From b0a046f81f5d4a42414a8d0025bab8fd814219d7 Mon Sep 17 00:00:00 2001 From: Jeremy Nelson Date: Sun, 22 Sep 2024 12:45:15 -0700 Subject: [PATCH] feat: Final copyedit of Training topic --- exploring-llms/training-llms.html | 20 ++++++++++---------- exploring-llms/training-llms.md | 22 ++++++++++------------ 2 files changed, 20 insertions(+), 22 deletions(-) diff --git a/exploring-llms/training-llms.html b/exploring-llms/training-llms.html index 7a24ee1..085eabd 100644 --- a/exploring-llms/training-llms.html +++ b/exploring-llms/training-llms.html @@ -27,21 +27,21 @@

23 September 2024

Training Large Language Models (LLMs)

One method for customizing and providing better context when using LLMs is to train the -LLM on your data. There are a number methods for fine-tuning LLMs including low-rank adaptation or LoRA1, -that allow you to fine-tune a small number of parameters of a LLM many parameters without a -large number of GPUs used for the initial training of these large models.

+model on your data. There are a number methods for fine-tuning LLMs including low-rank adaptation or LoRA1, +that allow you to fine-tune a small number of parameters of a model's many parameters without needing a +large number of GPUs.

OpenAI allows you to train one of their models through their API and custom GPTs. Google Gemni offers fine-tuning through the Gemini API as does Anthropic as explained in their documentation.

Training LLaMA models locally on a personal computer is possible -depending on the capacities of your local computer. Unfortunately, the process isn't +depending on the resources of your local computer. Unfortunately, the process isn't easy or straight-forward and require running Python code. If you -have an Apple computer, the mlx-lm package,
-which uses Apple Silicon GPU, can be used for fine-tuning open-source models like LLaMA. Another -possibility is using HuggingFace's peft package along -with HuggingFace's transformers to fine-tune models.

+have an Apple computer, the mlx-lm package, which uses Apple Silicon GPU, +can be used for fine-tuning open-source models like LLaMA. Another possibility is using +HuggingFace's peft package that along with HuggingFace's transformers, can be used +to fine-tune models.

Steps for Fine-tuning

-

OpenAI provides directions for [fine-tuning]2 a ChatGPT model that are +

OpenAI provides directions for fine-tuning2 a ChatGPT model that are general enough for most fine-tuning tasks:

  1. Prepare and upload training data
    2. @@ -50,7 +50,7 @@

    Steps for Fine-tuning

  2. Use your fine-tuned model

Create Inventory Instance Training Set

-

We will create a training set 1,000 random records from the FOLIO community's Quesnelia Bugfest +

From the FOLIO community's Quesnelia Bugfest instance. The training set consists of the denormalize Instance along with a text prompt.

Depending on the model and service, you may need reformat the training set to match the expected inputs to the model.

diff --git a/exploring-llms/training-llms.md b/exploring-llms/training-llms.md index 985ff46..e78dc79 100644 --- a/exploring-llms/training-llms.md +++ b/exploring-llms/training-llms.md @@ -1,8 +1,8 @@ # Training Large Language Models (LLMs) One method for customizing and providing better context when using LLMs is to train the -LLM on your data. There are a number methods for fine-tuning LLMs including low-rank adaptation or LoRA[^LORA], -that allow you to fine-tune a small number of parameters of a LLM many parameters without a -large number of GPUs used for the initial training of these large models. +model on your data. There are a number methods for fine-tuning LLMs including low-rank adaptation or LoRA[^LORA], +that allow you to fine-tune a small number of parameters of a model's many parameters without needing a +large number of GPUs. [OpenAI][OPENAI] allows you to train one of their models through their API @@ -10,16 +10,15 @@ and custom GPTs. Google Gemni offers fine-tuning through the [Gemini API](https: as does Anthropic as explained in their [documentation](https://www.anthropic.com/news/fine-tune-claude-3-haiku). Training [LLaMA][LLAMA] models locally on a personal computer is possible -depending on the capacities of your local computer. Unfortunately, the process isn't +depending on the resources of your local computer. Unfortunately, the process isn't easy or straight-forward and require running Python code. If you -have an Apple computer, the [mlx-lm][MLX_LM] package, -which uses Apple Silicon GPU, can be used for fine-tuning open-source models like LLaMA. Another -possibility is using HuggingFace's [peft][PEFT] package along -with HuggingFace's transformers to fine-tune models. - +have an Apple computer, the [mlx-lm][MLX_LM] package, which uses Apple Silicon GPU, +can be used for fine-tuning open-source models like LLaMA. Another possibility is using +HuggingFace's [peft][PEFT] package that along with HuggingFace's transformers, can be used +to fine-tune models. ## Steps for Fine-tuning -OpenAI provides directions for [fine-tuning][^OPENAI_FINETUNE] a ChatGPT model that are +OpenAI provides directions for fine-tuning[^OPENAI_FINETUNE] a ChatGPT model that are general enough for most fine-tuning tasks: 1. Prepare and upload training data @@ -29,8 +28,7 @@ general enough for most fine-tuning tasks: ### Create Inventory Instance Training Set - -We will create a training set 1,000 random records from the FOLIO community's [Quesnelia Bugfest](https://bugfest-quesnelia.int.aws.folio.org/) +From the FOLIO community's [Quesnelia Bugfest](https://bugfest-quesnelia.int.aws.folio.org/) instance. The training set consists of the denormalize Instance along with a text prompt. Depending on the model and service, you may need reformat the training set to match the expected inputs