README: remove references to files not in the GitHub repo (root requirements.txt, smoke_train.py)

README.md

@@ -150,10 +150,10 @@ We provide a standalone inference script [`infer_xaut.py`](https://github.com/X-
 python -m venv .venv
 source .venv/bin/activate
 # Install the PyTorch/torchaudio build for your accelerator first, then:
-pip install -r requirements.txt
+pip install transformers qwen-asr huggingface_hub
 ```
 
-Validated dependency versions are pinned in [`requirements.txt`](https://github.com/X-AuT/X-AuT/blob/main/requirements.txt). The repository does not pin PyTorch because CUDA, PPU, and CPU environments require different vendor builds.
+The repository does not pin PyTorch because CUDA, PPU, and CPU environments require different vendor builds.
 
 ### Quickstart
 
@@ -220,13 +220,7 @@ cd xaut-ft-simple
 bash run_ft.sh
 ```
 
-Before preparing a dataset, return to the repository root and validate model loading, audio preprocessing, forward/backward, and one optimizer step with the generated license-free sample:
-
-```bash
-python xaut-ft-simple/scripts/smoke_train.py --device-type cuda
-```
-
-The recorded PPU code-path check and its explicit full-checkpoint boundary are documented in [`xaut-ft-simple/VALIDATION.md`](https://github.com/X-AuT/X-AuT/blob/main/xaut-ft-simple/VALIDATION.md).
+For the recorded PPU code-path check and its explicit full-checkpoint boundary, see [`xaut-ft-simple/VALIDATION.md`](https://github.com/X-AuT/X-AuT/blob/main/xaut-ft-simple/VALIDATION.md).
 
 ## 📁 Checkpoint
 

README_zh.md

@@ -128,10 +128,10 @@
 python -m venv .venv
 source .venv/bin/activate
 # 先安装与本机加速卡匹配的 PyTorch/torchaudio，再执行：
-pip install -r requirements.txt
+pip install transformers qwen-asr huggingface_hub
 ```
 
-已验证的上层依赖版本记录在 [`requirements.txt`](https://github.com/X-AuT/X-AuT/blob/main/requirements.txt) 中。仓库不固定 PyTorch 版本，因为 CUDA、PPU 与 CPU 环境需要不同的厂商构建。
+仓库不固定依赖与 PyTorch 版本，因为 CUDA、PPU 与 CPU 环境需要不同的厂商构建。
 
 ### 快速开始
 
@@ -198,12 +198,6 @@ cd xaut-ft-simple
 bash run_ft.sh
 ```
 
-准备真实数据前，请回到仓库根目录，并用自动生成的无版权合成音频验证模型加载、音频预处理、前向/反向传播和一次优化器更新：
-
-```bash
-python xaut-ft-simple/scripts/smoke_train.py --device-type cuda
-```
-
 已执行的 PPU 训练路径验证及“未验证完整权重”的边界说明见 [`xaut-ft-simple/VALIDATION.md`](https://github.com/X-AuT/X-AuT/blob/main/xaut-ft-simple/VALIDATION.md)。
 
 ## 📁 检查点