Fix setting joiner dim (#2027)

Fixes incorrect computation of encoder_dim when encoder_dim is a comma-separated list of integers by ensuring numeric (not lexicographic) max is used.

Fixes #2018

- Replace int(max(params.encoder_dim.split(","))) (lexicographic max on strings) with max(_to_int_tuple(params.encoder_dim)) (numeric max).
- Apply the fix consistently across all affected training scripts.

egs/aishell/ASR/zipformer/train.py

@@ -639,7 +639,7 @@ def get_model(params: AttributeDict) -> nn.Module:
         encoder=encoder,
         decoder=decoder,
         joiner=joiner,
-        encoder_dim=int(max(params.encoder_dim.split(","))),
+        encoder_dim=max(_to_int_tuple(params.encoder_dim)),
         decoder_dim=params.decoder_dim,
         vocab_size=params.vocab_size,
         use_transducer=params.use_transducer,

egs/libriheavy/ASR/zipformer_prompt_asr/train_baseline.py

@@ -651,7 +651,7 @@ def get_transducer_model(params: AttributeDict) -> nn.Module:
         encoder=encoder,
         decoder=decoder,
         joiner=joiner,
-        encoder_dim=int(max(params.encoder_dim.split(","))),
+        encoder_dim=max(_to_int_tuple(params.encoder_dim)),
         decoder_dim=params.decoder_dim,
         joiner_dim=params.joiner_dim,
         vocab_size=params.vocab_size,

egs/libriheavy/ASR/zipformer_prompt_asr/train_bert_encoder.py

@@ -881,7 +881,7 @@ def get_transducer_model(params: AttributeDict) -> nn.Module:
         text_encoder=text_encoder,
         decoder=decoder,
         joiner=joiner,
-        encoder_dim=int(max(params.encoder_dim.split(","))),
+        encoder_dim=max(_to_int_tuple(params.encoder_dim)),
         decoder_dim=params.decoder_dim,
         joiner_dim=params.joiner_dim,
         vocab_size=params.vocab_size,

egs/mdcc/ASR/zipformer/train.py

@@ -586,7 +586,7 @@ def get_model(params: AttributeDict) -> nn.Module:
         encoder=encoder,
         decoder=decoder,
         joiner=joiner,
-        encoder_dim=int(max(params.encoder_dim.split(","))),
+        encoder_dim=max(_to_int_tuple(params.encoder_dim)),
         decoder_dim=params.decoder_dim,
         vocab_size=params.vocab_size,
     )

egs/tedlium3/ASR/zipformer/train.py

@@ -598,7 +598,7 @@ def get_model(params: AttributeDict) -> nn.Module:
         encoder=encoder,
         decoder=decoder,
         joiner=joiner,
-        encoder_dim=int(max(params.encoder_dim.split(","))),
+        encoder_dim=max(_to_int_tuple(params.encoder_dim)),
         decoder_dim=params.decoder_dim,
         joiner_dim=params.joiner_dim,
         vocab_size=params.vocab_size,

egs/wenetspeech/ASR/zipformer/train.py

@@ -590,7 +590,7 @@ def get_model(params: AttributeDict) -> nn.Module:
         encoder=encoder,
         decoder=decoder,
         joiner=joiner,
-        encoder_dim=int(max(params.encoder_dim.split(","))),
+        encoder_dim=max(_to_int_tuple(params.encoder_dim)),
         decoder_dim=params.decoder_dim,
         vocab_size=params.vocab_size,
     )

egs/wenetspeech/KWS/zipformer/train.py

@@ -647,7 +647,7 @@ def get_model(params: AttributeDict) -> nn.Module:
         encoder=encoder,
         decoder=decoder,
         joiner=joiner,
-        encoder_dim=int(max(params.encoder_dim.split(","))),
+        encoder_dim=max(_to_int_tuple(params.encoder_dim)),
         decoder_dim=params.decoder_dim,
         vocab_size=params.vocab_size,
     )