--- title: "PiSSA vs P-tuning-v2" type: "comparison" canonical_url: "https://www.graphcanon.com/compare/mulabpku-pissa-vs-thudm-p-tuning-v2" tools: ["mulabpku-pissa", "thudm-p-tuning-v2"] --- # PiSSA vs P-tuning-v2 *GraphCanon updated Aug 24, 2026* ## Verdict Pick PiSSA if piSSA targets efficient fine-tuning of large language models via principal singular values and vectors; pick P-tuning-v2 if p-tuning-v2 is an optimized deep prompt tuning strategy that can be used for tasks like natural language processing and parameter-efficient learning where few parameters need to be adjusted compared to full fine-tuning. [PiSSA](https://proceedings.neurips.cc/paper_files/paper/2024/file/db36f4d603cc9e3a2a5e10b93e6428f2-Paper-Conference.pdf) reports 430 GitHub stars, 23 forks, and 16 open issues, last pushed Jun 30, 2025. [P-tuning-v2](https://github.com/THUDM/P-tuning-v2) has 2.1k stars, 213 forks, and 35 open issues, last pushed Nov 16, 2023. Figures are from public GitHub metadata via [PiSSA's repository](https://github.com/MuLabPKU/PiSSA) and [P-tuning-v2's repository](https://github.com/THUDM/P-tuning-v2). | | [PiSSA](/tools/mulabpku-pissa.md) | [P-tuning-v2](/tools/thudm-p-tuning-v2.md) | | --- | --- | --- | | Tagline | Principal Singular Values and Singular Vectors Adaptation of Large Language Models | Optimized deep prompt tuning strategy comparable to fine-tuning across scales and tasks | | Stars | 430 | 2,077 | | Forks | 23 | 213 | | Open issues | 16 | 35 | | Language | Jupyter Notebook | Python | | Adopt for | PiSSA targets efficient fine-tuning of large language models via principal singular values and vectors. | P-tuning-v2 is an optimized deep prompt tuning strategy that can be used for tasks like natural language processing and parameter-efficient learning where few parameters need to be adjusted compared to full fine-tuning. | | Persona | - | - | | Runtime | - | - | | License | - | P-tuning-v2 is provided under the Apache-2.0 license which permits free use, modification and distribution as long as copyright and license notice are preserved. | | Categories | LLM Frameworks, Model Training | Model Training | ## Trust and health _Sourced signals - not a safety guarantee. No winner column._ | | [PiSSA](/tools/mulabpku-pissa.md) | [P-tuning-v2](/tools/thudm-p-tuning-v2.md) | | --- | --- | --- | | Days since push | 420d | 990d | | Open issues (now) | 16 | 35 | | Stars delta | +1 (30d) | Unknown | | Open issues delta | 0 (30d) | Unknown | | Full report | [trust report](/tools/mulabpku-pissa/trust.md) | [trust report](/tools/thudm-p-tuning-v2/trust.md) | ## Decision facts: PiSSA - **Adopt for:** PiSSA targets efficient fine-tuning of large language models via principal singular values and vectors. ## Decision facts: P-tuning-v2 - **Requirements:** Min 8 GB RAM; Experiments were conducted on NVIDIA DGX-A100; however, RTX 3090 or RTX 2080 Ti GPUs can suffice given certain conditions.; CUDA version should be at least 11.1 and specific versions of pytorch (1.7.1), torchvision (0.8.2) and torchaudio (0.7.2).; A conda environment setup with Python 3.8.5 is recommended for optimal performance. - **Adopt for:** P-tuning-v2 is an optimized deep prompt tuning strategy that can be used for tasks like natural language processing and parameter-efficient learning where few parameters need to be adjusted compared to full fine-tuning. - **License detail:** P-tuning-v2 is provided under the Apache-2.0 license which permits free use, modification and distribution as long as copyright and license notice are preserved. ## Choose when ### Choose PiSSA if… - PiSSA is primarily Jupyter Notebook; P-tuning-v2 is Python. - Tags unique to PiSSA: fine-tuning, peft, quantization. - Also covers LLM Frameworks. - You need to fine-tune a large language model efficiently with limited resources. ### Choose P-tuning-v2 if… - P-tuning-v2 is primarily Python; PiSSA is Jupyter Notebook. - Requirements: Min 8 GB RAM; Experiments were conducted on NVIDIA DGX-A100; however, RTX 3090 or RTX 2080 Ti GPUs can suffice given certain conditions.; CUDA version should be at least 11.1 and specific versions of pytorch (1.7.1), torchvision (0.8.2) and torchaudio (0.7.2).; A conda environment setup with Python 3.8.5 is recommended for optimal performance.. - Tags unique to P-tuning-v2: natural-language-processing, p-tuning, parameter-efficient-learning, pretrained-language-model. - For