[transformers] The tokenizer you are loading from '/Users/vodkaexpans/.cache/huggingface/hub/models--mlx-community--Devstral-Small-2-24B-Instruct-2512-4bit/snapshots/97228558fcbdd1adefd39d515155db3a894b37f7' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
Running warmup..
Timing with prompt_tokens=1024, generation_tokens=512, batch_size=1.
Trial 1:  prompt_tps=500.092, generation_tps=18.305, peak_memory=15.088, total_time=30.033
Trial 2:  prompt_tps=436.248, generation_tps=18.105, peak_memory=15.089, total_time=30.641
Trial 3:  prompt_tps=466.936, generation_tps=18.182, peak_memory=15.089, total_time=30.368
Trial 4:  prompt_tps=457.929, generation_tps=18.075, peak_memory=15.089, total_time=30.578
Trial 5:  prompt_tps=456.646, generation_tps=18.083, peak_memory=15.089, total_time=30.571
Averages: prompt_tps=463.570, generation_tps=18.150, peak_memory=15.089