From 792c5fa9f7665654bfe7b942cc6187531f92ad4f Mon Sep 17 00:00:00 2001
From: Yufeng Zhang <30943549+yufengzh@users.noreply.github.com>
Date: Sun, 25 Aug 2024 18:04:23 -0400
Subject: [PATCH] Add an example on how to customize dataparser with ns-train
 (#3385)

---
 docs/reference/cli/ns_train.md | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/docs/reference/cli/ns_train.md b/docs/reference/cli/ns_train.md
index 7b8a52620..65ed5206f 100644
--- a/docs/reference/cli/ns_train.md
+++ b/docs/reference/cli/ns_train.md
@@ -29,3 +29,11 @@ By default the nerfstudio dataparser is used. If you would like to use a differe
 ```bash
 ns-train {method} {dataparser} --help
 ```
+
+For example, if you want to specify the `eval_mode` of the nerfstudio dataparser to be `filename` when training your `splatfacto` model via the `ns-train` cli, you can do
+
+```
+ns-train splatfacto [method args] nerfstudio-data --eval-mode filename
+```
+
+Notice that the custom dataparser and its arguments are passed after specifying the training method and its arguments.