---
license: apache-2.0
tags:
  - tts
  - qwen3-tts
  - float16
  - speech-synthesis
base_model: Qwen/Qwen3-TTS-12Hz-1.7B-Base
---

# qwen3-tts-1.7b-base-fp16

FP16 conversion of [Qwen/Qwen3-TTS-12Hz-1.7B-Base](https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-Base).

## Changes from original
- All bfloat16/float32 weights cast to **float16**
- All `config.json` files updated: `torch_dtype: "float16"`
- Everything else (tokenizer, architecture) unchanged

## Why?
GTX 1650 and other Turing GPUs don't support bfloat16 natively.
Pre-converted fp16 weights load directly without runtime casting.

## Usage
```python
from qwen_tts import Qwen3TTSModel
model = Qwen3TTSModel.from_pretrained("owlninjam/qwen3-tts-1.7b-base-fp16", dtype=torch.float16)
```