receipt_id string | publisher string | date_utc timestamp[s] | evidence_class string | hardware dict | workload dict | stages list | throughput_comparison dict | projected_same_shape_wall_time_ratios dict | interpretation string | limitations list | provenance dict |
|---|---|---|---|---|---|---|---|---|---|---|---|
THC-RTX4050-OMNI-20260913 | The Hidden Canopy | 2026-09-13T00:00:00 | curated public summary of published project benchmark JSON | {
"manufacturer": "NVIDIA",
"device": "GeForce RTX 4050 Laptop GPU",
"architecture": "Ada",
"vram_gib_reported": 6.14
} | {
"description": "Omni reasoner native training, same 300-packet GA=1 workload shape",
"packets": 300,
"gradient_accumulation": 1
} | [
{
"name": "initial",
"reported_or_estimated_wall_hours": 33,
"retained_wall_hours": null,
"estimated_seconds_per_packet": null,
"estimated_training_seconds": null,
"estimated_checkpoint_seconds": null,
"estimated_total_wall_seconds_rounded": null
},
{
"name": "after three kernel ... | {
"initial_tokens_per_second_approximate": 82,
"optimized_tokens_per_second_approximate": 1900,
"throughput_ratio_approximate": 23.17
} | {
"v2_to_optimized": 22.17,
"first_attempt_to_optimized": 172.17
} | The approximately 22x and 170x end-to-end figures are projections from the optimized per-packet measurement and checkpoint estimates, NOT completed stopwatch timings for the full optimized 300-packet run. | [
"The projected current 11.5-minute wall time has not been represented as an observed full-run duration.",
"The 33-hour initial estimate and 4.25-hour retained v2 stage must not be described as a head-to-head independent benchmark across different models.",
"The speedups refer to this particular training shape o... | {
"public_source": "https://github.com/The-Hidden-Canopy/OMNI-Q_Weird_Stuff_Machine/blob/main/evidence/benchmark_results/omni_reasoner_speed_2026-09-13/results.json",
"supporting_internal_report": "The-Hidden-Canopy/IDA-TRAIN-V2 (private) docs/rtx4050-omni-native-training-report-2026-09-10.md"
} |
Three GPU optimization receipts
Three curated, hardware-specific research benchmark summaries from The Hidden Canopy. Published for reproducible discussion of measured or explicitly projected performance and its limits. The JSON documents are the three receipts; this README is only the dataset card.
| Receipt | Device | Headline | Status |
|---|---|---|---|
| AMD OpenCL hybrid | Radeon 610M / gfx1036 | 13.12 → 107.31 tokens/s; about 8.18× on Swift-shaped canary | Internally reported completed test, CPU/GPU hybrid |
| Omni native training | RTX 4050 Laptop | ~82 → ~1,900 tokens/s; ~170× projected cumulative wall-time improvement | Published source JSON; optimized full-run time is estimated |
| 1F1B pipeline | 2× RTX 5090 | 42,213 → 53,968 tokens/s; +27.9% on matched small ablation | Internally reported 20-microstep experiment |
Read before comparing
These are three different workload shapes, not three entries in a common model-throughput ranking. No claim is made that they outperform vendor-native NVFP4, AMD Instinct, or any unrelated production benchmark.
The RTX 4050 ~170× number is a projection, not a completed full-run stopwatch result. The AMD receipt includes work performed on CPU as well as GPU. The dual-5090 result is a short, matched pipeline ablation rather than a completed convergence test.
No private host addresses, user directory paths, infrastructure credentials, proprietary weights, or unrelated source files are part of this release. Original private-repository reports remain private. No independent third-party audit is implied.
The receipt IDs and source references are provided so the owner can produce the underlying run material to a qualified verifier.
- Downloads last month
- 45