Miscellaneous Updates -

* Fix Incorrect SD2 Inpainting Artifact & Broken Link to SDXL benchmark link
* Fix incorrect SD inpainting artifact
* Refactoring Performance Data
* Improve neuron-profile documentation
* Adding additional dependency for AL2023
* Update n2-helper.py
* Adding changes for Llama2 Training using aws batch
* Addressed PR comments
* Update training-trn1-samples.rst
* Formating fixes
* Adding Notes for Broken Tutorials

Author: aws-qieqingy

frameworks/torch/torch-neuronx/programming-guide/training/index.rst

@@ -8,7 +8,6 @@ Developer Guide  (``torch-neuronx``)
     /frameworks/torch/torch-neuronx/programming-guide/training/pytorch-neuron-programming-guide
     /frameworks/torch/torch-neuronx/programming-guide/training/pytorch-neuron-debug
     /frameworks/torch/torch-neuronx/programming-guide/torch-neuronx-profiling-dev-guide
-    /frameworks/torch/torch-neuronx/programming-guide/training/neuron-distributed-programming-guide
 
 
 .. dropdown::  Developer Guide

frameworks/torch/torch-neuronx/programming-guide/training/index.txt

@@ -1,4 +1,3 @@
 * :ref:`pytorch-neuronx-programming-guide`
 * :ref:`pytorch-neuronx-debug`
 * :ref:`torch-neuronx-dev-guide`
-* :ref:`neuronx-distributed`

frameworks/torch/torch-neuronx/training-troubleshooting.rst

@@ -61,17 +61,17 @@ On Ubuntu, if Apport is not running, core dump file name is by default "core" in
 
    echo '/tmp/core.%e.%p' | sudo tee /proc/sys/kernel/core_pattern
 
-For containers, install appropriate dependencies during docker build ("apt-get update && apt-get -y install build-essential gdb") and start the container with "--ulimit core=-1" to enable core dump and "-v /tmp/:/tmp/" to ensure core dumps to /tmp are preserved when container is stopped or deleted. Dependencies can also be installed after container is started.
+For containers, install appropriate dependencies during docker build ("apt-get update && apt-get -y install build-essential gdb") and start the container with ``--ulimit core=-1`` to enable core dump and ``-v /tmp/:/tmp/`` to ensure core dumps to ``/tmp`` are preserved when container is stopped or deleted. Dependencies can also be installed after container is started.
 
-On Ubuntu, core dumps can also handled by Apport which is disabled by default. To enable Apport, run ``sudo service apport start``. The ``/proc/sys/kernel/core_pattern`` is updated by Apport service. After a crash, look in /var/log/apport.log for the core dump file name, which should be in located in /var/lib/apport/coredump/.
+On Ubuntu, core dumps can also handled by Apport which is disabled by default. To enable Apport, run ``sudo service apport start``. The ``/proc/sys/kernel/core_pattern`` is updated by Apport service. After a crash, look in ``/var/log/apport.log`` for the core dump file name, which should be in located in ``/var/lib/apport/coredump/``.
 
 Once you have the core dump, you can use gdb to debug further (for Python applications, <executable> is ``python`` or ``python3``):
 
 .. code:: bash
 
    gdb <executable> <core file>
 
-If some process (i.e. XRT server) is killed due to out-of-memory on host (i.e. you see "Out of memory: Killed process <PID>" in syslog or dmesg), there won't be any core dump generated. However, you can change to it to kernel panic mode to trigger core dump by setting ``/proc/sys/vm/panic_on_oom`` to value of 1 on the host or from inside container.
+If some process (i.e. XRT server) is killed due to out-of-memory on host (i.e. you see ``Out of memory: Killed process <PID>`` in ``/var/log/syslog`` or output of ``dmesg``), there won't be any core dump generated. However, you can change to it to kernel panic mode to trigger core dump by setting ``/proc/sys/vm/panic_on_oom`` to value of 1 on the host or from inside container.
 
 On the host where you need ``sudo`` (this change will be reflected inside the container also):
 

frameworks/torch/torch-neuronx/tutorials/training/finetune_t5.rst

@@ -3,6 +3,9 @@
 Fine-tune T5 model on Trn1
 ================================
 
+.. note:: 
+   Update 01/03/24: This tutorial is currently broken and the AWS Neuron team is working on the fix.
+
 
 In this tutorial, we show how to fine-tune a Hugging Face (HF) T5 model 
 using HF trainer API. This example fine-tunes a `T5 model for

frameworks/torch/torch-neuronx/tutorials/training/zero1_gpt2.rst

@@ -3,6 +3,9 @@
 ZeRO-1 Tutorial
 ===============
 
+.. note::
+   Update 01/03/24: This tutorial is currently broken and the AWS Neuron team is working on the fix.
+
 What is ZeRO-1?
 ---------------
 
@@ -89,15 +92,13 @@ these arguments to the wrapper constructor:
 
 GPT2-XL Pretraining Tutorial
 ----------------------------
-Update 10/02:This tutorial is currently broken and the AWS Neuron team is working on the fix. 
 
-Table of Contents:
-///
--  Setup
--  Dataset
--  Training
+.. note::
+   Update 01/03/24: This tutorial is currently broken and the AWS Neuron team is working on the fix.
 
---------------
+.. contents:: Table of contents
+   :local:
+   :depth: 2
 
 Setup
 ~~~~~