scenarios requiring efficient use of resources, as P-tuning v2 requires fewer parameters than traditional fine-tuning methods. ## When NOT to use PiSSA - Insufficient flexibility in model adaptation is acceptable, prefer broader customization options. - Full fine-tuning of the entire model rather than just key components via peft. ## When NOT to use P-tuning-v2 - Avoid using P-tuning v2 when you have large-scale datasets and model architectures, as it might not be as effective compared to full fine-tuning methods. - If your project requires a high level of customization beyond prompt tuning or continuous prompts do not align with the problem complexity. ## Common questions ### What is the difference between PiSSA and P-tuning-v2? PiSSA: Principal Singular Values and Singular Vectors Adaptation of Large Language Models. P-tuning-v2: Optimized deep prompt tuning strategy comparable to fine-tuning across scales and tasks. See the comparison table for live GitHub stats and shared categories. ### When should I choose PiSSA over P-tuning-v2? Choose PiSSA over P-tuning-v2 when PiSSA is primarily Jupyter Notebook; P-tuning-v2 is Python; Tags unique to PiSSA: fine-tuning, peft, quantization; Also covers LLM Frameworks; You need to fine-tune a large language model efficiently with limited resources. ### When should I choose P-tuning-v2 over PiSSA? Choose P-tuning-v2 over PiSSA when P-tuning-v2 is primarily Python; PiSSA is Jupyter Notebook; Requirements: Min 8 GB RAM; Experiments were conducted on NVIDIA DGX-A100; however, RTX 3090 or RTX 2080 Ti GPUs can suffice given certain conditions.; CUDA version should be at least 11.1 and specific versions of pytorch (1.7.1), torchvision (0.8.2) and torchaudio (0.7.2).; A conda environment setup with Python 3.8.5 is recommended for optimal performance.; Tags unique to P-tuning-v2: natural-language-processing, p-tuning, parameter-efficient-learning, pretrained-language-model; For scenarios requiring efficient use of resources, as P-tuning v2 requires fewer parameters than traditional fine-tuning methods. ### When should I avoid PiSSA? Insufficient flexibility in model adaptation is acceptable, prefer broader customization options. Full fine-tuning of the entire model rather than just key components via peft. ### When should I avoid P-tuning-v2? Avoid using P-tuning v2 when you have large-scale datasets and model architectures, as it might not be as effective compared to full fine-tuning methods. If your project requires a high level of customization beyond prompt tuning or continuous prompts do not align with the problem complexity. ### Is PiSSA or P-tuning-v2 more popular on GitHub? P-tuning-v2 has more GitHub stars (2,077 vs 430). Stars measure visibility, not whether either tool fits your constraints. ### Are PiSSA and P-tuning-v2 open source? Yes - both are open-source projects on GitHub. ### Where can I find alternatives to PiSSA or P-tuning-v2? GraphCanon lists graph-backed alternatives at [PiSSA alternatives](/tools/mulabpku-pissa/alternatives) and [P-tuning-v2 alternatives](/tools/thudm-p-tuning-v2/alternatives) ([PiSSA markdown twin](/tools/mulabpku-pissa/alternatives.md), [P-tuning-v2 markdown twin](/tools/thudm-p-tuning-v2/alternatives.md)), ranked by typed relationship edges rather than popularity votes. ### Is there a machine-readable version of this comparison? Yes. The markdown twin at [this comparison](/compare/mulabpku-pissa-vs-thudm-p-tuning-v2.md) mirrors this page for agents and LLM crawlers, with the same stats table and FAQ answers. ### Which is better maintained, PiSSA or P-tuning-v2? PiSSA: Dormant. P-tuning-v2: Dormant. Compare maintenance labels, days since push, and release cadence in the trust section below - stars alone do not measure maintenance. ### Where are the full trust reports for PiSSA and P-tuning-v2? GraphCanon publishes per-repo trust reports with dated maintenance, provenance, and scan summaries: [PiSSA trust report](/tools/mulabpku-pissa/trust); [P-tuning-v2 trust report](/tools/thudm-p-tuning-v2/trust). --- **Machine-readable endpoints** - JSON: [`/api/graphcanon/graph?tool=mulabpku-pissa`](/api/graphcanon/graph?tool=mulabpku-pissa) - LLM index: [/llms.txt](/llms.txt) - Full corpus: [/llms-full.txt](/llms-full.txt) _GraphCanon - The knowledge graph for AI development. https://www.graphcanon.com/_