Fix weights not updating during training (#6068)

Fix weights not updating during training (#6039)

Summary:
Loaded weights were not updating in ET because the returned weights were clones, not the original weights.

Pull Request resolved: #6039

Test Plan:
```
> buck2 run fbcode//executorch/examples/llm_pte_finetuning:runner -- --cfg=fbcode/executorch/examples/llm_pte_finetuning/phi3_config.yaml --model_file=phi3_mini_lora.pte
Evaluating the model before training...
100%|██████████████████████████████████████████████████████████████████████████████████████| 3/3 [31:23<00:00, 627.98s/it]
Eval loss:  tensor(2.3778)
grad_start:`1`
param_start:`129`
100%|██████████████████████████████████████████████████████████████████████████████████████| 5/5 [52:29<00:00, 629.84s/it]
Losses:  [2.7152762413024902, 0.7890686988830566, 2.249271869659424, 1.4777560234069824, 0.8378427624702454]
100%|██████████████████████████████████████████████████████████████████████████████████████| 3/3 [30:35<00:00, 611.90s/it]
Eval loss:  tensor(0.8464)
```

Reviewed By: dpalmasan

Differential Revision: D64084552

Pulled By: dvorjackz

fbshipit-source-id: 9d478dda02f7bcaa5964d83d257d0db5bfe9feab
(cherry picked from commit 867c96a)

Co-authored-by: Jack Zhang <dvorjackz@gmail.com>

examples/llm_pte_finetuning/runner.py

@@ -94,7 +94,11 @@ def main() -> None:
             else:
                 labels = F.pad(labels, (0, max_seq_len - labels_size), value=0)
 
-            out = et_mod.forward((tokens, labels))
+            # Do not clone outputs, since we want the original weights to be returned
+            # for us to update with the gradients in-place.
+            # See https://github.com/pytorch/executorch/blob/main/extension/pybindings/pybindings.cpp#L736
+            # for more info.
+            out = et_mod.forward((tokens, labels), clone_outputs=False)  # pyre-ignore
 
             loss = out[0]
             losses.append(loss.item())