BugFix: Modify TPS metric calculation. Add default cpu threads for hybrid CPU system.

[hebangwen]

Summary

This PR introduces two minor fix for Llama runner:

1. Modify TPS metric calculation. Currently, TPS metric uses all token as the numerator, including the tokens from prefilling, not generated tokens. So we need to get the generated tokens by decoding, i.e. len(tokens) - len(prompt_tokens) - 1.
2. Use a default CPU threads setting. In hybrid CPU system, using all CPU threads will make a pretty high inference latency due to high workloads in efficient cores. So I set num_threads to 4 as a general setting.

Release notes: examples

Test plan

CPU: Intel Core-14900KF. 8 performance cores and 16 efficient cores, with total 32 CPU threads.

System: Ubuntu 22.04

Testing command:

python -m examples.models.llama.runner.native \
--model qwen3-4b \
--pte qwen3-4b.pte \
--tokenizer ~/.cache/huggingface/hub/models--Qwen--Qwen3-4B/snapshots/531c80e289d6cff3a7cd8c0db8110231d23a6f7a/tokenizer.json \
--tokenizer_config ~/.cache/huggingface/hub/models--Qwen--Qwen3-4B/snapshots/531c80e289d6cff3a7cd8c0db8110231d23a6f7a/tokenizer_config.json \
--prompt "Write a hello-world programme in c++. Code is:" \
--params examples/models/qwen3/4b_config.json \
--max_len 20 \
-kv \
--temperature 0.8

without this PR

INFO:root:Looking for libcustom_ops_aot_lib.so in executorch/extension/llm/custom_ops
INFO:root:Loading custom ops library: executorch/extension/llm/custom_ops/libcustom_ops_aot_lib.so
Warning - given vocab_size in params is unequal to tokenizer vocab size.
[program.cpp:134] InternalConsistency verification requested but not available
 #include <iostream> using namespace std;

Prefill time: 0.20537757873535156
Generation tok/s: 0.19834848596001395
Response: [671, 997, 366, 9665, 29, 1667, 4473, 1460, 26]

with this PR

INFO:root:Looking for libcustom_ops_aot_lib.so in executorch/extension/llm/custom_ops
INFO:root:Loading custom ops library: executorch/extension/llm/custom_ops/libcustom_ops_aot_lib.so
Warning - given vocab_size in params is unequal to tokenizer vocab size.
[program.cpp:134] InternalConsistency verification requested but not available
 #include <iostream> using namespace std;

Prefill time: 0.29981303215026855
Generation tok/s: 15.131355739212273
Generation tok/s: 37.82838934803068
Response: [671, 997, 366, 9665, 29, 1667, 4473, 1460, 26]

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/11063

• 📄 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, 1 Cancelled Job, 1 Unrelated Failure

As of commit fbc366a with merge base 80250f8 ():

NEW FAILURE - The following job has failed:

• Build documentation / build (buck2) / Build doc (gh)
  At least one of the pre-conditions you specified did not hold

CANCELLED JOB - The following job was cancelled. Please retry:

• pull / unittest-editable / macos / macos-job (gh)
  ##[error]The operation was canceled.

BROKEN TRUNK - The following job failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• pull / test-binary-size-linux-gcc / linux-job (gh) (trunk failure)
  /pytorch/executorch/kernels/portable/cpu/op_stack.cpp:129:26: error: comparison of integer expressions of different signedness: ‘size_t’ {aka ‘long unsigned int’} and ‘ssize_t’ {aka ‘long int’} [-Werror=sign-compare]

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

[facebook-github-bot]

Hi @hebangwen!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at cla@meta.com. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!

[hebangwen]

@pytorchbot label "release notes: examples"

[jackzhxng]

Thanks for the fix. Curious what speedup you saw from setting the num threads to your p-core count?

[kimishpatel]

Thank you for your contribution

[digantdesai]

Generation tok/s: 15.131355739212273
Generation tok/s: 37.82838934803068

Why two lines?

[github-actions]

Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as Stale.
Feel free to remove the Stale label if you feel this was a mistake.
If you are unable to remove the Stale label please contact a maintainer in order to do so.
If you want the bot to never mark this PR stale again, add the no-stale label.
Stale pull requests will automatically be closed after 30 days of inactivity.