Item detail
github.com

LoRA Speedrun — a fine-tuning leaderboard where every record is re-run three times on fresh seeds before it counts

LoRA Speedrun — a fine-tuning leaderboard where every record is re-run three times on fresh seeds before it counts is a developer tool in RepoRadar's Evals & Benchmarks section, holding Gold tier and a 'try now' verdict. Its strongest signal is workflow potential, scored 9.1 out of 10.

Score8.0
Popularity100.0
Risknone
TierGold
Score breakdown
Usefulness8.0
Novelty9.0
Momentum5.0
Maturity8.8
Open-source/build8.4
Evidence7.2
Workflow potential9.1
Setup ease6.4

Popularity is tracked separately. Support, ads, sponsorships, and tips never affect these signals.

Why it matters

Useful when you fine-tune small models and cannot tell which speedup claims survive replication. The result is a verified record of which tricks — sequence packing, completion-only masking, data pruning, aggressive learning-rate schedules, fused kernels — pay for themselves in wall-clock, and anyone can re-verify a record with one command on Modal's free monthly credits.

Who should use it

engineers training small models who want evidence-backed speedups instead of blog benchmarks teams deciding between packing, pruning, QLoRA and fused-kernel tricks for short runs anyone who wants a reproducible target for teaching or comparing training stacks

Who should skip it

Pass on LoRA Speedrun — a fine-tuning leaderboard where every record is re-run three times on fresh seeds before it counts if its scope or audience does not match what your team is building right now.

About this signal

LoRA Speedrun — a fine-tuning leaderboard where every record is re-run three times on fresh seeds before it counts is tracked by RepoRadar as a developer tool in the Evals & Benchmarks section. First seen —; the source record was last checked on 2026-09-11. The current verdict is 'try now' with a Gold tier and moderate setup difficulty. The standout signals for LoRA Speedrun — a fine-tuning leaderboard where every record is re-run three times on fresh seeds before it counts are workflow potential (9.1) and novelty (9.0), while momentum (5.0) trails — that balance shapes where it fits best. This page summarizes the public evidence on the linked source page and states where additional review is still needed.

How this item is evaluated

The LoRA Speedrun — a fine-tuning leaderboard where every record is re-run three times on fresh seeds before it counts record combines a 8.0/10 composite score with separate popularity (100.0), risk (none), and setup (moderate) signals. See the scoring methodology for the current weights and evidence definitions.

Questions worth asking before you adopt this

Putting this into practice? Read How to read AI benchmarks without getting fooled for the checklist behind this score.

Risk explanation

No inherent user-impacting risk found: everything runs in your own Modal account or on your own GPU, and official verification runs strangers' submissions only inside Modal's network-blocked, secretless sandbox; Quiet since early August: last push 2026-08-04 and the newest verified record is 2026-08-03. Treat it as a complete v1 artifact rather than a fast-moving race; Official times come only from the pinned Modal L40S; local runs on your own 24 GB+ card are explicitly not comparable; Do not run other people's submissions on your own hardware — use the sandbox verification path for anything you did not write.

Evidence links
Closest alternatives / related signals
fine-tuning lora benchmark leaderboard verification reproducibility gsm8k peft
Verification record

What RepoRadar actually verified

Discovered

Automated discovery and source capture. Last checked 2026-09-13T05:04:21.529456Z.

No editorial or hands-on review is claimed. This record remains at Discovered.

Verification sources

Longitudinal intelligence

How this decision record is moving

Raw history JSON →

2 dated snapshots retained from 2026-09-12 through 2026-09-13; see the snapshot index for explicit coverage gaps. Stars, version, release, pricing, integration, risk, maintenance, verdict, score, and momentum fields remain explicit even when a source has not reported them. Repository momentum is a normalized 0–10 RepoRadar signal; GitHub stars appear only where the popularity monitor retained exact timestamped observations.

RepoRadar score8.0 current · +0.0 net
Repository momentum9.0 current · +0.0 net
GitHub stars (observed)148 current · +0 net
GitHub stars148 exact observation
Versionv1.1
Last release2026-07-20T08:51:51Z
Maintenanceactive
Current risknone
Current verdicttry now
Pricing baselineNo structured commercial pricing baseline
Pricing checkedNot applicable or not recorded
Pricing freshnessNo dated commercial pricing review
Integrations baselineNo structured integrations recorded

Recent dated points

DateScoreMomentumStarsRiskVerdictMaintenance
2026-09-138.09.0148nonetry nowactive
2026-09-128.09.0148nonetry nowactive

Why the record changed

new entity

Newly added to RepoRadar's decision catalog.