fix eager run for cuda

[helunwencser]

Stack from ghstack (oldest at bottom):

• -> fix eager run for cuda #6365

This PR allows us to run the eager runner on GPU machines.

Test Plan:

python -m examples.models.llama.runner.eager \
          -c /home/lunwenh/models/1B_Instruct/consolidated.00.pth \
          -p /home/lunwenh/models/1B_Instruct/params.json \
          -t /home/lunwenh/models/1B_Instruct/tokenizer.model \
          -kv \
          -d bf16 \
          --max_seq_length 128 \
          --prompt "<|begin_of_text|><|start_header_id|>system<|end_header_id|>

You are a good assistant<|eot_id|><|start_header_id|>user<|end_header_id|>

What is the capital of France?<|eot_id|><|start_header_id|>assistant<|end_header_id|>"

Response: 
The capital of France is Paris.
 Tokens:
 [791, 6864, 315, 9822, 374, 12366, 13]

Differential Revision: D64713144

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/6365

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure

As of commit 13a3ade with merge base 1e9ebdf ():

NEW FAILURE - The following job has failed:

• Propose to merge ghstack orig PRs to main / Try to create a PR with ghstack /orig branch (gh)
  AttributeError: 'NoneType' object has no attribute 'splitlines'

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[helunwencser]

@helunwencser has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[helunwencser]

@helunwencser has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[helunwencser]

@helunwencser has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.