llm_counts

by harleyszhang · Agent Tool · ★ 114

About llm_counts

llmprofiler llm theoretical performance analysis tools and support params, flops, memory and latency analysis. 主要功能 支持 qwen2.5、qwen3 dense 系列模型。 支持张量并行推理模式。 支持 、、 等硬件以及主流 decoder-only 的自回归模型，可自行在配置文件中增加。 支持分析性能瓶颈，不同 是 还是 ，以及 的性能瓶颈。 支持输出每层和整个模型的参数量、计算量，内存和 。 推理时支持预填充和解码阶段分别计算内存和 latency、以及理论支持的最大 等等。 支持设置计算效率、内存读取效率（不同推理框架可能不一样，这个设置好后，可推测输出实际值）。 推理性能理论分析结果的格式化输出。 如何使用 使用方法，直接调用 文件中函数 函数并输入相关参数即可。 python def llmprofile(modelname="llama-13b", gpuname: str = "v100-sxm-32gb", bytesperparam: int = BYTESFP16, bs: int = 1, seqlen: int = 522, generatelen=1526, dszero: int = 0, dpsize: int = 1, tpsize:...

gpu-performancellamallmllm-inferenceprofilerpython3transformer

Quick Facts

Stars114
Forks10
LanguagePython
CategoryAgent Tool
Quality Score57.9863397002389/100
Open Issues1
Last Updated2025-07-11
Created2023-07-26
Platformspython
Est. Tokens~505k

Compatible Skills

These tools work well together with llm_counts for enhanced workflows:

  • vllm-cli — semantic(0.23)+complementary+rare_topics+same_lang+similar_pop+shared_platform (58%)
  • Noema-Declarative-AI — semantic(0.16)+complementary+rare_topics+same_lang+similar_pop+shared_platform (55%)
  • ai-agents-reality-check — semantic(0.19)+complementary+same_lang+similar_pop+shared_platform (52%)
  • bzm-mcp — semantic(0.18)+complementary+same_lang+similar_pop+shared_platform (51%)
  • data-analysis-llm-agent — semantic(0.18)+complementary+same_lang+similar_pop+shared_platform (51%)

More Agent Tool Tools

Explore other popular agent tool tools:

View all Agent Tool tools →

Popular Python Agent Tools

Frequently Asked Questions

What is llm_counts?

llm_counts is llm theoretical performance analysis tools and support params, flops, memory and latency analysis.. It is categorized as a Agent Tool with 114 GitHub stars.

What programming language is llm_counts written in?

llm_counts is primarily written in Python. It covers topics such as gpu-performance, llama, llm.

How do I install or use llm_counts?

You can find installation instructions and usage details in the llm_counts GitHub repository at github.com/harleyszhang/llm_counts. The project has 114 stars and 10 forks, indicating an active community.

View on GitHub → Browse Agent Tool tools