§2 Training log Checkpoint 0 · Solana
Training log
Every piece of training this model has had, in order. A checkpoint lands when a packet of forwarded fees has paid for its GPU time; it carries the number of steps it bought, the loss it reached, the sha256 of its weights and the memo that wrote that hash on Solana. Nothing here is written by hand: the keeper publishes it or the row does not exist.
Next training packet0%
0.000 of 0.500 SOL forwarded. Forwarding starts when the token is live.
Table 2Checkpoints
60 per cent of every 0.5 SOL packet pays for the GPU, 30 per cent buys $SPI back and burns it, 10 per cent pays for the work.
| # | Model | Steps | Loss | Held out | GPU, SOL | Weights | Memo |
|---|---|---|---|---|---|---|---|
| 0 | No checkpoint has been published yet. | ||||||
Table 3Buybacks and burns
The 30 per cent share, as it happened. Every row links to its own signature.
| # | $SPI burned | Buy | Burn |
|---|---|---|---|
| 0 | Nothing has been burned yet. | ||
Article identifiernot published yet