docs: works on modern transformers now; soften version note

README.md

@@ -37,19 +37,19 @@ A LoRA fine-tune of [**MOSS-TTS-Nano**](https://huggingface.co/OpenMOSS-Team/MOS
 
 The voice became the target Indian-English speaker **without** a reference clip, with intelligibility unchanged.
 
-## ⚠️ Requirements — version pin matters
-This model uses MOSS-TTS-Nano's **custom modeling code (`trust_remote_code`)**, which targets a
-specific Transformers version. On **newer Transformers** you will see errors such as
-`TypeError: unsupported operand type(s) for |: 'list' and 'set'` and/or `MISSING ..._lm_head.weight`
-(the LM heads are *tied* weights restored on load — a version mismatch breaks the tying). Pin:
+## Requirements
+This repo's custom modeling code includes a **cross-version compatibility fix**, so it loads on
+both `transformers==4.57.1` and **modern Transformers (tested 5.12.1)** — the older
+`TypeError: unsupported operand type(s) for |: 'list' and 'set'` is resolved. Install:
 
 ```bash
-# GPU (Blackwell/most NVIDIA): CUDA 12.8 wheels
-pip install torch==2.7.0 torchaudio==2.7.0 --index-url https://download.pytorch.org/whl/cu128
-# REQUIRED pinned versions (same as MOSS-TTS-Nano upstream)
-pip install "transformers==4.57.1" soundfile sentencepiece numpy huggingface_hub
-# CPU-only: pip install torch==2.7.0 torchaudio==2.7.0   (default index)
+pip install transformers torch torchaudio soundfile sentencepiece numpy huggingface_hub
+# GPU (Blackwell/most NVIDIA), if needed:
+#   pip install torch==2.7.0 torchaudio==2.7.0 --index-url https://download.pytorch.org/whl/cu128
 ```
+`torchaudio` is required (the modeling code imports it). The `MISSING ..._lm_head.weight` line in
+the load log is **cosmetic** — those heads are *tied* weights, rebound to the embeddings on load.
+For exact parity with the training environment you may still pin `transformers==4.57.1`.
 
 ## Usage
 ```python