![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2062758%2Ff9faa54f9c0a89e43aa7267cba97b23e%2Fimage%20(14).png?generation=1768576990418111&alt=media\")
We’re excited to share an inference workflow for RNAPro for part 2 of the Stanford RNA 3D Folding competition ! RNAPro reflects where the field landed after last year’s breakthroughs.
\nRNAPro is a state-of-the-art RNA 3D folding model developed through a synthesis of ideas from the previous Kaggle competition hosts and top-performing teams. It integrates several components that proved valuable in the previous competition and subsequent work:
\nThis combination represents some of the best insights from the previous competition !
\n⭐ The code is available on GitHub ⭐
\n
We’re sharing two Kaggle notebooks that together create a complete inference and submission workflow:
\n\nThese notebooks are intended to provide a strong baseline and starting point for the community, while being completely open.
\nWe are looking forward to seeing how the community builds on this baseline !
\nWe’re excited to share an inference workflow for RNAPro for part 2 of the Stanford RNA 3D Folding competition ! RNAPro reflects where the field landed after last year’s breakthroughs.
\nRNAPro is a state-of-the-art RNA 3D folding model developed through a synthesis of ideas from the previous Kaggle competition hosts and top-performing teams. It integrates several components that proved valuable in the previous competition and subsequent work:
\nThis combination represents some of the best insights from the previous competition !
\n⭐ The code is available on GitHub ⭐
\n
We’re sharing two Kaggle notebooks that together create a complete inference and submission workflow:
\n\nThese notebooks are intended to provide a strong baseline and starting point for the community, while being completely open.
\nWe are looking forward to seeing how the community builds on this baseline !
\nMy assumption is that over the next few days you will learn lots of ways your git hub install directions have issues :)
\nI will start with mine - I absolutely hate with a strong passion code that downloads a large zip, extracts it and than deletes it. In my case I have 4 local machines were I play with kaggle. I would like to copy the zip to my NAS and than get things installed on all 4 machines. Sadly - Did not notice the rm's until it was too late.
\ncd release_data\nmkdir kaggle; cd kaggle\ncurl -L -o stanford-rna-3d-folding-all-atom-train-data.zip https://www.kaggle.com/api/v1/datasets/download/rhijudas/stanford-rna-3d-folding-all-atom-train-data\nunzip stanford-rna-3d-folding-all-atom-train-data.zip\nrm stanford-rna-3d-folding-all-atom-train-data.zip\n\n# Get MSAs from https://www.kaggle.com/datasets/rhijudas/clone-of-stanford-rna-3d-modeling-competition-data\ncurl -L -o stanford-rna-3d-folding.zip https://www.kaggle.com/api/v1/datasets/download/rhijudas/clone-of-stanford-rna-3d-modeling-competition-data\nmkdir MSA_v2\nunzip stanford-rna-3d-folding.zip \"MSA_v2/*\" -d /MSA_v2\nrm stanford-rna-3d-folding.zip\ncd ..\nIn line 9, you have: unzip stanford-rna-3d-folding.zip \"MSA_v2/*\" -d /MSA_v2
\nThe flag -d /MSA_v2 attempts to extract the files to the root of your file system (e.g., /MSA_v2), which will likely fail with a \"Permission denied\" error unless you are running as root.
\nSince you created a local directory in line 8 (mkdir MSA_v2), you likely meant to extract to that local folder. You should remove the leading slash:
", "rawMarkdown": "In line 9, you have: unzip stanford-rna-3d-folding.zip \"MSA_v2/*\" -d /MSA_v2\n\nThe flag -d /MSA_v2 attempts to extract the files to the root of your file system (e.g., /MSA_v2), which will likely fail with a \"Permission denied\" error unless you are running as root.\n\nSince you created a local directory in line 8 (mkdir MSA_v2), you likely meant to extract to that local folder. You should remove the leading slash:", "votes": 1, "replies": [ { "id": 3392319, "postDate": "2026-01-16T17:41:59.493Z", "content": "I am guessing you changed folders and names at some point - but this does not work as no scripts folder seems to exist in anything I created so far. But that py can be found in the RNAPro/preprocess/ folder.
\npython3 scripts/gen_ccd_cache.py
This one takes some time - guessing that a few folks are hitting that server right now :) Right now for me on a 1TB Xfinity system I had 1 hour and 10 minutes of download time. Did the Xfinity speed test and my current download is 922 Mbps - so assuming the time is mostly the result of the server rather than my system.
