Upload README.md with huggingface_hub

README.md

@@ -37,33 +37,31 @@ FormalASR-1.7B is a fine-tuned ASR (Automatic Speech Recognition) model based on
 
 ## Usage
 
+### Installation
+
+```bash
+pip install -U qwen-asr
+```
+
 ### HuggingFace
 
 ```python
 import torch
-from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq
+from qwen_asr import Qwen3ASRModel
 
-model_id = "TaurenMountain/FormalASR-1.7B"
-
-processor = AutoProcessor.from_pretrained(model_id)
-model = AutoModelForSpeechSeq2Seq.from_pretrained(
-    model_id,
-    torch_dtype=torch.bfloat16,
-    device_map="auto"
+model = Qwen3ASRModel.from_pretrained(
+    "TaurenMountain/FormalASR-1.7B",
+    dtype=torch.bfloat16,
+    device_map="cuda:0",
+    max_new_tokens=512,
 )
 
-# Load audio (16kHz, mono)
-import librosa
-audio, sr = librosa.load("your_audio.wav", sr=16000)
-
-inputs = processor(audio, sampling_rate=16000, return_tensors="pt")
-inputs = {k: v.to(model.device) for k, v in inputs.items()}
-
-with torch.no_grad():
-    generated_ids = model.generate(**inputs)
+results = model.transcribe(
+    audio="your_audio.wav",
+    language="Chinese",
+)
 
-transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
-print(transcription)
+print(results[0].text)
 ```
 
 ### 魔搭社区（ModelScope）
@@ -71,30 +69,24 @@ print(transcription)
 ```python
 import torch
 from modelscope import snapshot_download
-from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq
+from qwen_asr import Qwen3ASRModel
 
 # 下载模型到本地（首次运行自动下载）
 model_dir = snapshot_download("TaurenMountain/FormalASR-1.7B")
 
-processor = AutoProcessor.from_pretrained(model_dir)
-model = AutoModelForSpeechSeq2Seq.from_pretrained(
+model = Qwen3ASRModel.from_pretrained(
     model_dir,
-    torch_dtype=torch.bfloat16,
-    device_map="auto"
+    dtype=torch.bfloat16,
+    device_map="cuda:0",
+    max_new_tokens=512,
 )
 
-# Load audio (16kHz, mono)
-import librosa
-audio, sr = librosa.load("your_audio.wav", sr=16000)
-
-inputs = processor(audio, sampling_rate=16000, return_tensors="pt")
-inputs = {k: v.to(model.device) for k, v in inputs.items()}
-
-with torch.no_grad():
-    generated_ids = model.generate(**inputs)
+results = model.transcribe(
+    audio="your_audio.wav",
+    language="Chinese",
+)
 
-transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
-print(transcription)
+print(results[0].text)
 ```
 
 ## Training Details