minor fix for docstr and default param. (#1490)

* Update train.py and README.md

README.md

@@ -74,6 +74,9 @@ The [LibriSpeech][librispeech] recipe supports the most comprehensive set of mod
   - LSTM-based Predictor
   - [Stateless Predictor](https://research.google/pubs/rnn-transducer-with-stateless-prediction-network/)
 
+#### Whisper
+  - [OpenAi Whisper](https://arxiv.org/abs/2212.04356) (We support fine-tuning on AiShell-1.)
+
 If you are willing to contribute to icefall, please refer to [contributing](https://icefall.readthedocs.io/en/latest/contributing/index.html) for more details.
 
 We would like to highlight the performance of some of the recipes here.

egs/aishell/ASR/whisper/train.py

@@ -19,7 +19,7 @@
 Usage:
 
 #fine-tuning with deepspeed zero stage 1
-torchrun --nproc-per-node 8 ./whisper/train.py \
+torchrun --nproc_per_node 8 ./whisper/train.py \
   --max-duration 200 \
   --exp-dir whisper/exp_large_v2 \
   --model-name large-v2 \
@@ -28,7 +28,7 @@ torchrun --nproc-per-node 8 ./whisper/train.py \
   --deepspeed_config ./whisper/ds_config_zero1.json
 
 # fine-tuning with ddp
-torchrun --nproc-per-node 8 ./whisper/train.py \
+torchrun --nproc_per_node 8 ./whisper/train.py \
   --max-duration 200 \
   --exp-dir whisper/exp_medium \
   --manifest-dir data/fbank_whisper \
@@ -136,7 +136,7 @@ def get_parser():
     parser.add_argument(
         "--exp-dir",
         type=str,
-        default="pruned_transducer_stateless7/exp",
+        default="whisper/exp",
         help="""The experiment dir.
         It specifies the directory where all training related
         files, e.g., checkpoints, log, etc, are saved