24 skills · sorted by relevance
Configure large language model pretraining with torchtitan across multiple GPUs. Covers ways to split training work, launch jobs on computing clusters, enable Float8 numerical processing, and save or resume training through distributed checkpoints.
Configure large language model pretraining with torchtitan across multiple GPUs. Covers ways to split training work, launch jobs on computing clusters, enable Float8 numerical processing, and save or resume training through distributed checkpoints.
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
The English summary is being prepared. Showing the original description.
Builds LLM pretraining corpora from Common Crawl: extract, filter, deduplicate, decontaminate, mix, synthesize. Use when curating or ablating a pretraining corpus or data pipeline.
The English summary is being prepared. Showing the original description.
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
The English summary is being prepared. Showing the original description.
Set up large language model pretraining with TorchTitan and PyTorch. Configure training across multiple GPUs, combine four ways of splitting the workload, enable Float8 calculations on H100 hardware, and manage checkpoints for saving and resuming training.
Work with Waypoint models to analyze microbiome samples, the communities of microbes in a sample. Convert taxonomic abundance tables, generate numerical sample representations, configure further training for prediction tasks, and evaluate models on the Compass benchmark.
Fornece treinamento distribuído de LLM nativo do PyTorch usando torchtitan com paralelismo 4D (FSDP2, TP, PP, CP). Use para fazer pré-treinamento de Llama 3.1, DeepSeek V3 ou modelos customizados em escala de 8 a 512+ GPUs com Float8, torch.compile e checkpoint distribuído.
The English summary is being prepared. Showing the original description.
Plan further training for AI agent models by diagnosing gaps in knowledge, output consistency, or decision making. Compare continued pretraining, supervised fine-tuning, reinforcement learning, and distillation, then design training data, feedback environments, and evaluation checks.
Build and troubleshoot TorchDrug 0.2.1 workflows for molecular property prediction, molecule generation, protein modeling, and knowledge graph reasoning. Connect datasets, models, training tasks, and evaluation while checking version compatibility and common implementation errors.
Decide whether an LLM needs additional training or whether retrieval and better prompts will suffice. Match available examples, preferences, or verifiable results to a training method, then assess model size and memory requirements before starting.
Use LitGPT to explore language model implementations and train models on custom data. Covers pretrained model loading, LoRA fine-tuning that updates a small set of parameters, pretraining, and preparing trained models for deployment.
Use LitGPT to explore language model implementations and train models on custom data. Covers pretrained model loading, LoRA fine-tuning that updates a small set of parameters, pretraining, and preparing trained models for deployment.
Plan, configure, and chain repo-native Nemotron customization steps into single-step or multi-step pipelines: curation, translation, SFT/PEFT (AutoModel or Megatron-Bridge), pretraining/CPT, RL alignment (DPO/RLVR/GRPO/RLHF), BYOB/MCQ benchmarks, checkpoint conversion, ModelOpt optimization, env profiles, and evaluation of trained checkpoints or existing/hosted endpoints. Use when a request names a Nemotron step or workflow, or asks to clean, translate, train, fine-tune, align, convert, optimize, evaluate, or compose these into a pipeline. Do NOT use for frontend/dashboard/visualization work, generic ML advice, billing/access, or non-Nemotron coding tasks.
The English summary is being prepared. Showing the original description.
Continue KERMT pretraining on a custom SMILES corpus with a grover_base, cmim, or hybrid checkpoint. Use a local checkpoint or optionally download a pinned Hugging Face model bundle using HF_TOKEN if configured. Run containerized training and write model bundles, prepared data, logs, and checkpoints to user-selected host directories.
The English summary is being prepared. Showing the original description.
Convert a grover_base checkpoint (encoder-only or encoder + vocab heads) into a hybrid checkpoint by adding a randomly-initialized cMIM decoder + latent_dist, then continue pretraining on the user's corpus as hybrid (vocab + contrast). Effectively kermt-continue-pretrain with a one-time ckpt-conversion step prepended.
The English summary is being prepared. Showing the original description.
PyTorch-based ML platform for drug discovery: graph molecular representation learning, property prediction (ADMET, activity), retrosynthesis, drug-target interaction (DTI), and pretraining on large molecular datasets. Provides GNN layers (GraphConv, GAT, MPNN), pretrained models, and benchmark datasets.
The English summary is being prepared. Showing the original description.
