spb/localvm-research Public License
Running LLMs larger than memory on a consumer Mac — falsification-driven research: margin-gated deferred refinement, out-of-core verification on Apple Silicon. TR-01 published.
Python 63.2%
JavaScript 23.5%
CSS 11.8%
Shell 0.9%
Makefile 0.5%
1loading mlx-community/Qwen3-1.7B-bf16 …2
Fetching 9 files: 0%| | 0/9 [00:00<?, ?it/s]
Fetching 9 files: 100%|██████████| 9/9 [00:00<00:00, 5740.38it/s]348 trajectories in 48s4 traced 12/485 traced 24/486 traced 36/487 traced 48/488trace shape (6144, 28, 384)910wrote /Users/simon-pierreboucher/Desktop/localvm-research/results/expA_weight_concentration/20260812T044829Z/results.json (+ block_energy_trace.npz for expB)11overall block-64: {"top_frac_energy": {"0.1": 0.5811625123023987, "0.2": 0.7119550108909607, "0.4": 0.8502882122993469, "0.6": 0.9268651604652405}, "frac_needed_for": {"0.9": 0.47826662517729257, "0.95": 0.6121286816067165, "0.99": 0.8104762122744605}}12neuron-level : {"top_frac_energy": {"0.1": 0.8779218658982288, "0.2": 0.9387342477483409, "0.4": 0.9827663398658236, "0.6": 0.9962594589839379}, "frac_needed_for": {"0.9": 0.12640662041921463, "0.95": 0.1993891048052954, "0.99": 0.37685486532392953}}13