From 1d86bd87fbf5fe89d083c8c40275d6d4ef74d411 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?= <sigbjorn.skjaeret@scala.com>
Date: Sat, 13 Apr 2024 12:09:09 +0200
Subject: [PATCH 1/7] Fix --split-max-size

Byte size calculation was done on int and overflowed.
---
 examples/gguf-split/gguf-split.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/examples/gguf-split/gguf-split.cpp b/examples/gguf-split/gguf-split.cpp
index 24acbf02a4eed..39c75e0a7a802 100644
--- a/examples/gguf-split/gguf-split.cpp
+++ b/examples/gguf-split/gguf-split.cpp
@@ -59,10 +59,10 @@ static size_t split_str_to_n_bytes(std::string str) {
     int n;
     if (str.back() == 'M') {
         sscanf(str.c_str(), "%d", &n);
-        n_bytes = n * 1024 * 1024; // megabytes
+        n_bytes = (size_t)n * 1024 * 1024; // megabytes
     } else if (str.back() == 'G') {
         sscanf(str.c_str(), "%d", &n);
-        n_bytes = n * 1024 * 1024 * 1024; // gigabytes
+        n_bytes = (size_t)n * 1024 * 1024 * 1024; // gigabytes
     } else {
         throw std::invalid_argument("error: supported units are M (megabytes) or G (gigabytes), but got: " + std::string(1, str.back()));
     }

From 6738215a10bd4d128357b264eba1382303ef26ba Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?= <sigbjorn.skjaeret@scala.com>
Date: Sat, 13 Apr 2024 16:01:34 +0200
Subject: [PATCH 2/7] add tests.sh

---
 examples/gguf-split/tests.sh | 81 ++++++++++++++++++++++++++++++++++++
 1 file changed, 81 insertions(+)
 create mode 100644 examples/gguf-split/tests.sh

diff --git a/examples/gguf-split/tests.sh b/examples/gguf-split/tests.sh
new file mode 100644
index 0000000000000..033c5f3858b21
--- /dev/null
+++ b/examples/gguf-split/tests.sh
@@ -0,0 +1,81 @@
+#!/bin/bash
+
+set -eu
+
+if [ $# -lt 1 ]
+then
+  echo "usage:   $0 path_to_build_binary [path_to_temp_folder]"
+  echo "example: $0 ../../build/bin ../../tmp"
+  exit 1
+fi
+
+if [ $# -gt 1 ]
+then
+  TMP_DIR=$2
+else
+  TMP_DIR=/tmp
+fi
+
+set -x
+
+SPLIT=$1/gguf-split
+MAIN=$1/main
+WORK_PATH=$TMP_DIR
+CUR_DIR=$(pwd)
+
+# 1. Get a model
+(
+  cd $WORK_PATH
+  "$CUR_DIR"/../../scripts/hf.sh --repo ggml-org/gemma-1.1-2b-it-Q8_0-GGUF --file gemma-1.1-2b-it.Q8_0.gguf
+)
+echo PASS
+
+# 2. Split with max tensors strategy
+$SPLIT --split-max-tensors 28  $WORK_PATH/gemma-1.1-2b-it.Q8_0.gguf $WORK_PATH/ggml-model-split
+echo PASS
+echo
+
+# 2b. Test the sharded model is loading properly
+$MAIN --model $WORK_PATH/ggml-model-split-00001-of-00006.gguf --random-prompt --n-predict 32
+echo PASS
+echo
+
+# 3. Merge
+$SPLIT --merge $WORK_PATH/ggml-model-split-00001-of-00006.gguf $WORK_PATH/ggml-model-merge.gguf
+echo PASS
+echo
+
+# 3b. Test the merged model is loading properly
+$MAIN --model $WORK_PATH/ggml-model-merge.gguf --random-prompt --n-predict 32
+echo PASS
+echo
+
+# 4. Split with no tensor in metadata
+#$SPLIT --split-max-tensors 32 --no-tensor-in-metadata $WORK_PATH/ggml-model-merge.gguf $WORK_PATH/ggml-model-split-32-tensors
+#echo PASS
+#echo
+
+# 4b. Test the sharded model is loading properly
+#$MAIN --model $WORK_PATH/ggml-model-split-32-tensors-00001-of-00006.gguf --random-prompt --n-predict 32
+#echo PASS
+#echo
+
+# 5. Merge
+#$SPLIT --merge $WORK_PATH/ggml-model-split-32-tensors-00001-of-00006.gguf $WORK_PATH/ggml-model-merge-2.gguf
+#echo PASS
+#echo
+
+# 5b. Test the merged model is loading properly
+#$MAIN --model $WORK_PATH/ggml-model-merge-2.gguf --random-prompt --n-predict 32
+#echo PASS
+#echo
+
+# 6. Split with size strategy
+$SPLIT --split-max-size 2G $WORK_PATH/ggml-model-merge.gguf $WORK_PATH/ggml-model-split-2G
+echo PASS
+echo
+
+# 6b. Test the sharded model is loading properly
+$MAIN --model $WORK_PATH/ggml-model-split-2G-00001-of-00002.gguf --random-prompt --n-predict 32
+echo PASS
+echo

