Daniel Avila

Fine-tune large language models with LoRA and QLoRA while limiting GPU memory use. Guides training settings and adapter management, using small sets of task-specific weights that can be saved, switched, or merged into a base model.

中文简介正在准备中。当前显示英语简介。

• Fine-tuning an LLM with limited GPU memory
• Choosing LoRA settings and training layers
• Switching task-specific adapters on one base model

Deploy language models with vLLM through an OpenAI-compatible API or offline batch processing. Tune response speed, throughput, and GPU memory use with quantization, which reduces numerical precision, and execution across multiple GPUs.

中文简介正在准备中。当前显示英语简介。

• Deploy an OpenAI-compatible language model API
• Generate text from many prompts in a batch
• Serve quantized models with limited GPU memory

Run machine learning workloads on Modal's cloud GPUs without managing servers. Guides Python configuration for training, inference, batch processing, and API deployment, including automatic scaling, persistent storage, scheduled jobs, and troubleshooting.

中文简介正在准备中。当前显示英语简介。

• Running GPU inference without managing servers
• Deploying a model as an automatically scaling API
• Processing large batches in parallel

Design LLM applications using patterns for answers grounded in documents, agents that use tools, and prompt testing. Implementation examples also cover monitoring response quality, latency, costs, and failures during operation.

中文简介正在准备中。当前显示英语简介。

• Building an assistant that answers from internal documents
• Choosing an architecture for an agent that uses tools
• Comparing and versioning prompts

Guides experiments that inspect and change internal model values using nnsight. Covers local PyTorch models and remote experiments through NDIF, including replacing intermediate values to test how particular layers affect a model’s predictions.

中文简介正在准备中。当前显示英语简介。

• Inspecting layer outputs and attention patterns
• Replacing internal values to compare predictions
• Running large-model experiments through NDIF

Guides the use of RWKV, a language model architecture that carries information forward in a compact state. Covers streaming text generation, long document processing, fine-tuning, and memory comparisons with Transformers.

中文简介正在准备中。当前显示英语简介。

• Build streaming text generation
• Process long documents in chunks
• Compare memory needs with Transformers

Build image-based conversations with LLaVA, an AI model that combines image understanding with language. Covers setup, model selection, image descriptions, follow-up questions, and memory-saving options for running models on a GPU.

中文简介正在准备中。当前显示英语简介。

• Building a chatbot that answers questions about images
• Generating image descriptions and captions
• Exploring a scene through follow-up questions

Use LangSmith to record LLM application inputs, outputs, and timing, investigate failures, evaluate answers against test datasets, and monitor production usage. It also guides automated quality checks and analysis of token usage and costs.

中文简介正在准备中。当前显示英语简介。

• Investigating failures in LLM applications
• Evaluating answers against test datasets
• Monitoring production latency and costs

Run Gemini CLI to review code, assess technical plans, and analyze large collections of files. Choose a model and execution settings for the task, then capture findings about security, maintainability, architecture, or technical debt.

中文简介正在准备中。当前显示英语简介。

• Reviewing security and quality across multiple files
• Checking architecture plans for missing components
• Analyzing the structure of a large codebase

Learn how GPT language models work by preparing text, training a small model, and generating samples. Covers CPU experiments, custom datasets, further training of pretrained GPT-2 models, and reproducing GPT-2 with multiple GPUs.

中文简介正在准备中。当前显示英语简介。

• Train a small model on Shakespeare using a CPU
• Prepare custom text for model training
• Further train a pretrained GPT-2 model

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.

中文简介正在准备中。当前显示英语简介。

• Studying language model architecture through readable code
• Fine-tuning a model on custom instruction data
• Configuring LoRA training for a single GPU

Develop Garry's Mod addons with guidance on Lua scripts, file organization, assets, and packaging. Covers server and client execution contexts, API lookup, and debugging, with in-game testing included in the workflow.

中文简介正在准备中。当前显示英语简介。

• Planning a new Garry's Mod addon
• Writing Lua scripts for server and client contexts
• Organizing addon assets and directories

Combine capabilities from multiple fine-tuned LLMs without retraining using mergekit. Choose a merging method, configure how models are blended, and evaluate the resulting model on tasks such as math, code generation, and conversation.

