Execution attempt

Sign in with GitHub
← Experiment E7 · Does the embedding initialisation of an APT4 transplant of Qwen2.5-1.5B damage formal domains more than prose before training, does the ranking of FOCUS, FVT and random initialisation depend on the domain, and do the differences persist under continued pretraining that updates only the embeddings?

Execution attempt · A63 · planned

Trains fastText auxiliary embeddings on the first 200 MB of the Polish training text tokenized with APT4 and freezes the 32,000-row matrix.

Succeeded · Registered by @stw2 via agent. Reported and received times are kept apart; no computation or result is verified.

Pinned source and configuration

https://github.com/stw2/tokenizer-science-tax @ e54c7e02bb1ea89dbaff6911c87331d705602b1e

Reference checked 2026-09-14 13:12 UTC. Later commits, branches or plan changes do not retarget this attempt.

Command
python scripts/01_train_aux_fasttext.py --mb 200 --dim 200 --epochs 3
Working directory
experiments/E07-embedding-init
Configuration paths
None
Parameters
mb 200; dim 200; epochs 3; workers 4 (default); skip-gram; negative 5; window 5; character n-grams 3 to 6; min_count 1; seed 20260705; trained in 272 s
Environment
3.12.13; gensim fastText on CPU, 4 worker threads; Apple silicon, 128 GB unified memory
Output directory
Not recorded

Inputs

Materials the registrant named when registering this attempt, by content identity; obtainability is derived from their location reports. Nothing is fetched or verified.

Delivered events

  1. Registered

    #1

    Trains fastText auxiliary embeddings on the first 200 MB of the Polish training text tokenized with APT4 and freezes the 32,000-row matrix.

    reported · received · @stw2 via agent · posted to the Thread

  2. Started

    #2

    Started.

    reported · received · @stw2 via agent · attempt only

  3. Succeeded

    #3

    Completed with exit code 0.

    Exit code 0.

    • aux_matrix.npy · https://github.com/stw2/tokenizer-science-tax/blob/e54c7e02bb1ea89dbaff6911c87331d705602b1e/experiments/E07-embedding-init/data/aux/aux_matrix.npy · public · sha256 2b628ab6451f… · 25600128 bytes · M352
    • aux_manifest.json · https://github.com/stw2/tokenizer-science-tax/blob/e54c7e02bb1ea89dbaff6911c87331d705602b1e/experiments/E07-embedding-init/data/aux/aux_manifest.json · public · sha256 a68b094ddf91… · 733 bytes · M353
    • aux_manifest.json as produced · Archived private repository, experiments/E7/data/aux/aux_manifest.json at commit b6867a4; the public copy makes the source path repository-relative and replaces a knowledge-graph label · unavailable · sha256 344bab5b9139… · 696 bytes · M354
    • aux_corpus.txt · Not redistributed: ask the Room owner, or rebuild with scripts/01_train_aux_fasttext.py at the pinned revisions · restricted · sha256 c5c41c30410f… · 322070593 bytes · M355

    reported · received · @stw2 via agent · posted to the Thread

Report an event

The registrant’s capture tool normally delivers start and outcome events. Reporting here is the same author report with the browser as the reported time; it does not observe the process.

This attempt has a delivered outcome. A new execution is a new attempt.