Add Qwen text encoder path override

[bghira]

Summary

Adds qwen_text_encoder_model_name_or_path as a first-class option for Qwen text encoder components.

• exposes the option in the WebUI/CLI field registry and env-file mapping
• routes shared and custom Qwen text encoder loaders through the override when a model has one Qwen text encoder
• logs a warning and ignores the override for model families that define multiple Qwen text encoders
• updates OPTIONS and FLUX.2 quickstart docs, including existing translations

Why

Qwen-based models currently load their Qwen text encoder from the model package or an upstream Qwen repository. Users with a local or alternate Qwen-compatible text encoder need a supported config path instead of replacing files inside the Hugging Face cache.

Validation

• python -m unittest tests.test_qwen_text_encoder_override tests.test_hunyuanvideo_model tests.test_ace_step_model tests.test_acestep_lora_targets tests.test_krea2_model tests.test_anima_model tests.test_boogu_image_model tests.test_kandinsky5_models tests.test_model_field_registry tests.test_parser_type_override -v
• python -m compileall -q on touched Python files
• git diff --cached --check
• staged file scan for local absolute user paths before commit returned no matches

Note: the separate feature worktree does not contain its own .venv, so local hooks that hardcode .venv/bin/python could not complete there. The validation above was rerun successfully using the project venv from the main checkout before committing.

[Copilot]

Pull request overview

Adds a first-class configuration override (qwen_text_encoder_model_name_or_path) so users can point Qwen-based model families at an alternate/local Qwen-compatible text encoder without modifying the Hugging Face cache, and wires this through relevant model loaders and docs.

Changes:

• Introduces qwen_text_encoder_model_name_or_path across the config surface area (WebUI field registry + env mapping) and routes Qwen text encoder resolution through it.
• Updates multiple Qwen-using model families to honor the override (and to ignore + warn when multiple Qwen encoders exist).
• Adds targeted unit tests and updates OPTIONS + FLUX.2 quickstart documentation (including existing translations).

Reviewed changes

Copilot reviewed 29 out of 29 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
tests/test_qwen_text_encoder_override.py	New unit coverage for override path/subfolder resolution, warnings, CLI parsing, and env mapping.
tests/test_hunyuanvideo_model.py	Verifies HunyuanVideo prefers the new Qwen override over the legacy path.
tests/test_acestep_lora_targets.py	Extends test config with the new override field.
tests/test_ace_step_model.py	Adds coverage ensuring Ace Step v15 layout uses the override.
simpletuner/simpletuner_sdk/server/services/field_registry/sections/model.py	Registers the new WebUI/CLI field for the Qwen text encoder override.
simpletuner/helpers/models/common.py	Centralizes Qwen override detection, warnings, subfolder/path resolution helpers, and integrates them into tokenizer/text encoder loading.
simpletuner/helpers/models/hunyuanvideo/model.py	Prefers qwen_text_encoder_model_name_or_path when selecting the Qwen text encoder source.
simpletuner/helpers/models/flux2/model.py	Allows Klein Qwen3 text encoder loader to use the new override as the preferred explicit path.
simpletuner/helpers/models/ace_step/model.py	Allows v15 layout to resolve even if shared tokenizer dir is absent when an override is provided; uses override as tokenizer path.
simpletuner/helpers/models/anima/model.py	Routes Anima Qwen tokenizer/text encoder selection through the override (including single-file support path).
simpletuner/helpers/models/ideogram/model.py	Uses override as Qwen source and clears subfolders when override is set.
simpletuner/helpers/models/longcat_image/model.py	Loads text processor from override root when provided; otherwise uses model subfolder layout.
simpletuner/helpers/models/qwen_image/model.py	Routes processor loading through shared Qwen processor path/subfolder resolvers.
simpletuner/helpers/models/mageflow/model.py	Routes processor loading through shared Qwen processor path/subfolder resolvers.
simpletuner/helpers/models/krea2/model.py	Routes processor loading through shared Qwen processor path/subfolder resolvers.
simpletuner/helpers/models/boogu_image/model.py	Routes processor loading through shared Qwen processor path/subfolder resolvers.
simpletuner/helpers/configuration/env_file.py	Adds env var → CLI arg mapping for the new override.
documentation/OPTIONS.md	Documents the new CLI option and updates autogenerated usage/options excerpts.
documentation/OPTIONS.zh.md	Translation update for the new option.
documentation/OPTIONS.ja.md	Translation update for the new option.
documentation/OPTIONS.pt-BR.md	Translation update for the new option.
documentation/OPTIONS.es.md	Translation update for the new option.
documentation/OPTIONS.hi.md	Translation update for the new option.
documentation/quickstart/FLUX2.md	Updates Klein guidance to reference the new Qwen override option.
documentation/quickstart/FLUX2.zh.md	Translation update for Klein guidance.
documentation/quickstart/FLUX2.ja.md	Translation update for Klein guidance.
documentation/quickstart/FLUX2.pt-BR.md	Translation update for Klein guidance.
documentation/quickstart/FLUX2.es.md	Translation update for Klein guidance.
documentation/quickstart/FLUX2.hi.md	Translation update for Klein guidance.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.