中文简介正在准备中。当前显示英语简介。

• Combining math, coding, and conversation capabilities
• Blending two models at different ratios
• Comparing variants made with different merging methods

Extend the amount of text a language model can process at once. Compares RoPE, YaRN, ALiBi, and position interpolation, with implementation examples, additional training guidance, and memory optimization approaches for long document processing.

中文简介正在准备中。当前显示英语简介。

• Extend an existing model to handle longer inputs
• Compare methods for representing token positions
• Plan additional training with long documents

Adapt machine learning training to run across multiple GPUs or machines with Ray Train. Covers changes to existing code, comparing training settings with Ray Tune, and saving checkpoints so training can resume after failures.

中文简介正在准备中。当前显示英语简介。

• Adapting existing PyTorch code for multiple GPUs
• Training HuggingFace models across multiple machines
• Comparing learning rates and batch sizes with Ray Tune

Organize PyTorch training code with LightningModule and Trainer. Follow examples for validation, logging, model saving, and distributed training across multiple GPUs, with callbacks that add monitoring or stop training when progress stalls.

中文简介正在准备中。当前显示英语简介。

• Converting an existing PyTorch training loop
• Adding validation, logs, and model checkpoints
• Configuring training across multiple GPUs

Helps configure and troubleshoot DeepSpeed for training large models across multiple devices. Covers memory optimization, parallel training, and DeepNVMe storage transfers, with examples for tuning data movement and configuring models that divide work among specialized components.

中文简介正在准备中。当前显示英语简介。

• Configuring distributed training for large models
• Tuning tensor transfers to NVMe storage
• Setting up models with specialized expert components

Adapt PyTorch training scripts for multiple GPUs or machines with Accelerate. This skill guides code changes, configuration, and launch commands, including mixed precision, which combines numerical formats, and DeepSpeed or FSDP integration for distributed training.

中文简介正在准备中。当前显示英语简介。

• Moving single-GPU training to multiple GPUs
• Configuring mixed precision training
• Setting up DeepSpeed or FSDP

Split independent debugging or repair tasks among multiple AI agents. Define a focused scope, constraints, and expected report for each agent, then review their changes, check for conflicts, and test the combined result.

中文简介正在准备中。当前显示英语简介。

• Investigating unrelated failures across test files
• Assigning separate subsystems to different agents
• Writing focused tasks with clear constraints

Run Gemini research tasks that plan searches, read sources, and combine findings into cited reports. Specify a report structure, monitor progress, and continue earlier research with follow-up questions for market, technical, historical, or comparative studies.

中文简介正在准备中。当前显示英语简介。

• Research a market and its competitors
• Review technical literature
• Compare products or technologies

Adapt an AI agent’s approach to brainstorming, implementation, debugging, review, teaching, or release preparation. Defines how each mode should ask questions, prioritize work, and present results, with automatic mode selection and explicit switching examples.

中文简介正在准备中。当前显示英语简介。

• Explore several ideas before implementation
• Investigate a bug through hypotheses and checks
• Organize review findings by severity

Build LlamaIndex applications that retrieve relevant material before generating answers, a method called RAG. Connect documents, websites, and databases to support document questions, conversational search, structured summaries, and checks of answer relevance and source consistency.

中文简介正在准备中。当前显示英语简介。

• Building question answering over internal documents
• Creating a chatbot that remembers the conversation
• Searching information across files, websites, and databases

Gives AI agents persistent project memory for architecture, implementation patterns, and decisions. Supports saving, searching, and retrieving knowledge, automatically syncing it with project documentation, and viewing memory usage through a dashboard.

中文简介正在准备中。当前显示英语简介。

• Save architecture decisions for later reference
• Search saved implementation patterns by type or tag
• Retrieve a specific design record by key

Manage multiple local command-line agents in separate tmux sessions. Start and stop agents, assign tasks, and follow their logs, with support for scheduling recurring work through cron. Requires tmux and python3.

中文简介正在准备中。当前显示英语简介。

• Run multiple local command-line agents in parallel
• Assign tasks to individual agents
• Follow agent logs and monitor output