Revert incorrect looping-fix banner — mlx_lm already handles multi-eos from generation_config.json

README.md

@@ -9,33 +9,18 @@ tags:
 pipeline_tag: text-generation
 ---
 
-> # 🚨 IMPORTANT — RE-DOWNLOAD `tokenizer_config.json` IF YOU GOT INFINITE LOOPS
->
-> **Fixed 2026-04-13.** Earlier versions of this repo had a tokenizer bug that caused the model to **loop forever** in stock `mlx_lm` and other loaders.
->
-> **Gemma-4 emits `<end_of_turn>` (id 106) at the end of an assistant turn**, but the original `tokenizer_config.json` only listed `<eos>` (id 1) as the stop token. Stock loaders never detected the actual end-of-turn marker → infinite loop.
->
-> ## How to fix:
->
-> ### Option A — Re-download just the tokenizer config (fastest)
-> ```bash
-> huggingface-cli download JANGQ-AI/Gemma-4-26B-A4B-it-JANG_4M tokenizer_config.json --local-dir ./your-model-dir
-> ```
->
-> ### Option B — Re-download the whole repo
-> ```bash
-> huggingface-cli download JANGQ-AI/Gemma-4-26B-A4B-it-JANG_4M --local-dir ./your-model-dir
-> ```
->
-> ### Option C — Pass the stop tokens manually
-> ```python
-> stop_token_ids = [1, 106, 50]  # <eos>, <end_of_turn>, <end_of_image>
-> ```
->
-> **The model weights are unchanged — you only need to update `tokenizer_config.json`.**
+# Gemma-4-26B-A4B-it-JANG_4M
 
----
+JANG-quantized Gemma-4 MoE for Apple Silicon. Created by Jinho Jang — eric@jangq.ai.
 
-# Gemma-4-26B-A4B-it-JANG_4M
+See the full [JANGQ-AI collection](https://huggingface.co/JANGQ-AI) for all profiles.
+
+## Loading
+
+```python
+from mlx_lm import load, generate
+model, tokenizer = load("JANGQ-AI/Gemma-4-26B-A4B-it-JANG_4M")
+print(generate(model, tokenizer, "Hello", max_tokens=256))
+```
 
-JANG quantized Gemma-4 MoE for Apple Silicon. See [JANGQ-AI](https://huggingface.co/JANGQ-AI) for the full collection.
+Stock `mlx_lm` picks up the multi-stop-token list (`[1, 106, 50]`) automatically from `generation_config.json` — no manual configuration required.