ocrvlm_pilot_q35
on OCR page labels, trained on A100 40GB (Google Cloud, spot). Started 4 Oct 2026, ended 4 Oct 2026.
Hypothesis
Short pilot of fine-tuning a 2B open vision-language model (Qwen3.5-2B) on our page-OCR labels, to set the learning rate.
What we learned
300 steps, validation only; its settings carried into OCR-VLM v1.
Scores
No scores published yet.
Curves
Training loss
Cross-entropy per token, by step. The first few percent of the run, far higher, run off the top; hover or the table has every value.
Held-out bits per byte
The validation split, evaluated during training, by step. Lower is better.
Throughput
Tokens per second, by step.
Model
- Preset
- —
- Parameters
- —
- Outside embeddings
- —
- Layers · heads · width
- — · — · —
- Tokenizer
- Qwen3.5 tokenizer
Data
- Slice
- OCR page labels
- Words
- —
- Training tokens
- —
- Passes
- —
- Tokens seen
- 11,877,635
Compute
- GPU
- A100 40GB
- Where
- Google Cloud (spot)
- Steps
- 300 / 300
- GPU hours
- 1.37
- Cost
- $1.80
- Spot restarts
- 0
Lineage
gcpa100spot