fix: check for tokenizer eos_token in ModelInfo response (IBM#93)

Use consistent logic to determine the eos_token_id in ModelInfo as it is
in other functions by falling back to the tokenizer's `eos_token_id`
attribute if the model config does not have an `eos_token_id`.

Fixes the behavior for a model that does not have an eos_token_id
in the model config

Resolves IBM#91

Signed-off-by: Travis Johnson <[email protected]>

Author: tjohnson31415

server/text_generation_server/server.py

@@ -79,7 +79,7 @@ async def ModelInfo(self, request: generate_pb2.ModelInfoRequest, context) -> ge
         return generate_pb2.ModelInfoResponse(
             model_type=ModelInfoResponse.ModelType.SEQ2SEQ_LM
                 if isinstance(self.model, Seq2SeqLM) else ModelInfoResponse.ModelType.CAUSAL_LM,
-            eos_token=self.model.config.eos_token_id,
+            eos_token=getattr(self.model.tokenizer, 'model_eos_token_id', self.model.tokenizer.eos_token_id),
             batch_padding=not isinstance(self.model, FlashCausalLM),
             memory_scaling_model=self.memory_scaling_model,
         )