“experiments”的搜索结果

审查REA的CI与测试集，验证是否存在漏报缺陷的判定或测试数据偏差，并依据实际测量结果，在保留必要缺陷检测能力的前提下评估运行成本。

• 调查漏报缺陷的CI判定
• 核对测试数据与当前规范的差异
• 测量缓慢测试集中的无效开销

帮助编写和检查PyTorch模型与训练代码，涵盖数据加载、实验可复现性、训练状态保存与恢复，并提供改善GPU显存占用和运行速度的方法。

• 编写与审查PyTorch训练代码
• 排查实验结果不一致的原因
• 改善加载速度与GPU显存占用

指导通过评价回答的强化学习对大语言模型进行后训练。结合Megatron-LM与SGLang，涵盖数据准备、并行执行及自定义生成流程。

• 开始语言模型的GRPO训练
• 并行运行回答生成与训练
• 训练使用工具的智能体

根据Archify生成的实际图表和代理工作记录，评估准确性、可读性及完成时间，通过前后对比实验改进布局、连线路径和制作流程。

• 比较图表布局和连线路径
• 调查反复修改图表的原因
• 验证图表制作流程的改进

Create ad copy and campaigns for Facebook and Instagram, review advertising spend and delivery results, and manage catalogs and product feeds. The skill also checks advertising policy questions through a dedicated policy tool.

中文简介正在准备中。当前显示英语简介。

• Draft ad copy and plan a campaign
• Review campaign spending and delivery results
• Manage product catalogs and feed refresh schedules

Fit probabilistic models to single-cell measurements for integration across experiments, cell annotation, combined RNA and protein or ATAC analysis, and gene expression comparisons. The skill covers input requirements, training checks, reference mapping, and interpretation of model uncertainty.

中文简介正在准备中。当前显示英语简介。

• Integrating single-cell RNA data across experimental batches
• Assigning cell labels using partial annotations or a reference
• Analyzing RNA with protein or ATAC measurements

Simulate lithium-ion battery charging, discharging, and rest with PyBaMM. Track voltage and capacity, record parameter sources, check how numerical settings affect predictions, and compare simulated curves with measured cycling data.

中文简介正在准备中。当前显示英语简介。

• Predicting voltage changes during charge, discharge, and rest
• Comparing simulation results with measured cycling data
• Checking how numerical settings affect predictions

Organize PyTorch models and data pipelines with PyTorch Lightning, automate training, save checkpoints, and track experiments. Includes templates and guidance for configuring training across multiple GPUs or TPUs.

中文简介正在准备中。当前显示英语简介。

• Restructure PyTorch training code into LightningModules
• Build reusable data preparation and loading pipelines
• Configure checkpoints, early stopping, and experiment logs

Measure fluid velocity from pairs of particle images with OpenPIV. The skill covers image preparation, velocity calculation, correction of unreliable vectors, plotting, and calculation of vorticity and strain rate to examine flow patterns.

中文简介正在准备中。当前显示英语简介。

• Measure velocity fields from experimental image pairs
• Detect and replace unreliable velocity vectors
• Calculate vorticity and strain rate

Prepare experiments for Ginkgo Cloud Lab by choosing a protocol, organizing required inputs, checking configured pricing, and navigating browser ordering. Covers protein expression, RNA synthesis, stability assays, and setup of measurement methods.

中文简介正在准备中。当前显示英语简介。

• Choosing a protein expression and purification protocol
• Preparing inputs for mRNA or circRNA synthesis
• Reviewing stability assays and measurement method setup

Process next-generation sequencing data with deepTools: convert alignment files into coverage tracks, normalize signals, compare samples, and check quality. Create heatmaps and profiles around genes or peak regions for ChIP-seq, RNA-seq, and ATAC-seq experiments.

中文简介正在准备中。当前显示英语简介。

• Converting BAM files to normalized bigWig tracks
• Checking sequencing quality and replicate agreement
• Plotting heatmaps around transcription start sites

Organize experiment photos, audio, and notes into Markdown logs with structured metadata. Assign experiment and sample batch IDs, archive original attachments, and optionally connect records to Feishu group messages or an Obsidian notebook.