From d42add49d4530fd5547ac77f0313f8dae557c41b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?= <sigbjorn.skjaeret@scala.com>
Date: Sat, 13 Apr 2024 16:04:00 +0200
Subject: [PATCH 3/7] add examples test scripts to ci run

Will autodiscover examples/*/tests.sh scripts and run them.
---
 ci/run.sh | 49 +++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)

diff --git a/ci/run.sh b/ci/run.sh
index 19776b5f7c6ff..8521e1f074339 100755
--- a/ci/run.sh
+++ b/ci/run.sh
@@ -153,6 +153,52 @@ function gg_sum_ctest_release {
     gg_printf '```\n'
 }
 
+# test_scripts_debug
+
+function gg_run_test_scripts_debug {
+    cd ${SRC}
+
+    set -e
+
+    (find ./examples -mindepth 2 -maxdepth 2 -name 'tests.sh' -execdir time bash "{}" "$SRC/build-ci-debug/bin" "$MNT/models" \;) 2>&1 | tee -a $OUT/${ci}-scripts.log
+
+    set +e
+}
+
+function gg_sum_test_scripts_debug {
+    gg_printf '### %s\n\n' "${ci}"
+
+    gg_printf 'Runs test scripts in debug mode\n'
+    gg_printf '- status: %s\n' "$(cat $OUT/${ci}.exit)"
+    gg_printf '```\n'
+    gg_printf '%s\n' "$(cat $OUT/${ci}-scripts.log)"
+    gg_printf '```\n'
+    gg_printf '\n'
+}
+
+# test_scripts_release
+
+function gg_run_test_scripts_release {
+    cd ${SRC}
+
+    set -e
+
+    (find ./examples -mindepth 2 -maxdepth 2 -name 'tests.sh' -execdir time bash "{}" "$SRC/build-ci-release/bin" "$MNT/models" \;) 2>&1 | tee -a $OUT/${ci}-scripts.log
+
+    set +e
+}
+
+function gg_sum_test_scripts_release {
+    gg_printf '### %s\n\n' "${ci}"
+
+    gg_printf 'Runs test scripts in release mode\n'
+    gg_printf '- status: %s\n' "$(cat $OUT/${ci}.exit)"
+    gg_printf '```\n'
+    gg_printf '%s\n' "$(cat $OUT/${ci}-scripts.log)"
+    gg_printf '```\n'
+    gg_printf '\n'
+}
+
 function gg_get_model {
     local gguf_3b="$MNT/models/open-llama/3B-v2/ggml-model-f16.gguf"
     local gguf_7b="$MNT/models/open-llama/7B-v2/ggml-model-f16.gguf"
@@ -642,6 +688,9 @@ test $ret -eq 0 && gg_run ctest_release
 if [ -z ${GG_BUILD_LOW_PERF} ]; then
     test $ret -eq 0 && gg_run embd_bge_small
 
+    test $ret -eq 0 && gg_run test_scripts_debug
+    test $ret -eq 0 && gg_run test_scripts_release
+
     if [ -z ${GG_BUILD_VRAM_GB} ] || [ ${GG_BUILD_VRAM_GB} -ge 8 ]; then
         if [ -z ${GG_BUILD_CUDA} ]; then
             test $ret -eq 0 && gg_run open_llama_3b_v2

From 18ed9ed57f55627cdc51978ebb81210d017d86e3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?= <sigbjorn.skjaeret@scala.com>
Date: Sat, 13 Apr 2024 22:22:29 +0200
Subject: [PATCH 4/7] move WORK_PATH to a subdirectory