", "rawMarkdown": "I am guessing you changed folders and names at some point - but this does not work as no scripts folder seems to exist in anything I created so far. But that py can be found in the RNAPro/preprocess/ folder. \n\n`python3 scripts/gen_ccd_cache.py`\n\nThis one takes some time - guessing that a few folks are hitting that server right now :) Right now for me on a 1TB Xfinity system I had 1 hour and 10 minutes of download time. Did the Xfinity speed test and my current download is 922 Mbps - so assuming the time is mostly the result of the server rather than my system. " }, { "id": 3392320, "postDate": "2026-01-16T17:42:10.560Z", "content": "\n\nMy assumption is that over the next few days you will learn lots of ways your git hub install directions have issues :)
\n
Yes, I already caught a few installing for inference on Kaggle, but I did not setup for training (yet?) so it can have issues.
\nWe'll fix them asap, thanks for the feedback :)
", "rawMarkdown": "> My assumption is that over the next few days you will learn lots of ways your git hub install directions have issues :)\n\nYes, I already caught a few installing for inference on Kaggle, but I did not setup for training (yet?) so it can have issues.\n\nWe'll fix them asap, thanks for the feedback :) " }, { "id": 3392321, "postDate": "2026-01-16T17:43:43.590Z", "content": "\n\npython3 scripts/gen_ccd_cache.py
\n
This one is known, I have a MR that fixes it
", "rawMarkdown": "> python3 scripts/gen_ccd_cache.py\n\nThis one is known, I have a MR that fixes it " }, { "id": 3392345, "postDate": "2026-01-16T18:33:19.933Z", "content": "In case you did not guess :) I am working on use for multiple local machines, so my issues might be a bit different than a kaggle install.
", "rawMarkdown": "In case you did not guess :) I am working on use for multiple local machines, so my issues might be a bit different than a kaggle install." } ] } ] }, { "id": 3419651, "postDate": "2026-03-11T09:28:33.200Z", "content": "thanks for share
", "rawMarkdown": "thanks for share" } ], "comments": [ { "id": 3392306, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2026-01-16T17:12:23.863000", "content": "My assumption is that over the next few days you will learn lots of ways your git hub install directions have issues :)
\nI will start with mine - I absolutely hate with a strong passion code that downloads a large zip, extracts it and than deletes it. In my case I have 4 local machines were I play with kaggle. I would like to copy the zip to my NAS and than get things installed on all 4 machines. Sadly - Did not notice the rm's until it was too late.
\ncd release_data\nmkdir kaggle; cd kaggle\ncurl -L -o stanford-rna-3d-folding-all-atom-train-data.zip https://www.kaggle.com/api/v1/datasets/download/rhijudas/stanford-rna-3d-folding-all-atom-train-data\nunzip stanford-rna-3d-folding-all-atom-train-data.zip\nrm stanford-rna-3d-folding-all-atom-train-data.zip\n\n# Get MSAs from https://www.kaggle.com/datasets/rhijudas/clone-of-stanford-rna-3d-modeling-competition-data\ncurl -L -o stanford-rna-3d-folding.zip https://www.kaggle.com/api/v1/datasets/download/rhijudas/clone-of-stanford-rna-3d-modeling-competition-data\nmkdir MSA_v2\nunzip stanford-rna-3d-folding.zip \"MSA_v2/*\" -d /MSA_v2\nrm stanford-rna-3d-folding.zip\ncd ..\nIn line 9, you have: unzip stanford-rna-3d-folding.zip \"MSA_v2/*\" -d /MSA_v2
\nThe flag -d /MSA_v2 attempts to extract the files to the root of your file system (e.g., /MSA_v2), which will likely fail with a \"Permission denied\" error unless you are running as root.
\nSince you created a local directory in line 8 (mkdir MSA_v2), you likely meant to extract to that local folder. You should remove the leading slash:
", "votes": 1, "replies": [ { "id": 3392319, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2026-01-16T17:41:59.493000", "content": "I am guessing you changed folders and names at some point - but this does not work as no scripts folder seems to exist in anything I created so far. But that py can be found in the RNAPro/preprocess/ folder.
\npython3 scripts/gen_ccd_cache.py
This one takes some time - guessing that a few folks are hitting that server right now :) Right now for me on a 1TB Xfinity system I had 1 hour and 10 minutes of download time. Did the Xfinity speed test and my current download is 922 Mbps - so assuming the time is mostly the result of the server rather than my system.
