From 6d78228bd77c739b4b7425560638c9d46c15fa05 Mon Sep 17 00:00:00 2001
From: Gregory Kielian <gkielian@google.com>
Date: Thu, 19 Dec 2024 15:21:03 +0000
Subject: [PATCH] Organize benchmark, inspections, modules

Removed unused modules, moved benchmarks and ckpt inspections
---
 .gitmodules                                              | 6 ------
 README.md                                                | 6 +++---
 bench.py => benchmarks/bench.py                          | 0
 softmax_sweep.py => benchmarks/softmax_sweep.py          | 0
 inspect_ckpts.py => checkpoint_analysis/inspect_ckpts.py | 0
 {steering_vector_util => demos}/obtain_vector.sh         | 0
 modules/llm.c                                            | 1 -
 modules/nanoGPT                                          | 1 -
 8 files changed, 3 insertions(+), 11 deletions(-)
 rename bench.py => benchmarks/bench.py (100%)
 rename softmax_sweep.py => benchmarks/softmax_sweep.py (100%)
 rename inspect_ckpts.py => checkpoint_analysis/inspect_ckpts.py (100%)
 rename {steering_vector_util => demos}/obtain_vector.sh (100%)
 delete mode 160000 modules/llm.c
 delete mode 160000 modules/nanoGPT

diff --git a/.gitmodules b/.gitmodules
index b652164cb6..d7708a16ef 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -1,9 +1,3 @@
 [submodule "data/template/whisper.cpp"]
 	path = data/template/whisper.cpp
 	url = https://github.com/ggerganov/whisper.cpp.git
-[submodule "modules/nanoGPT"]
-	path = modules/nanoGPT
-	url = https://github.com/karpathy/nanoGPT
-[submodule "modules/llm.c"]
-	path = modules/llm.c
-	url = https://github.com/karpathy/llm.c.git
diff --git a/README.md b/README.md
index 8dd829b1ff..09e8d571f3 100644
--- a/README.md
+++ b/README.md
@@ -174,7 +174,7 @@ parent directory.
 
 Example usage:
 ```bash
-python3 inspect_ckpts.py --directory ./out --sort loss
+python3 checkpoint_analysis/inspect_ckpts.py --directory ./out --sort loss
 ```
 
 ![image](./images/inspect_ckpts.png)
@@ -183,7 +183,7 @@ This can be wrapped with color via the watch command for a realtime dashboard.
 
 For example to look at all checkpoint files in the out directory:
 ```bash
-watch --color 'python3 inspect_ckpts.py --directory ./out --sort loss'
+watch --color 'python3 checkpoint_analysis/inspect_ckpts.py --directory ./out --sort loss'
 ```
 
 As with remainder of the repo, this script is provided as a base to open up for
@@ -194,7 +194,7 @@ additional community contributions.
 If using tensorboard for logging, we have provided a convenience script:
 
 ```bash
-bash start_tensorboard.sh
+source ./logging/start_tensorboard.sh
 ```
 
 You can view live validation loss updates on url: [http://localhost:6006](http://localhost:6006)
diff --git a/bench.py b/benchmarks/bench.py
similarity index 100%
rename from bench.py
rename to benchmarks/bench.py
diff --git a/softmax_sweep.py b/benchmarks/softmax_sweep.py
similarity index 100%
rename from softmax_sweep.py
rename to benchmarks/softmax_sweep.py
diff --git a/inspect_ckpts.py b/checkpoint_analysis/inspect_ckpts.py
similarity index 100%
rename from inspect_ckpts.py
rename to checkpoint_analysis/inspect_ckpts.py
diff --git a/steering_vector_util/obtain_vector.sh b/demos/obtain_vector.sh
similarity index 100%
rename from steering_vector_util/obtain_vector.sh
rename to demos/obtain_vector.sh
diff --git a/modules/llm.c b/modules/llm.c
deleted file mode 160000
index 2c9213731e..0000000000
--- a/modules/llm.c
+++ /dev/null
@@ -1 +0,0 @@
-Subproject commit 2c9213731ecbe6eff6b7ec597d3b760e9fe423ae
diff --git a/modules/nanoGPT b/modules/nanoGPT
deleted file mode 160000
index 9755682b98..0000000000
--- a/modules/nanoGPT
+++ /dev/null
@@ -1 +0,0 @@
-Subproject commit 9755682b981a45507f6eb9b11eadef8cb83cebd5