Fix for dataset and model variables Errors when downloading fairseq models

in the original implementation, the dataset and model variables were causing errors due to uninitialized in the TTS fairseq models.

Changes Made:

Introduced a line to split model_name into model_attributes based on the '/' delimiter. This ensures that the different components of the model name (language, dataset, model) can be separately and accurately accessed.
Added lines to explicitly extract lang, dataset, and model from model_attributes using their respective indices. This change makes the code more robust and clear, as it directly links the variables to their sources:
lang is now set to the second component (model_attributes[1]),
dataset to the third (model_attributes[2]),
model to the fourth (model_attributes[3]).

TTS/utils/manage.py

@@ -260,8 +260,9 @@ class ModelManager(object):
     def _set_model_item(self, model_name):
         # fetch model info from the dict
         if "fairseq" in model_name:
+            model_attributes = model_name.split("/")
             model_type = "tts_models"
-            lang = model_name.split("/")[1]
+            lang = model_attributes[1]
             model_item = {
                 "model_type": "tts_models",
                 "license": "CC BY-NC 4.0",
@@ -270,6 +271,9 @@ class ModelManager(object):
                 "description": "this model is released by Meta under Fairseq repo. Visit https://github.com/facebookresearch/fairseq/tree/main/examples/mms for more info.",
             }
             model_item["model_name"] = model_name
+            dataset = model_attributes[2]
+            model = model_attributes[3]
+            
         elif "xtts" in model_name and len(model_name.split("/")) != 4:
             # loading xtts models with only model name (e.g. xtts_v2.0.2)
             # check model name has the version number with regex