---
 examples/gguf-split/tests.sh | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/examples/gguf-split/tests.sh b/examples/gguf-split/tests.sh
index 033c5f3858b21..d6079b497ecc8 100644
--- a/examples/gguf-split/tests.sh
+++ b/examples/gguf-split/tests.sh
@@ -20,9 +20,11 @@ set -x
 
 SPLIT=$1/gguf-split
 MAIN=$1/main
-WORK_PATH=$TMP_DIR
+WORK_PATH=$TMP_DIR/gguf-split
 CUR_DIR=$(pwd)
 
+mkdir -p "$WORK_PATH"
+
 # 1. Get a model
 (
   cd $WORK_PATH

From e53bc29c25a1740676e3de9acfe3e3dc554ed512 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?= <sigbjorn.skjaeret@scala.com>
Date: Sun, 14 Apr 2024 11:04:02 +0200
Subject: [PATCH 5/7] clean up before and after test

---
 examples/gguf-split/tests.sh | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/examples/gguf-split/tests.sh b/examples/gguf-split/tests.sh
index d6079b497ecc8..879522f7e2512 100644
--- a/examples/gguf-split/tests.sh
+++ b/examples/gguf-split/tests.sh
@@ -25,6 +25,9 @@ CUR_DIR=$(pwd)
 
 mkdir -p "$WORK_PATH"
 
+# Clean up in case of previously failed test
+rm -f $WORK_PATH/ggml-model-split*.gguf $WORK_PATH/ggml-model-merge*.gguf
+
 # 1. Get a model
 (
   cd $WORK_PATH
@@ -81,3 +84,6 @@ echo
 $MAIN --model $WORK_PATH/ggml-model-split-2G-00001-of-00002.gguf --random-prompt --n-predict 32
 echo PASS
 echo
+
+# Clean up
+rm -f $WORK_PATH/ggml-model-split*.gguf $WORK_PATH/ggml-model-merge*.gguf

From 708a0b05165f366e24bb4f0a9141e4bcc7b08437 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?= <sigbjorn.skjaeret@scala.com>
Date: Sun, 14 Apr 2024 11:15:49 +0200
Subject: [PATCH 6/7] explicitly define which scripts to run

---
 ci/run.sh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/ci/run.sh b/ci/run.sh
index 8521e1f074339..085dfd42faf32 100755
--- a/ci/run.sh
+++ b/ci/run.sh
@@ -160,7 +160,7 @@ function gg_run_test_scripts_debug {
 
     set -e
 
-    (find ./examples -mindepth 2 -maxdepth 2 -name 'tests.sh' -execdir time bash "{}" "$SRC/build-ci-debug/bin" "$MNT/models" \;) 2>&1 | tee -a $OUT/${ci}-scripts.log
+    (cd ./examples/gguf-split && time bash tests.sh "$SRC/build-ci-debug/bin" "$MNT/models") 2>&1 | tee -a $OUT/${ci}-scripts.log
 
     set +e
 }
@@ -183,7 +183,7 @@ function gg_run_test_scripts_release {
 
     set -e
 
-    (find ./examples -mindepth 2 -maxdepth 2 -name 'tests.sh' -execdir time bash "{}" "$SRC/build-ci-release/bin" "$MNT/models" \;) 2>&1 | tee -a $OUT/${ci}-scripts.log
+    (cd ./examples/gguf-split && time bash tests.sh "$SRC/build-ci-release/bin" "$MNT/models") 2>&1 | tee -a $OUT/${ci}-scripts.log
 
     set +e
 }

From 650db0f25f28cdd1ceb53b57b052794283d12419 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?= <sigbjorn.skjaeret@scala.com>
Date: Sun, 14 Apr 2024 12:00:05 +0200
Subject: [PATCH 7/7] add --split-max-size to readme

---
 examples/gguf-split/README.md | 1 +
 1 file changed, 1 insertion(+)

diff --git a/examples/gguf-split/README.md b/examples/gguf-split/README.md
index ddb1f76497aed..ad1d86651b46d 100644
--- a/examples/gguf-split/README.md
+++ b/examples/gguf-split/README.md
@@ -5,5 +5,6 @@ CLI to split / merge GGUF files.
 **Command line options:**
 
 - `--split`: split GGUF to multiple GGUF, default operation.
+- `--split-max-size`: max size per split in `M` or `G`, f.ex. `500M` or `2G`.
 - `--split-max-tensors`: maximum tensors in each split: default(128)
 - `--merge`: merge multiple GGUF to a single GGUF.