SWE-PolyBench (full set sample): every task, with its verdict
SWE-PolyBench, published by Amazon Science. Paper authors: Muhammad Shihab Rashid, Christian Bock, Yuan Zhuang, Alexander Buchholz, Tim Esler, Simon Valentin, Luca Franceschi, Martin Wistuba, Prabhu Teja Sivaprasad, Woo Jung Kim, Anoop Deoras, Giovanni Zappella and Laurent Callot. SWE-PolyBench (arXiv) · This page covers a random 100-task sample of the 2,110 tasks (AmazonScience/SWE-PolyBench on Hugging Face). The sample is the 100 instance IDs with the lowest SHA-256 of "own1:swe-polybench:<instance_id>", so anyone can redraw it. · Report a correction · flags show what the public text and git history contain: public exposure was possible, nothing more.
| Task | Verdict | Issue title |
|---|---|---|
| huggingface__transformers-13573 | no overlap found (rule check) | Upcasting of attention computation for reliable pretraining of GPT-2 models |
| huggingface__transformers-17053 | no overlap found (rule check) | Bug: Finetuning large models resume checkpoint error |
| huggingface__transformers-25429 | no overlap found (rule check) | NLLB MoE router_state referenced before assignment |
| huggingface__transformers-25765 | no overlap found (rule check) | RuntimeError: The size of tensor a (16) must match the size of tensor b (16000) at non-singleton dimension 2 |
| huggingface__transformers-26568 | no overlap found (rule check) | SWIN2SR: Allow to choose number of in_channels and out_channels |
| huggingface__transformers-29838 | no overlap found (rule check) | Trainer: Functions to inspect model and optimizer status |
| huggingface__transformers-31128 | no overlap found (rule check) | get_wsd_schedule gets passed num_training_steps because not handled |