feat: add masked language modeling ingestor template

[shujaatTracebloc]

Summary

• Add MASKED_LANGUAGE_MODELING to TaskCategory constants and get_all_categories()
• Add MLM validator mapping in validators_mapping.py — validates .txt file extensions in sequences/ directory, plus table name and duplicate checks
• Create templates/masked_language_modeling/ with full ingestion script, README, and 5 sample PrimeKG-style random walk sequences
• Add examples/yaml/masked_language_modeling.yaml for CLI-based ingestion

Key difference from text_classification

MLM is self-supervised — no label_column is needed. The CSV manifest only contains filename and extension columns. Masking is applied on-the-fly by the training client during training.

Data format

data/
├── labels_file.csv           filename, extension
├── sequences/
│     seq_0000001.txt         "Lepirudin indication Huntington phenotype_present Chorea"
│     seq_0000002.txt         "Soliris target TP53 associated_with Breast_cancer"
└── tokenizer.json            (placed on dataset path, not ingested)

Files changed

File	Change
tracebloc_ingestor/utils/constants.py	Add MASKED_LANGUAGE_MODELING to TaskCategory
tracebloc_ingestor/utils/validators_mapping.py	Add MLM validator chain
templates/masked_language_modeling/	New template with script, README, sample data
examples/yaml/masked_language_modeling.yaml	New YAML config example

Test plan

• Verify TaskCategory.is_valid_category("masked_language_modeling") returns True
• Verify map_validators(TaskCategory.MASKED_LANGUAGE_MODELING, ...) returns correct validator chain
• Run masked_language_modeling.py template against sample data
• Verify ingested records have correct filename, extension, and no label column errors

🤖 Generated with Claude Code

---

Note

Medium Risk
Adds a new ingestion category that changes CLI schema validation, convention resolution, and validator selection, which could affect YAML-driven runs if misconfigured. Main risk is around sidecar path resolution (sequences/SRC_PATH) and the new TokenizerValidator failing datasets missing tokenizer.json or required tokens.

Overview
Adds first-class masked language modeling (MLM) ingestion support across the declarative YAML/CLI path.

The PR extends TaskCategory, the ingest JSON schema, and CLI conventions to recognize category: masked_language_modeling, require a sequences sidecar directory, derive SRC_PATH from it, and default MLM to DataFormat.TEXT with .txt file options.

It wires in an MLM-specific validator chain (including a new TokenizerValidator that enforces presence of tokenizer.json with [MASK]/[PAD]) and adds a complete templates/masked_language_modeling/ example plus an examples/yaml/masked_language_modeling.yaml config.

^{Reviewed by Cursor Bugbot for commit 5e774ec. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

👋 Heads-up — Code review queue is at 18 / 8

Above the WIP limit. The team convention is to review existing PRs before opening new work.

Open PRs currently in Code review (oldest first):

• .github#41 — fix(profile): point install snippets to canonical i.sh / i.ps1 URLs · author: @saadqbal · reviewer: @LukasWodka
• client#100 — feat(client): scaffolding for stateless requests-proxy auth (HC-1) · author: @saadqbal · no reviewer assigned
• client-runtime#35 — feat(removing package for dev for merge #21): jobs-manager submit-ingestion-run HTTP endpoint · author: @saadqbal · no reviewer assigned
• client-runtime#36 — Bump requests from 2.32.3 to 2.33.0 in /Node-deploy · author: @dependabot · no reviewer assigned
• client-runtime#37 — Bump black from 25.1.0 to 26.3.1 in /Node-deploy · author: @dependabot · no reviewer assigned
• docs#3 — Block bots from crawling Mintlify static assets · author: @LukasWodka · reviewer: @saadqbal
• frontend-app#465 — chore(deps): bump next from 15.5.9 to 15.5.18 · author: @dependabot · no reviewer assigned
• frontend-app#466 — chore(deps): bump systeminformation from 5.28.5 to 5.31.6 · author: @dependabot · no reviewer assigned
• frontend-app#467 — chore(deps): bump axios from 0.21.4 to 0.31.1 · author: @dependabot · no reviewer assigned
• start-training#16 — chore: merge main into develop to resolve PR Csv ingestor #15 conflicts · author: @saadqbal · reviewer: @LukasWodka

Pull from review before opening new work. (This is a nudge from the kanban WIP check, not a block.)

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit fd3578e. Configure here.}