Add MLX support for draft model

[metascroy]

Adds Eagle3 MLX support. MLX supports more features than CUDA, including sharing embedding weights between draft/base model, quantization of draft model, runtime selection of chains, and runtime temperature sampling (not just greedy).

Export command:

python -m executorch.examples.models.eagle3.export \
  --backend mlx --target-model gemma4_31b \
  --target ./gemma-4-31B-it-HQQ-INT4 --draft "$SPEC" \
  --output-dir ./eagle3_mlx --chain 3 --max-prefill 256 \
  --quantize-draft --share-draft-embedding

Build command:

 make eagle3-mlx 

Run command:

./cmake-out/examples/models/eagle3/eagle3_speculator_runner \
  --model_path ./eagle3_mlx/model.pte \
  --tokenizer_path ./gemma-4-31B-it-HQQ-INT4/tokenizer.json \
  --prompt "Write a short joke about saving RAM." \
  --max_new_tokens 128 --chain 3 --temperature 0.8
I tokenizers:regex.cpp:27] Registering override fallback regex
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
E0000 00:00:1781159777.559320 23022209 re2.cc:237] Error parsing '((\<pad\>|ool\|\>1\x00\x00\
             �\<t|respo|\<tool_call\|\>|\<bos\>|\<\|tool_response\>|\<\|think\|\>|\x0...': invalid UTF-8
I tokenizers:re2_regex.cpp:27] Re2 failed to compile regex: ((\<pad\>|ool\|\>1\x00\x00\
       �\<t|respo|\<tool_call\|\>|\<bos\>|\<\|tool_response\>|\<\|think\|\>|\x00\x00\\\<|\<tool_response\|\>|\<mask\>|\<\|\"\|\>|all\|\>j\x00\x00\\|\<channel\|\>|\<\|turn\>|\<turn\|\>|\<\|image\>|\<\|$
I tokenizers:regex_lookahead.cpp:27] Creating PCRE2 regex
I tokenizers:pcre2_regex.cpp:48] PCRE2 UTF-8 validation failed at offset 27: UTF-8 error: byte 2 top bits not 0x80. Retrying without UTF flags.
Prompt tokens: 24, chain K=3, max_new=128
Why did the computer programmer refuse to buy a new car?

Because he didn't have enough **space**, but he had plenty of **room** for improvement.

***

(Or, if you prefer a one-liner:)

**I tried to write a joke about saving RAM, but I decided to just keep it in my head to save space.**<turn|>
[timing] verify=2946ms draft=154ms over 33 rounds (89.3 / 4.7 ms per round)

--- EAGLE-3 speculative decode ---
generated 75 tokens in 33 rounds (tau=2.242)
decode: 23.60 tok/s (3.178 s)
PyTorchObserver {"prefill_token_per_sec":84.507,"decode_token_per_sec":23.5997,"prompt_tokens":24,"generated_tokens":75,"model_load_start_ms":1781159777286,"model_load_end_ms":1781159778444,"inference_start_ms":1781159778444,"inference_end_ms":1781159781906,"prompt_eval_end_ms":1781159778728,"first_token_ms":1781159778728,"aggregate_sampling_time_ms":0,"SCALING_FACTOR_UNITS_PER_SECOND":1000}

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20210

• 📄 Preview Python docs built from this PR

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

❌ 4 New Failures, 2 Unclassified Failures

As of commit 263bc33 with merge base dc55469 ():

NEW FAILURES - The following jobs have failed:

• pull / unittest / macos / macos-job (gh)
  The process '/usr/bin/git' failed with exit code 1
• pull / unittest-buck / macos / macos-job (gh)
  The process '/usr/bin/git' failed with exit code 1
• pull / unittest-editable / macos / macos-job (gh)
  The process '/usr/bin/git' failed with exit code 1
• pull / unittest-nxp-neutron / linux-job (gh)
  RuntimeError: Command docker exec -t 6f6b6a2ab0adcc085b5c6892d36aa29e332c7563c739c9ffa27110ea3a41d3fa /exec failed with exit code 1

UNCLASSIFIED FAILURES - DrCI could not classify the following jobs because the workflow did not run on the merge base. The failures may be pre-existing on trunk or introduced by this PR:

• Build Windows Wheels / pytorch/executorch / build-wheel-py3_10-cpu (gh) (this job did not run on the merge base, so DrCI cannot tell whether the failure is pre-existing)
• Build Windows Wheels / pytorch/executorch / upload / upload-wheel-py3_10-cpu (gh) (this job did not run on the merge base, so DrCI cannot tell whether the failure is pre-existing)
  Unable to download artifact(s): Artifact not found for name: pytorch_executorch__3.10_cpu_x64

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