", "votes": 0, "replies": [] }, { "id": 3392320, "author_name": "Theo Viel", "author_url": "", "post_date": "2026-01-16T17:42:10.560000", "content": "\n\nMy assumption is that over the next few days you will learn lots of ways your git hub install directions have issues :)
\n
Yes, I already caught a few installing for inference on Kaggle, but I did not setup for training (yet?) so it can have issues.
\nWe'll fix them asap, thanks for the feedback :)
", "votes": 0, "replies": [] }, { "id": 3392321, "author_name": "Theo Viel", "author_url": "", "post_date": "2026-01-16T17:43:43.590000", "content": "\n\npython3 scripts/gen_ccd_cache.py
\n
This one is known, I have a MR that fixes it
", "votes": 0, "replies": [] }, { "id": 3392345, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2026-01-16T18:33:19.933000", "content": "In case you did not guess :) I am working on use for multiple local machines, so my issues might be a bit different than a kaggle install.
", "votes": 0, "replies": [] } ] } ] }, { "id": 3419651, "author_name": "Cat D", "author_url": "", "post_date": "2026-03-11T09:28:33.200000", "content": "thanks for share
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "3392269": "We’re excited to share an inference workflow for RNAPro for part 2 of the Stanford RNA 3D Folding competition ! RNAPro reflects where the field landed after last year’s breakthroughs. \n\n## What is RNAPro? \n\nRNAPro is a state-of-the-art RNA 3D folding model developed through a synthesis of ideas from the previous Kaggle competition hosts and top-performing teams. It integrates several components that proved valuable in the previous competition and subsequent work: \n- RibonanzaNet2 as an encoder for sequence and pairwise features\n- RNA MSAs to provide rich input data beyond a single sequence\n- 3D structure templates \n- RNA fine-tuned Protenix to take the above inputs and predict 3D structure\n\nThis combination represents some of the best insights from the previous competition !\n\n*⭐ [The code is available on GitHub](https://github.com/NVIDIA-Digital-Bio/RNAPro) ⭐*\n\n.png?generation=1768576990418111&alt=media)\n\n## What's included here? \n\nWe’re sharing two Kaggle notebooks that together create a complete inference and submission workflow: \n- [Template preparation notebook](https://www.kaggle.com/code/theoviel/stanford-rna-3d-folding-pt2-templates)\n- [Inference notebook](https://www.kaggle.com/code/theoviel/stanford-rna-3d-folding-pt2-rnapro-inference)\n\nThese notebooks are intended to provide a strong baseline and starting point for the community, while being completely open. \n\nWe are looking forward to seeing how the community builds on this baseline !\n\n## Further resources: \n\n- [Preprint summarizing past competition insights including RNAPro](https://www.biorxiv.org/content/10.64898/2025.12.30.696949v1.full.pdf)\n- [RNAPro code](https://github.com/NVIDIA-Digital-Bio/RNAPro)\n- Model weights: [Private-Best](https://huggingface.co/nvidia/RNAPro-Private-Best-500M), [Public-Best](https://huggingface.co/nvidia/RNAPro-Public-Best-500M)\n- [Offline inference code](https://www.kaggle.com/code/jaejohn/rnapro-inference-with-tbm) by @jaejohn", "3392306": "My assumption is that over the next few days you will learn lots of ways your git hub install directions have issues :)\n\nI will start with mine - I absolutely hate with a strong passion code that downloads a large zip, extracts it and than deletes it. In my case I have 4 local machines were I play with kaggle. I would like to copy the zip to my NAS and than get things installed on all 4 machines. Sadly - Did not notice the rm's until it was too late. \n\n```python\ncd release_data\nmkdir kaggle; cd kaggle\ncurl -L -o stanford-rna-3d-folding-all-atom-train-data.zip https://www.kaggle.com/api/v1/datasets/download/rhijudas/stanford-rna-3d-folding-all-atom-train-data\nunzip stanford-rna-3d-folding-all-atom-train-data.zip\nrm stanford-rna-3d-folding-all-atom-train-data.zip\n\n# Get MSAs from https://www.kaggle.com/datasets/rhijudas/clone-of-stanford-rna-3d-modeling-competition-data\ncurl -L -o stanford-rna-3d-folding.zip https://www.kaggle.com/api/v1/datasets/download/rhijudas/clone-of-stanford-rna-3d-modeling-competition-data\nmkdir MSA_v2\nunzip stanford-rna-3d-folding.zip \"MSA_v2/*\" -d /MSA_v2\nrm stanford-rna-3d-folding.zip\ncd ..\n```\n", "3419651": "thanks for share" } }