中文简介正在准备中。当前显示英语简介。

• Turning experiment photos and voice notes into consistent logs
• Tracking the same sample batch across multiple experiments
• Archiving original files with links from experiment logs

Plan A/B tests to compare page or messaging changes. Helps define hypotheses, choose success metrics, estimate sample sizes, allocate traffic, and interpret results, while also organizing a repeatable process for prioritizing and documenting experiments.

中文简介正在准备中。当前显示英语简介。

• Compare headline or button-copy variants
• Plan sample sizes and test duration around available traffic
• Analyze results alongside metrics that must not worsen

Design machine learning workflows from data preparation through training, validation, deployment, and monitoring. Use templates and guidance to connect stages, track versions and experiments, handle failures, and plan gradual releases with rollback options.

中文简介正在准备中。当前显示英语简介。

• Designing an end-to-end machine learning pipeline
• Building reproducible training workflows
• Comparing models before production deployment

Convert a research paper into structured files covering its problem, claims, concepts, experiments, and figure or table evidence. Validate the generated directory, ingest it into OpenViking, and check that the target resource was created.

中文简介正在准备中。当前显示英语简介。

• Import a paper PDF into OpenViking
• Organize a paper's claims and experimental evidence
• Preserve figures and tables as evidence files

Build and run evaluations for AI applications using Phoenix. Start with observed failures, create checks in code or use a language model as a judge, run experiments, and validate automated judgments against human labels.

中文简介正在准备中。当前显示英语简介。

• Turning observed failures into evaluation criteria
• Building code-based or language-model evaluators
• Evaluating retrieval and answer faithfulness in RAG

Create Markdown documents for time-limited technical investigations before implementation. Define the decision to resolve, research tasks, success criteria, and deadlines, then record evidence, recommendations, and follow-up actions in a consistent structure.

中文简介正在准备中。当前显示英语简介。

• Investigating API capabilities before integration
• Comparing architecture options before implementation
• Planning experiments to check performance assumptions

Create and manage Arize datasets for evaluation and experiments using the ax CLI. Import records from files, append examples, inspect dataset versions, and export a chosen version for offline analysis, with checks for field consistency and export completeness.

中文简介正在准备中。当前显示英语简介。

• Creating evaluation datasets from files
• Adding examples to an existing test set
• Inspecting dataset metadata and versions

Create and run evaluations in Arize using an LLM as the judge. Define scoring criteria, connect them to actual data fields, evaluate recorded operations or experiment results, and set up continuous scoring for new data.

中文简介正在准备中。当前显示英语简介。

• Check answers for unsupported claims
• Score experiment results for correctness
• Continuously evaluate new production records

Improve LLM prompts using Arize production traces, evaluation results, and human feedback. Extract current instructions, identify recurring failures, revise the prompt, and compare experiments to measure changes in output quality and catch regressions.

中文简介正在准备中。当前显示英语简介。

• Improve a prompt using failed production traces
• Fix responses that use the wrong output format
• Review unsupported answers in retrieval-based responses

Run repeated change-and-evaluation cycles for one predefined improvement mission within an explicit runtime limit. Save each evaluation as JSON and each decision in readable Markdown logs, with support for periodic reruns through Claude Code native cron.

中文简介正在准备中。当前显示英语简介。

• Repeating improvements under a fixed evaluation method
• Keeping evaluation results and decision logs
• Running experiments within a time limit

Review decisions through Nassim Nicholas Taleb’s thinking about rare, severe losses and who bears their consequences. Identify hidden dependencies, compare potential gains with losses, and examine whether a strategy benefits from disruption or becomes more vulnerable.

中文简介正在准备中。当前显示英语简介。

• Examining rare events that could cause irreversible losses
• Checking whether advisers bear the consequences of their advice
• Finding dependencies that make a business fragile

Examine product costs, technical plans, and industry assumptions through a framework drawn from Elon Musk’s public record. Compare prices with material costs, question unnecessary requirements, and explore whether to build components internally or outsource them.

中文简介正在准备中。当前显示英语简介。

• Reviewing product costs against material costs
• Questioning requirements before automating a process
• Comparing internal production with outsourcing

Set up model training with torchforge for reinforcement learning, where rewards guide improvement. Covers training configurations, custom reward and loss functions, and distributed execution with Monarch and TorchTitan, keeping algorithm experiments separate from infrastructure concerns.

中文简介正在准备中。当前显示英语简介。

• Train mathematical reasoning models with GRPO
• Experiment with custom reward and loss functions
• Scale training across multiple GPUs or nodes