task stringclasses 1
value | confirmed_target dict | current_poc_run dict | measured_throughput dict | aws_status dict | roles dict | note_on_access stringclasses 1
value |
|---|---|---|---|---|---|---|
vilhyra image manipulation dataset - proof-of-concept build | {
"images": 300000,
"videos": 100000,
"note": "combined totals across all categories - confirmed by founder, not to be second-guessed by any AI tool"
} | {
"target_base_images_this_run": 200,
"downloaded_so_far": 40,
"manipulated_outputs_so_far": 0,
"wikimedia_rate_limit_hits_this_run": 4,
"process_alive": true
} | {
"source": "Wikimedia Commons, single machine, single IP",
"observed_rate": "~60 images/hour (dominated by 600s rate-limit cooldowns)",
"implication": "reaching 300,000 images from this single source/IP alone would take ~5,000 hours (~208 days) - scaling (parallel IPs/regions, additional data sources) is require... | {
"instances_running": 0,
"cost_incurred": "$0.00",
"gpu_quota_us_east_1": "CASE_OPENED (pending)",
"gpu_quota_us_east_2": "PENDING (submitted as cheaper backup region)",
"note": "AWS reserved for the training step only, per explicit founder instruction - not used for dataset building at this time"
} | {
"builder": "Google AI Studio",
"supervisor_enforcer_prompt_engineer": "Claude",
"note": "Claude verifies/fixes real bugs and reports real measurements - does not change agreed specs/targets"
} | Google Studio's session has Search grounding only, no direct URL fetch tool - this file is relayed manually by Claude for now, but stays live at this URL for any future session/tool that can fetch it directly. |
No dataset card yet
- Downloads last month
- 3