Downloads · 30 days
0
LamTNguyen/LESA-FLUX-cache-research-artifacts
LESA-FLUX-cache-research-artifacts is a machine learning model from LamTNguyen. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for pytorch. The card lists the license as other.
Public, reproducible artifacts for the A100 zero-warmup FLUX.1-dev campaign and the data-matched LearniBridge-100 reproduction. This repository contains trained checkpoints, the 100-by-50 teacher feature cache, exact…
Downloads · 30 days
0
Access
Public
Updated Sep 20, 2026
Repo size
150 GB
Likes
0
Public
Click a slice to open those files.
.uint16144 GB · 96%
From the Hugging Face model README
Public, reproducible artifacts for the A100 zero-warmup FLUX.1-dev campaign and the data-matched LearniBridge-100 reproduction. This repository contains trained checkpoints, the 100-by-50 teacher feature cache, exact prompt/seed and schedule manifests, logs, DB50/DB200 images and metrics, HPSv2.1 results, qualitative sheets, and invalidated runs retained under explicitly named provenance folders.
The original FLUX.1-dev weights are not redistributed. Obtain them from
black-forest-labs/FLUX.1-dev under their license. Exact revision and SHA256
are recorded in ARTIFACT_MANIFEST.json.
Primary results:
campaigns/a100_zero_warmup/results/metrics/summary.jsoncampaigns/learnibridge_100_matched/results/summary.jsoncampaigns/learnibridge_100_matched/results/summary.csvTeacher cache:
campaigns/a100_zero_warmup/results/teacher_cache/z_pre.uint16campaigns/a100_zero_warmup/results/teacher_cache/velocity.uint16campaigns/a100_zero_warmup/results/teacher_cache/manifest.jsonAll valid comparison rows use zero-warmup inference: full backbone at
0,N,2N,..., with cache reuse beginning at solver step 1.