SKU/Artículo: AMZ-B0FVMQ3BX8
HANDS-ON LLM FINE-TUNING WITH LORA AND QLORA: Step-by-step code examples for training custom models with Hugging Face, PEFT, and bitsandbytes on real datasets
Format:
Paperback
Detalles del producto
Disponibilidad:
En stock
En stock
Peso con empaque:
0.52 kg
0.52 kg
Devolución:
Sí
Sí
Condición
Nuevo
Nuevo
Producto de:
Amazon
Amazon
Viaja desde
USA
USA
Sobre este producto
- Train useful LLMs on real hardware and real datasets with a clear, repeatable recipe from setup to deployment.Many teams struggle to move from toy notebooks to stable training runs that meet cost, speed, and quality targets. Version mismatches, brittle data formatting, and unclear evaluation make results hard to trust.This book gives you a grounded path that works end to end. You will pin a reliable stack, prepare high quality chat data, fine tune with LoRA and QLoRA, scale on real GPUs, evaluate with rigorous tools, and ship models you can serve and maintain.Set up a dependable stack, pin Python PyTorch CUDA and enable fast attention with sdpa or flashattention 2Apply tokenizer apply chat template correctly for your base model and mask labels for supervised fine tuningClean datasets with normalization deduplication quality filters and pii redaction that hold up in productionIncrease throughput with sequence packing and length management without leaking loss across boundariesTrain adapters with peft configure ranks targets dropout and validate deltas with quick eval and error triageRun qlora with bitsandbytes nf4 and double quantization set correct dtypes gradients and k bit training hooksScale on multi gpu with accelerate and fsdp plus axolotl yaml workflows for consistent experimentsUse unsloth recipes for fast sft on a single gpu with tight memory budgetsChoose advanced adapters dora adalora vera and target_parameters for moe and fused qkv blocksMerge and compose adapters with merge and unload ties merging and adapterfusionOptimize preferences with trl dpo plus orpo and simpo alternatives with stable hyperparametersEvaluate with lm evaluation harness and lighteval track runs in weights and biases or mlflowRun instruction eval with arena hard auto and alpacaeval 2 length controlled settingsVerify long context with ntk aware scaling yarn and reliability testsPackage and publish adapters or merged weights to the hub with solid model cards and versioningServe multiple adapters with vllm routing and hot swap or deploy with tgi and export gguf for llama cppOperate with cost planning batch calculators governance licensing and a release checklist including kernels templates and fsdpThis is a code heavy guide with working snippets that you can paste into real projects, from training scripts to serving commands, so you make progress on day one.Get the guide that turns fine tuning into a dependable process, grab your copy today.
AR$104.244
49% OFF
AR$53.461
IMPORT EASILY
By purchasing this product you can deduct VAT with your RUT number
AR$104.244
49% OFF
AR$53.461
Pagá fácil y rápido con Mercado Pago o MODO
Llega en 8 a 12 días hábiles
con envío
Tienes garantía de entrega
Este producto viaja
de USA ![]()
a tus manos en
![]()
a tus manos en
Compra protegida
Disfruta de una experiencia de compra segura y confiable
Tarjetas de Crédito y Débito
Visa
Mastercard
American Express
Pagos a través de PayPal
Compra procesada en dólares con dinero en cuenta o tarjetas internacionales.
Paypal
Pagos a través de MODO
Compra procesada en pesos desde la App de tu banco.
PayPal
