From a3f7724fb5d47924dbc8710c911315e4002e5e9b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Yuan-Ting=20Hsieh=20=28=E8=AC=9D=E6=B2=85=E5=BB=B7=29?= Date: Tue, 5 Mar 2024 17:32:33 -0800 Subject: [PATCH] Fix NEMO links (#2385) --- .github/workflows/markdown-links-check.yml | 3 ++- .github/workflows/mlc_config.json | 7 +++++++ integration/nemo/examples/README.md | 8 ++++---- integration/nemo/examples/peft/README.md | 4 ++-- integration/nemo/examples/prompt_learning/README.md | 2 +- .../nemo/examples/prompt_learning/prompt_learning_20B.md | 4 ++-- .../nemo/examples/supervised_fine_tuning/README.md | 2 +- 7 files changed, 19 insertions(+), 11 deletions(-) create mode 100644 .github/workflows/mlc_config.json diff --git a/.github/workflows/markdown-links-check.yml b/.github/workflows/markdown-links-check.yml index 186c14d178..56fe4e7982 100644 --- a/.github/workflows/markdown-links-check.yml +++ b/.github/workflows/markdown-links-check.yml @@ -28,5 +28,6 @@ jobs: with: max-depth: -1 use-verbose-mode: 'yes' + config-file: '.github/workflows/mlc_config.json' check-modified-files-only: 'yes' - base-branch: 'dev' + base-branch: 'main' diff --git a/.github/workflows/mlc_config.json b/.github/workflows/mlc_config.json new file mode 100644 index 0000000000..3cff6d5621 --- /dev/null +++ b/.github/workflows/mlc_config.json @@ -0,0 +1,7 @@ +{ + "ignorePatterns": [ + { + "pattern": "catalog.ngc.nvidia.com" + } + ] +} \ No newline at end of file diff --git a/integration/nemo/examples/README.md b/integration/nemo/examples/README.md index 2e5bd21c2a..2584a984e8 100644 --- a/integration/nemo/examples/README.md +++ b/integration/nemo/examples/README.md @@ -1,16 +1,16 @@ # Examples of NeMo-NVFlare Integration ### [Parameter-Efficient Fine-Tuning (PEFT) with NeMo](./peft/README.md) -In this example, we utilize NeMo's [PEFT](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/nlp/nemo_megatron/peft/landing_page.html) using NVFlare's new Client API (minimal code changes required to run a NeMo script in FL) +In this example, we utilize NeMo's [PEFT](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/v1.22.0/nlp/nemo_megatron/peft/landing_page.html) using NVFlare's new Client API (minimal code changes required to run a NeMo script in FL) methods to showcase how to adapt a large language model (LLM) to a downstream task, such as financial sentiment predictions. ### [Supervised fine-tuning (SFT) with NeMo and NVFlare](./supervised_fine_tuning/README.md) -An example of using [NVIDIA FLARE](https://nvflare.readthedocs.io/en/main/index.html) +An example of using NVIDIA FLARE with NeMo for [supervised fine-tuning (SFT)](https://github.com/NVIDIA/NeMo-Megatron-Launcher#5152-sft-training) to fine-tune all parameters of a large language model (LLM) on supervised data to teach the model how to follow user specified instructions. ### [Prompt learning with NeMo and NVFlare](./prompt_learning/README.md) -An example of using [NVIDIA FLARE](https://nvflare.readthedocs.io/en/main/index.html) -with NeMo for [prompt learning](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/stable/nlp/nemo_megatron/prompt_learning.html) using NVFlare's Learner API +An example of using NVIDIA FLARE +with NeMo for [prompt learning](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/v1.17.0/nlp/nemo_megatron/prompt_learning.html) using NVFlare's Learner API to adapt a large language model (LLM) to a downstream task. diff --git a/integration/nemo/examples/peft/README.md b/integration/nemo/examples/peft/README.md index 01d116272d..458204a69e 100644 --- a/integration/nemo/examples/peft/README.md +++ b/integration/nemo/examples/peft/README.md @@ -1,13 +1,13 @@ ## Parameter-Efficient Fine-Tuning (PEFT) with NeMo -In this example, we utilize NeMo's [PEFT](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/nlp/nemo_megatron/peft/landing_page.html) +In this example, we utilize NeMo's [PEFT](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/v1.22.0/nlp/nemo_megatron/peft/landing_page.html) methods to showcase how to adapt a large language model (LLM) to a downstream task, such as financial sentiment predictions. With one line configuration change, you can try different PEFT techniques such as [p-tuning](https://arxiv.org/abs/2103.10385), [adapters](https://proceedings.mlr.press/v97/houlsby19a.html), or [LoRA](https://arxiv.org/abs/2106.09685), which add a small number of trainable parameters to the LLM that condition the model to produce the desired output for the downstream task. -For more details, see the [PEFT script](https://github.com/NVIDIA/NeMo/blob/main/examples/nlp/language_modeling/tuning/megatron_gpt_peft_tuning.py) in NeMo, which we adapt using NVFlare's Lightning client API to run in a federated scenario. +For more details, see the [PEFT script](https://github.com/NVIDIA/NeMo/blob/v1.22.0/examples/nlp/language_modeling/tuning/megatron_gpt_peft_tuning.py) in NeMo, which we adapt using NVFlare's Lightning client API to run in a federated scenario. ## Dependencies The example was tested with the [NeMo 23.10 container](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/nemo). diff --git a/integration/nemo/examples/prompt_learning/README.md b/integration/nemo/examples/prompt_learning/README.md index 0e2125a92f..5198458cf7 100644 --- a/integration/nemo/examples/prompt_learning/README.md +++ b/integration/nemo/examples/prompt_learning/README.md @@ -1,6 +1,6 @@ ## Prompt Learning with NeMo -In this example, we utilize NeMo's [prompt learning](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/stable/nlp/nemo_megatron/prompt_learning.html) +In this example, we utilize NeMo's [prompt learning](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/v1.17.0/nlp/nemo_megatron/prompt_learning.html) feature to showcase how to adapt a large language model (LLM) to a downstream task such as financial sentiment predictions. As the prompt learning technique shown in the example is p-tuning, which adds a small prompt encoder network to the LLM diff --git a/integration/nemo/examples/prompt_learning/prompt_learning_20B.md b/integration/nemo/examples/prompt_learning/prompt_learning_20B.md index 15d0bd72a1..670f2bc484 100644 --- a/integration/nemo/examples/prompt_learning/prompt_learning_20B.md +++ b/integration/nemo/examples/prompt_learning/prompt_learning_20B.md @@ -1,13 +1,13 @@ ## Federated p-tuning using a 20 billion parameter GPT model -In this example, we utilize NeMo's [prompt learning](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/stable/nlp/nemo_megatron/prompt_learning.html) +In this example, we utilize NeMo's [prompt learning](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/v1.17.0/nlp/nemo_megatron/prompt_learning.html) feature to showcase how to adapt a large language model (LLM) to a downstream task such as financial sentiment predictions. The prompt learning technique shown in the example is p-tuning which adds a small prompt encoder network to the LLM to produce virtual tokens that guide the model toward the desired output of the downstream task. -For more details on how to change hyperparameters for prompt learning in NeMo, see this [tutorial](https://github.com/NVIDIA/NeMo/blob/v1.22.0/tutorials/nlp/Multitask_Prompt_and_PTuning.ipynb) which is also the basis for this NVFlare tutorial. +For more details on how to change hyperparameters for prompt learning in NeMo, see this [tutorial](https://github.com/NVIDIA/NeMo/blob/v1.17.0/tutorials/nlp/Multitask_Prompt_and_PTuning.ipynb) which is also the basis for this NVFlare tutorial. ## Dependencies This example running a 20B GPT model requires more computational resources. diff --git a/integration/nemo/examples/supervised_fine_tuning/README.md b/integration/nemo/examples/supervised_fine_tuning/README.md index 4fb829ea7d..a55e759bf3 100644 --- a/integration/nemo/examples/supervised_fine_tuning/README.md +++ b/integration/nemo/examples/supervised_fine_tuning/README.md @@ -226,7 +226,7 @@ Below is the accuracy of the models on these tasks. As shown, FedAvg is able to generate a model with the best overall performance. ## Inference -We use NeMo's [inference script](https://github.com/NVIDIA/NeMo/blob/main/examples/nlp/language_modeling/megatron_gpt_eval.py) for generation task with models after SFT. +We use NeMo's [inference script](https://github.com/NVIDIA/NeMo/blob/v1.20.0/examples/nlp/language_modeling/megatron_gpt_eval.py) for generation task with models after SFT. Below, we define some test examples to feed to the SFT model to see its predictions. First, we ask the model to generate an answer to an open question: "Tell me an interesting fact about space travel."