update alpaca gpt4 to use dataset entry (#2869)

#2827

Update alpaca gpt4 to use dataset entry.

I ran 
```bash
~$ python check_dataset_appearances.py -d alpaca_gpt4 --cache_dir .cache --mode sft
'Found the following occurances in TRAIN alpaca_gpt4:'
{   re.compile('\\[\\d+(?:,\\s*\\d+)*?\\]'): [   '[3, 45, 99, 2, 8, 6, 72]',
                                                 '[10, 8, 7, 4]',
                                                 '[1, 2, 3, 4, 5]',
                                                 '[7, 3, 4, 6, 2]']}
'Found the following occurances in VAL alpaca_gpt4:'
{   'openai': [   'u’re approved, get your API key.\n'
                  '\n'
                  '2. Install the `openai` library in your Python environment '
                  'using p']}
```
Checked all the occurances for the references, but all are programming
related and have nothing to do with references, so this looks fine:
```python
DatasetEntry(questions=['Re-order the integer list given in the input field such that 
                                         all odd numbers are first and even numbers are last.
                                         \n[2, 3, 8, 45, 6, 99, 72]'], 
                      answers=['[3, 45, 99, 2, 8, 6, 72]'],
                      context=None, 
                      lang=None, 
                      length=None, 
                      quality=None, 
                      humor=None, 
                      creativity=None)
```

model/model_training/custom_datasets/qa_datasets.py

@@ -603,7 +603,7 @@ def __init__(self, cache_dir: str | Path, mode: str = "sft", input_max_length: i
             if (conv := self._process_instruction(line, input_max_length)) is not None:
                 self.rows.append(conv)
 
-    def _process_instruction(self, row: dict[str, str], input_max_length: int) -> list[str] | None:
+    def _process_instruction(self, row: dict[str, str], input_max_length: int) -> DatasetEntry | None:
         # discard items that are too long: when checked on 2023-04-17 this was just one item in the whole dataset with length above 2048.
         # And 12 above 1024.
         if len(row["input"]) + len(row["instruction"]) > input_max_length:
@@ -615,18 +615,17 @@ def _process_instruction(self, row: dict[str, str], input_max_length: int) -> li
             or (not row["input"])
             or (row["input"].lower() in row["instruction"].lower())
         ):
-            return [row["instruction"], row["output"]]
+            return DatasetEntry(questions=[row["instruction"]], answers=[row["output"]])
         # Concatenate the instruction and input.
         else:
             linking_char = random.choice(LINKING_CHARS)
-            return [f"{row['instruction']}{linking_char}{row['input']}", row["output"]]
+            return DatasetEntry(
+                questions=[f"{row['instruction']}{linking_char}{row['input']}"], answers=[row["output"]]
+            )
 
     def __len__(self) -> int:
         return len(self.rows)
 
     def __getitem__(self, index: int) -> list[str] | tuple[str]:
-        dialogue: list[str] = self.rows[index]
-        if self.mode == "sft":
-            return dialogue
-        elif self.mode == "rl":
-            return tuple(dialogue[:-1])
+        dialogue = self.rows[index]
+        return dialogue