README.md

metadata

license: other
license_name: per-record
language:
  - pl
pretty_name: Polish technical mega
task_categories:
  - text-generation
library_name: datasets
size_categories:
  - 10M<n<100M
tags:
  - polish
  - technical
  - cnc
  - industrial-maintenance
  - automation
  - plc
  - robotics
  - dynaword
  - parquet
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train.parquet
      - split: validation
        path: data/validation.parquet
      - split: test
        path: data/test.parquet

Polish technical mega

Large deduplicated Polish technical corpus covering CNC, machine tools, industrial maintenance, automation, PLC, robotics, mechatronics, engineering, energy systems and open technical textbooks.

This is a document/chunk corpus, not a question-answer set. Licences are retained per record; provenance and source hashes are included.

It contains 22331 unique documents and 40468328 proxy tokens.

Split	Documents	Tokens
train	20095	36294819
valid	1135	2157694
test	1101	2015815