_records": 993298,
  "record_breakdown": {
    "full_data/neuromorphic_data.jsonl": 813973,
    "full_data/node_sync_harvest.jsonl": 120322,
    "full_data/ghost_market_log.jsonl": 31573,
    "full_data/qubic_ticks_snn.jsonl": 27430
  },
  "serving_note": "All nine configs load from Parquet. The four v2 configs are served from fidelity-verified conversions under v2_parquet/ because a script-less Hub dataset applies ONE builder (inferred from the first config's files) to every config, so JSONL and Parquet configs cannot share a card. full_data/*.jsonl remain the canonical cleaned exports, byte-untouched. Only representation change: qubic_ticks.timestamp is timestamp[ms] (was timestamp[s]); values identical.",
  "v3": {
    "schema_version": "3.0.0",
    "teacher_policy_version": "1.0.0",
    "built_by": "spikenaut-etl build-v3 (rmems/spikenaut-telemetry-etl PR #7, merge commit 983323bdd6679832747009c6f4b8d17c0d2ea961)",
    "note": "v3 rows are derived from full_data/neuromorphic_data.jsonl and are not additional records; action_proposals and safety_filter_log are typed empty schema artifacts (zero teacher labels over the v2 backfill by design).",
    "episode_len": 4096,
    "outcome_horizon_steps": 64,
    "n_episodes": 199,
    "embargo_episodes": ["gpu-000139", "gpu-000169"],
    "config_rows": {
      "gpu_telemetry_v3": {"full": 813973},
      "qubic_signals": {"full": 813973},
      "state_telemetry": {"train": 569344, "validation": 118784, "test": 117653},
      "outcomes": {"train": 569344, "validation": 118784, "test": 117653},
      "action_proposals": {"train": 0, "validation": 0, "test": 0},
      "safety_filter_log": {"train": 0, "validation": 0, "test": 0},
      "encoding_params": {"train": 36}
    }
  },
  "generated_by": "https://github.com/rmems/spikenaut-telemetry-etl",
  "collected_by": "https://github.com/rmems/Theseus-Quarry",
  "tags": [
    "blockchain",
    "neural-networks",
    "spiking-neural-networks",
    "snn",
    "kaspa",
    "monero",
    "qubic",
    "dynex",
    "quai",
    "verus",
    "telemetry",
    "gpu",
    "mining",
    "hft",
    "time-series",
    "neuromorphic",
    "fpga",
    "reinforcement-learning",
    "rlds",
    "action-proposal",
    "safety-filter",
    "vllm",
    "nvml"
  ]
}