Spaces:
Running on Zero
Running on Zero
Commit History
22 days ago
-
Refine game onboarding for an ML-familiar audience
Verified signatureHF visitors can know basic ML while encountering AI Learns to Drive for the first time. The guide previously reached result delivery without enough context for interpreting a driver or planning the next experiment. This revision introduces the driving observations and controls, evolutionary search, and the headless lab. It then walks through the existing A01/Formula/CPU baseline, explains fitness versus driving measurements, preserves a checkpoint, and presents evaluation and continuation choices before recipes and batches. Concise links point to the game wiki while retaining this lab's actual defaults and supported behavior. The persistent Help links now open at the game orientation. Quick start remains compact and closed by default. Existing recipe, batch, freezing, storage, and challenge-reference anchors remain available. Validation: 8 integrated UI workflow checks passed; 13 unique help anchors and all 18 internal/app-to-help links are valid. First-run values and result labels …be0e625 -
Improve quick-start help and recipe/batch editing (#1)
- Improve quick-start help and recipe/batch editing (62f13e52057f284d4896c2455f73b41c97cea118) - Guard batch launches with the visible saved plan and current run setup (cedfddfd679847e29e34051ebb9cebbf62dae434) - Explain the compact tutorial, named recipes and ordered batch workflow (33ab696f7c03dc42ba954fafdd15e4e978f33406) - Update release integrity entries for the reviewed frontend files (d0fc87b078774efe0384ca189b9d9cd9ce8abae9)bae512d -
Default training and frozen evaluation to 60 Hz decisions (batch count 1)
Verified signaturee70da5c
26 days ago
-
Raise measured training limits
Verified signature2e4c103 -
Allow bounded multi-step training batches
Verified signaturee291629 -
Fix post-run settings editing and run-size feedback
Verified signature76acecb -
Add selective input, layer and output freezing
Verified signature46ae7f5 -
Improve onboarding, useful defaults and continuation guidance
Verified signature2a138ca -
Add current-policy star checks without campaign progression
Verified signaturebe1eee5 -
Add distinct winners, frozen evaluation and private-bucket recipes
Verified signature6c60ea0 -
Add optional HF bucket consent, automatic unique saves, history and restore picker
Verified signature0e86155 -
Add reviewed opt-in private bucket save and restore
Verified signature3315dab
27 days ago
-
Add matched four-vehicle support to ALD training
Verified signaturePreserve vehicle Type 0–3 through network import, training and export. Select the packaged trusted vehicle scene, enforce original control-output rules and report vehicle identity for optional global Steam estimates. Keep the existing CPU-private and GPU-protected Space visibility settings and package member boundary unchanged.5a037eb -
Fix original-game export loading and add opt-in Steam comparison
Verified signatureKeep existing Space access settings and simulator binaries. Fix required Id, Stats and Snapshots in exported game networks. Add optional read-only Steam comparison, run-wide best lap and bounded request admission. Formula remains the only hosted vehicle.a6e1920 -
Complete standard game metadata in five-output network exports
Verified signatureKeep existing Space access settings and simulator binaries. Expand hosted adapter and editor to all five Formula controls, 20 inputs and ten hidden layers, with original sensor defaults and backward compatibility.928b9a7 -
Support all five Formula controls and maximum game networks
Verified signatureKeep existing Space access settings and simulator binaries. Expand hosted adapter and editor to all five Formula controls, 20 inputs and ten hidden layers, with original sensor defaults and backward compatibility.ce0b812 -
Prepare public beta: in-page confirmations, feedback and safe GPU errors
Verified signatureKeep existing Space access settings. Add recipe-import recovery, capability gates, GitHub issue drafts and bounded CUDA reference instructions; preserve simulator binaries.63c56ce -
Beta r2: traceable run receipts and focused UI fixes
Verified signatureed1475b -
Audit single-engine UI wiring and release GPU tickets on quota failure
Verified signatureCorrect browser network status after batches, reward preset labels, map sizing and quota guidance. Add immediate discard for quota-rejected pending inputs; retain 120-second expiry for ambiguous failures.24cc077
28 days ago
-
Refresh single-engine training UI, track map, and short-lived results
Verified signatureUpdate to upstream b90b8af6 CI artifacts; keep CPU private and GPU Protected. Replace stale dual-engine files with the bounded runner, browser recipes and built-in track map. Results expire after five minutes; server retention is opt-in per single run.623f6f8
-
Simplify optional GPU account controls and explain connection recovery
Verified signaturee63a295 -
Add 28 accurately labeled built-in replica tracks and synchronized selectors
Verified signature76e0bfa -
Add optional HF OAuth visitor identity for Protected GPU requests
Verified signaturee49e31a -
Fix Protected standalone GPU entry without adding login
Verified signature9b81009 -
Deploy r10 RC2 runner interface beta
Verified signatureClean 26-file package; configurable training, browser-local recipes and network editor, bounded batches, existing ephemeral delivery policy. Hosted verification follows deployment. No hardware or visibility change.a6d3251 -
Restore private TiGa file workflow with HF wrapper guidance
Verified signature8751b29
-
Self-contained startup and first-party private HF recovery
Verified signature1d0ef9e -
Private CPU interface integration with ephemeral result lifecycle
Verified signature2031f77 -
Pin simulator d8ab74c; explicit math profile and fail-closed CUDA worker
Verified signature95a1fbc
-
Reserve private GPU comparison deployment; explicitly blocked pending compatible backend
Verified signaturea3b395b