400 件 ・ 関連度順
概要と使いどころ
Quantização de pesos consciente da ativação para compressão de LLM de 4-bit com aceleração 3x e perda mínima de acurácia. Use ao implantar modelos grandes (7B-70B) em memória GPU limitada, quando precisa de inferência mais rápida que GPTQ com melhor preservação de acurácia, ou para modelos com instruções ajustadas e multimodais. Vencedor do Prêmio Melhor Artigo MLSys 2024.
日本語の概要は準備中です。原文の説明を表示しています。
Otimiza atenção em transformers com Flash Attention para ganho de 2-4x em velocidade e redução de 10-20x em memória. Use ao treinar/executar transformers com sequências longas (>512 tokens), ao encontrar problemas de memória GPU com atenção, ou quando precisa de inferência mais rápida. Suporta SDPA nativo do PyTorch, biblioteca flash-attn, H100 FP8 e sliding window attention.
日本語の概要は準備中です。原文の説明を表示しています。
Quantiza LLMs para 8-bit ou 4-bit com redução de memória de 50-75% e perda mínima de acurácia. Use quando a memória GPU é limitada, precisa ajustar modelos maiores ou quer inferência mais rápida. Suporta formatos INT8, NF4, FP4, treinamento QLoRA e otimizadores 8-bit. Funciona com HuggingFace Transformers.
日本語の概要は準備中です。原文の説明を表示しています。
Framework para construir aplicações com LLM usando agentes, chains e RAG. Suporta múltiplos provedores (OpenAI, Anthropic, Google), 500+ integrações, agentes ReAct, tool calling, gerenciamento de memória e recuperação de vector stores. Use para construir chatbots, sistemas de perguntas e respostas, agentes autônomos ou aplicações RAG. Ideal para prototipagem rápida e deployments em produção.
日本語の概要は準備中です。原文の説明を表示しています。
Mescle múltiplos modelos ajustados usando mergekit para combinar capacidades sem retreinar. Use ao criar modelos especializados misturando expertise específica de domínio (math + coding + chat), melhorando performance além de modelos únicos, ou experimentando rapidamente variantes de modelos. Cobre SLERP, TIES-Merging, DARE, Task Arithmetic, mesclagem linear e estratégias de deploy em produção.
日本語の概要は準備中です。原文の説明を表示しています。
Reduza o tamanho de LLMs e acelere a inferência usando técnicas de pruning como Wanda e SparseGPT. Use para comprimir modelos sem retreinamento, alcançando 50% de esparsidade com perda mínima de acurácia, ou ativando inferência mais rápida em aceleradores de hardware. Cobre pruning não estruturado, pruning estruturado, esparsidade N:M, pruning por magnitude e métodos one-shot.
日本語の概要は準備中です。原文の説明を表示しています。
Framework de orquestração multi-agent para colaboração autônoma de IA. Use ao construir equipes de agentes especializados trabalhando juntos em tarefas complexas, quando você precisa de colaboração de agentes baseada em papéis com memória, ou para workflows em produção exigindo execução sequencial/hierárquica. Construído sem dependências de LangChain para execução enxuta e rápida.
日本語の概要は準備中です。原文の説明を表示しています。
Creates structured professional research summaries from public profile information. Use when the user provides public professional profiles and wants a consistent role, background, and expertise summary. Must avoid inferring sensitive personal attributes.
日本語の概要は準備中です。原文の説明を表示しています。
Identifies likely duplicate profiles or companies using available record evidence. Use when the user provides a list and wants duplicate groups, match confidence, and merge recommendations. Describe records as likely duplicates unless there is an exact deterministic match.
日本語の概要は準備中です。原文の説明を表示しています。
Groups companies by industry, business model, size, maturity, market, or strategic fit. Use when the user provides company records and wants consistent segment assignments with definitions and confidence.
日本語の概要は準備中です。原文の説明を表示しています。
Classifies professional profiles into buyer, influencer, user, decision-maker, founder, recruiter, or other personas. Use when the user provides profiles and wants persona labels with evidence and confidence. Must not infer sensitive personal attributes.
日本語の概要は準備中です。原文の説明を表示しています。
Turns company records into structured company research summaries. Use when the user provides company names or records and wants a consistent overview, market position, and relevance assessment.
日本語の概要は準備中です。原文の説明を表示しています。
Identifies missing information required to complete qualification, segmentation, or outreach preparation. Use when the user provides records and a goal and wants prioritized research gaps with sources.
日本語の概要は準備中です。原文の説明を表示しています。
Converts a job description into a structured candidate-sourcing workflow. Use when the user provides a role and wants candidate criteria, search terms, source recommendations, and a screening process. Must avoid sourcing on sensitive personal attributes.
日本語の概要は準備中です。原文の説明を表示しています。
Qualifies and scores B2B prospects against a supplied ideal customer profile. Use when the user provides a lead list, company records, prospect data, or qualification criteria and wants prioritization, fit analysis, or disqualification reasons.
日本語の概要は準備中です。原文の説明を表示しています。
Compares a group of companies by products, positioning, market focus, and differentiators. Use when the user provides several company records and wants a structured competitive comparison.
日本語の概要は準備中です。原文の説明を表示しています。
Creates a repeatable process for researching and comparing companies. Use when the user wants a structured, reusable company-research workflow with objectives, sources, fields, and validation.
日本語の概要は準備中です。原文の説明を表示しています。
Evaluates potential channel, integration, affiliate, or strategic partners. Use when the user provides partner profiles plus partnership goals and wants fit scoring, partnership type, and outreach priority.
日本語の概要は準備中です。原文の説明を表示しています。
Normalizes company names, industries, domains, locations, and descriptions. Use when the user provides company records and wants consistent canonical fields with confidence and unresolved items flagged.
日本語の概要は準備中です。原文の説明を表示しています。
Normalizes and improves the consistency of prospect-list data. Use when the user provides a messy prospect or lead export and wants standardized records, flagged issues, and a cleanup summary.
日本語の概要は準備中です。原文の説明を表示しています。
Turns an ICP or research goal into a plan for finding and scraping suitable directories. Use when the user has a target audience but is unsure where to find them, and wants source categories, queries, and an extraction workflow.
日本語の概要は準備中です。原文の説明を表示しています。
Converts inconsistent job titles into standardized role, department, and seniority labels. Use when the user provides records with free-text titles and wants consistent role family, department, and seniority fields.
日本語の概要は準備中です。原文の説明を表示しています。
Assesses the quality, composition, and usefulness of a lead list. Use when the user provides a lead list and wants completeness, distribution, quality issues, and recommended next actions.
日本語の概要は準備中です。原文の説明を表示しています。
Divides a lead list into useful commercial segments. Use when the user provides a combined lead list and wants segments with criteria, sizes, recommended treatment, and priority.
日本語の概要は準備中です。原文の説明を表示しています。