Operate Baichuan-7B model loading, architecture inspection, C-Eval/MMLU evaluation preflights, and DeepSpeed pretraining setup from self-contained repo guidance.
The English summary is being prepared. Showing the original description.
Use BERT-pytorch for corpus preparation, vocabulary building, and tiny BERT pretraining workflows.
The English summary is being prepared. Showing the original description.
Use when polishing, diagnosing, tailoring, or exporting resumes for LLM, RAG, Agent, Agentic RL, post-training, pretraining, AIGC, search/ranking, multimodal, AI backend, or LLM algorithm internships from raw resume text, a materials folder, and/or a target job description. Audits evidence, maps JD fit, enforces truth boundaries, writes polished and targeted resumes, generates interviewer-style grilling questions, answer cards, evidence-upgrade plans, and optional open-source project recommendations without fabricating experience.
The English summary is being prepared. Showing the original description.
Think and work like an expert Computer Vision Scientist. Use when a task calls for Computer Vision Scientist judgment. Reasons from image formation, projective geometry (pinhole intrinsics and extrinsics, epipolar/PnP/bundle adjustment, similarity-scale ambiguity), and COCO/LVIS AP mechanics through DINOv3/SigLIP 2 foundation baselines, RF-DETR/YOLO26/SAM 3 models, COLMAP 4/GLOMAP and VGGT geometry, pycocotools/TrackEval/BOP evaluation, and CVPR reporting and EU AI Act limits while treating train–test and pretraining leakage, AP evaluation-setting gaming, preprocessing mismatches (EXIF, BGR, aliased resizing), label noise, and camera-convention and scale errors as first-class failure modes.
The English summary is being prepared. Showing the original description.
Audit whether a paper's EVALUATION DESIGN actually measures what it claims and whether its reporting is complete — the validity layer family D (experiment-forensics) cannot reach. Three patterns: train/test leakage means the reported score may not measure generalization (HP-EVAL-LEAKAGE — adopts the Kapoor & Narayanan 8-type / 3-category leakage taxonomy; the illegitimate-proxy / sampling-bias / pretraining-contamination subtypes hand off as needs_external_check, naming but NEVER running Oren-2023 exchangeability / Shi-2023 Min-K% / Golchin-2023 Time-Travel / BIG-bench canary); a load-bearing LLM judge is conflicted (same model/family as a compared system) or unvalidated (no human-agreement, no bias control) (HP-JUDGE-VALIDITY); a declared condition/metric is dropped or switched to favor the method, or 'best' is chosen with no held-out set (HP-SELECTIVE-REPORTING). Verdict-bearing at L0/L1 from the DESCRIBED protocol — NOT repo-gated like experiment-forensics; L2 only CONFIRMS against split/preprocessing/result files. A fresh cross-model reviewer (gpt-5.6-sol xhigh, read-only, fresh thread per pass) PROPOSES findings, each span-anchored to a ledger claim_id; tools/adjudicate_findings.py DECIDES the verdict. Leakage and under-reporting are usually HONEST methodological errors — every finding describes a discrepancy to CHECK, never an accusation. An LLM generating GROUND-TRUTH labels is HP-FAKE-GT (experiment-forensics) — routed there, not here. Emits eval-design-forensics.findings.json; computes NO verdict. Detect-only. Triggers: "eval design audit", "evaluation validity", "train/test leakage", "data leakage", "is the score measuring generalization", "LLM judge bias", "is the judge validated", "selective reporting", "cherry-picked results", "评估设计审计", "评测有效性", "数据泄漏", "训练测试集泄漏", "裁判模型有没有验证", "选择性报告".
The English summary is being prepared. Showing the original description.
Builds a GPT and BPE tokenizer from scratch. Use when implementing autograd, attention, a nanoGPT loop, muP hyperparameter transfer, WSD annealing, loss spikes, or mid-training.
The English summary is being prepared. Showing the original description.
Fine-tune LLMs and VLMs using Unsloth on HF Jobs (Hugging Face on-demand cloud GPUs). Use when users want to fine-tune language models, train VLMs (Vision Language Models), do continued pretraining, domain adaptation, or run UV scripts on HF Jobs. Triggers on requests involving Unsloth training, HF Jobs GPU training, Qwen3-VL fine-tuning, Gemma VLM training, or LoRA fine-tuning on cloud GPUs.
The English summary is being prepared. Showing the original description.