[LAB 01 · 固定表示：标签预算与读出](/course/notebooks/4dd8605ed9904820919ea9e13d9e7d7c.html)

# 4 · 写出竞争解释 · 实现代码
[独立阅读](/course/notebook-blocks/b141f26853a042f4970bc2e0023a04ed.html)
按问题继续

固定CNN与缓存特征、标签预算、kNN/linear和独立评价
展开完整实现代码
```
# 当前单元无法确认宿主平台或是否使用全新进程；unknown 不等于 False。
verification={'identity':'v17 independent frozen readout','fresh_process':'unknown','local_run':'unknown','cloud_run':'unknown','source_model_unchanged':all(torch.equal(v,before[k]) for k,v in model.state_dict().items()),'source_target_sample_overlap':False,'source_target_writer_overlap':'unknown','heldout_used_for_tuning':False,'clip_loaded':False,'environment':{'python':platform.python_version(),'torch':torch.__version__,'numpy':np.__version__,'sklearn':sklearn.__version__},'checkpoint_sha256':sha_file(ROOT/'assets/cnn-plain-state.pt'),'target_sha256':sha_file(ROOT/'assets/target.npz')}
(OUT/'verification.json').write_text(json.dumps(verification,indent=2));print(json.dumps(verification,indent=2))

```