OCR pilot test

OCR Done

ocr_pilot100k_test

, trained on L4 24GB (AWS, spot). Started 6 Oct 2026, ended 6 Oct 2026.

Hypothesis

Two books plus benchmark pages, to check the bulk OCR pipeline end to end before the full run.

What we learned

Pipeline check on a spot GPU.

Scores

No scores published yet.

Curves

No training curves were published for this run.Its scores above are what it published.

Model

Preset
—
Parameters
—
Outside embeddings
—
Layers · heads · width
— · — · —
Tokenizer
—

OCR-VLM v1 (vLLM)

Data

Slice
—
Words
—
Training tokens
—
Passes
—
Tokens seen
—

Compute

GPU
L4 24GB
Where
AWS (spot)
Steps
—
GPU hours
0.57
Cost
$0.52
Spot restarts
0

Lineage

awsspot