bonsai: depth0/experts32
Pruned and distilled from Qwen3-30B-A3B-Base for TypeScript code completion. Part of the bonsai project, which applies the Minitron structured-pruning and knowledge-distillation recipe (Muralidharan et al. 2024, Sreenivas et al. 2024) to a Mixture-of-Experts model, a pruning axis neither paper covers.
What was pruned
- Depth: 0 of 48 layers removed (48 kept)
- Experts: 32 of 128 routed experts removed per layer (96 kept)
- Result: ~23.28B parameters (76.3% of baseline's ~30.5B)
Training
500-step fixed-loss short schedule (75-step warmup, peak LR 1e-4), distilled against the uncorrected teacher. Best-validation checkpoint, step 29.
Results
HumanEval-TS (TypeScript MultiPL-E), full 159-problem suite, temperature 0.2, top-p 0.95:
| pass@1 | |
|---|---|
| Qwen3-30B-A3B-Base (unpruned) | 0.717 (114/159) |
| This model | 0.604 (96/159) |
84.2% of baseline accuracy at 76.3% of baseline's parameters.
Notes
Mid-range experts-only cut. See depth0/experts64 for the recommended candidate at a smaller size.
Project
Other checkpoints from this sweep: bonsai collection. Code and full writeup: github.com/SartajBhuvaji/bonsai (private repo).