Update Transformers quickstart for PR 46180

Pin the verified rebased PR head, document structured language/context/hotword inputs, and remove the obsolete serialized default prompt field.

README.md

@@ -40,7 +40,7 @@ With [PyTorch installed for your platform](https://pytorch.org/get-started/local
 
 ```bash
 python -m pip install accelerate librosa \
-  "transformers @ https://github.com/huggingface/transformers/archive/d58b6371091473c7bd1486f3147950c6e076ace9.zip"
+  "transformers @ https://github.com/huggingface/transformers/archive/4e199f7990ab38838210fc02726f34f6f25d0097.zip"
 ```
 
 ```python
@@ -56,11 +56,15 @@ model = AutoModelForSpeechSeq2Seq.from_pretrained(
 )
 
 audio_url = "https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512/resolve/main/example/en.mp3"
-inputs = processor.apply_transcription_request(audio=audio_url, return_tensors="pt").to(model.device)
+inputs = processor.apply_transcription_request(
+    audio=audio_url,
+    language="en",
+    return_tensors="pt",
+).to(model.device)
 
 generated_ids = model.generate(**inputs, max_new_tokens=200)
 generated_ids = generated_ids[:, inputs.input_ids.shape[1] :]
-print(processor.batch_decode(generated_ids, skip_special_tokens=True)[0])
+print(processor.decode(generated_ids, skip_special_tokens=True)[0])
 ```
 
 Expected transcription:
@@ -69,4 +73,16 @@ Expected transcription:
 The tribal chieftain called for the boy, and presented him with fifty pieces of gold.
 ```
 
-For batch inference, training, `torch.compile`, benchmarks, and the full model description, see the [Transformers documentation](https://github.com/huggingface/transformers/blob/main/docs/source/en/model_doc/fun_asr_nano.md) and the [original model card](https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512).
+For domain-specific transcription, pass contextual information with `prompt` and hotwords with `keywords`; the checkpoint chat template builds the complete instruction:
+
+```python
+inputs = processor.apply_transcription_request(
+    audio=audio_url,
+    language="en",
+    prompt="A tribal story involving a chieftain and a boy.",
+    keywords=["tribal chieftain", "fifty pieces of gold"],
+    return_tensors="pt",
+).to(model.device)
+```
+
+`language` accepts Chinese, English, and Japanese as ISO codes, full English names, or native names. For batch inference, training, `torch.compile`, benchmarks, and the full model description, see the [Transformers documentation](https://github.com/huggingface/transformers/blob/4e199f7990ab38838210fc02726f34f6f25d0097/docs/source/en/model_doc/fun_asr_nano.md) and the [original model card](https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512).

processor_config.json

@@ -1,6 +1,5 @@
 {
   "audio_token": "<|object_ref_start|>",
-  "default_transcription_prompt": "Transcribe the audio:",
   "feature_extractor": {
     "feature_extractor_type": "FunAsrNanoFeatureExtractor",
     "feature_size": 80,