机器翻译模型评估与BLEU计算
使用Hugging Face Transformers和sacrebleu库,对预训练的机器翻译模型进行评估。支持从制表符分隔的CSV文件读取源文本和参考翻译,计算BLEU分数,并提供Seq2SeqTrainer的compute_metrics函数实现。
Browse reusable Agent Skills, each with a clear purpose and practical guidance.
使用Hugging Face Transformers和sacrebleu库,对预训练的机器翻译模型进行评估。支持从制表符分隔的CSV文件读取源文本和参考翻译,计算BLEU分数,并提供Seq2SeqTrainer的compute_metrics函数实现。
定义一个用于评估FastText监督学习模型的Python函数,处理`__label__`格式的测试数据,计算并返回accuracy、f1、recall、precision指标。
使用Python手动计算卡方统计量和临界值,检验观测频数数据是否符合指定均值的泊松分布,参考特定代码风格实现。
Reviews specific chapters of a product accountability report against a provided SDLC structure, evaluating adherence, logical flow, and technical quality.
Performs comprehensive Solidity security audits (Markdown table) or validates specific proposed vulnerabilities. Adheres to strict table schemas, severity ratings, and evidence requirements.
Validate Draw Things LoRA training end to end with draw-things-cli, including tiny-dataset training, loss and scaler checks, checkpoint sanity, and base-versus-LoRA generation comparison.
根据用户提供的测试用例列表生成测试结果描述,或将功能描述转化为包含特定字段的测试用例。支持将测试结果总结为单一段落。
提供镜像量子电路基准测试的步骤说明,并使用Python和Qiskit库编写代码实现电路构建、模拟运行及结果评估。
Review citation integrity across data/citations.yml and @citation tags using bactopia-citations --validate. Detects orphan citation keys (defined in the yml but never referenced) and workflow @citation keys that don't resolve to a yml entry. Use this skill whenever the user asks to review citations, check citation integrity, audit citations.yml, find orphan citations, clean up unused citations, validate workflow @citation tags, or verify that every tool cited in a workflow has a matching entry in the citations file.
Review GroovyDoc accuracy across modules and subworkflows using bactopia-lint. Checks @output/@input field matching, @modules/@subworkflows lists, citation keys, tag ordering, and formatting. Use when asked to review GroovyDoc, check documentation accuracy, validate module/subworkflow docs, or audit GroovyDoc.