ollama-test, enhance to me like pv-test with the parameter looping, also fix some variables being set via env and add QUIET to suppress all but the table.

rc_shared.sh

@@ -1566,10 +1566,21 @@ alias ollama-cpu="ollama_threads "
 
 function ollama_benchmark_usage() {
 	cat <<- EOF
-		USAGE: ollama-benchmark CONTEXT [THREADS] [iGPU] [MODEL] [-h | --help]
+		USAGE: ollama-benchmark CONTEXTS [THREADS] [IGPUS] [MODELS] [-h | --help]
 
 		Tests installed models at a certain context, thread count, and iGPU setting.
 
+		PARAMETERS
+
+		   CONTEXTS
+		      Sent as num_ctx to ollama. No default.
+		   THREADS
+		      Sent as num_threads to ollama. Defaults to 0.
+		   IGPUS
+		      Values to set OLLAMA_IGPU_ENABLE to. Defaults to 1.
+		   MODELS
+		      Model names to test. Defaults to all of `ollama ls`.
+
 		EXAMPLES:
 
 		   Test models with at a context of 32000, default threads, and iGPU enabled.
@@ -1585,6 +1596,11 @@ function ollama_benchmark_usage() {
 		   context=64000 igpu=0 ollama-benchmark
 		   context=64000 model=qwen3.5:4b ollama-benchmark
 
+		The parameters can also act as arrays to cause extra looping.
+
+		   ollama-benchmark "32000 64000 128000 256000" "4 7 0" "0 1" \
+		      "qwen3.8:27b qwen3.6:35b qwen3.6:27b qwen3.5:35b qwen3.5:27b"
+
 		Aliases:
 
 		   ollama-test
@@ -1599,50 +1615,40 @@ function ollama_benchmark() {
 		return 0
 	fi
 
-	typeset -i context
+	contexts="${1:-$contexts}"
-	if [[ -z $context || $context == 0 ]]; then
+	if [[ -z "$contexts" ]]; then
-		context="$1"
-	fi
-	if [[ -z "$context" ]]; then
 		echo " ERROR: Context is required."
 		ollama_benchmark_usage
 		return 1
 	fi
 
-	typeset -i threads
+	threads="${2:-$threads}"
-	if [[ -z $threads ]]; then
-		threads="$2"
-	fi
 	if [[ -z $threads ]]; then
 		threads=0
 	fi
 
-	declare igpu
+	igpus="${3:-$igpus}"
-	if [[ -z "$igpu" ]] ; then
+	if [[ -z "$igpus" ]] ; then
-		igpu="$3"
+		igpus=1
-	fi
-	if [[ -z "$igpu" ]] ; then
-		igpu=1
-	fi
-	export OLLAMA_IGPU_ENABLE=$igpu
-	export OLLAMA_VULKAN=$igpu
-
-	declare model
-	if [[ -z "$model" ]] ; then
-		model="$4"
 	fi
 
-	echo -e "\n Sharing run parameters..."
+	models="${4:-$models}"
-	cat <<- EOF
+	if [[ -z "$models" ]]; then
+		models="`ollama ls | grep -v NAME | awk '{print $1}'`"
+	fi
+
+	DEBUG="${DEBUG}"
+	QUIET="${QUIET}"
+	[[ -z "$QUIET" ]] && echo -e "\n Sharing run parameters..."
+	[[ -z "$QUIET" ]] && cat <<- EOF
 		DEBUG=$DEBUG
+		QUIET=$QUIET
 		context=$context
 		threads=$threads
 		igpu=$igpu
-		OLLAMA_IGPU_ENABLE=$OLLAMA_IGPU_ENABLE
-		OLLAMA_VULKAN=$OLLAMA_IGPU_ENABLE
 		model=$model
 	EOF
-	echo " Done!"
+	[[ -z "$QUIET" ]] && echo " Done!"
 
 	# Variables
 
@@ -1651,11 +1657,13 @@ function ollama_benchmark() {
 	col_ctx=8
 	col_ram=6
 	col_thr=7
+	col_gpu=4
 	col_cpu=9
 	col_spd=11 col_spd_round=2
 	col_tim=8
 
 	log_file="ollama_benchmark.log.txt"
+	[[ -z "$QUIET" ]] && echo -e "\n Logging ollama to '$log_file'..."
 
 	declare champ_{speed,time}_{model,num}
 
@@ -1670,49 +1678,39 @@ function ollama_benchmark() {
 	# Main
 
 	if [[ ! -d /sdcard && -d /etc/sv ]]; then
-		echo -e "\n Stopping ollama runit service..."
+		[[ -z "$QUIET" ]] && echo -e "\n Stopping ollama runit service..."
-		sudo sv disable ollama
+		sudo sv disable ollama >/dev/null 2>&1
-		sudo sv stop ollama
+		sudo sv stop ollama >/dev/null 2>&1
 		sleep 5
-		echo " Done!"
+		[[ -z "$QUIET" ]] && echo " Done!"
 	fi
 
-	echo -e "\n Stopping any ollama processes..."
+	[[ -z "$QUIET" ]] && echo -e "\n Starting main loop... \n"
-	killall ollama
-	sleep 5
-	echo " Done!"
 
-	echo -e "\n Starting background ollama service, logging to '$log_file'..."
+	printf "| %-${col_mdl}s | %${col_hdd}s | %${col_ctx}s | %${col_ram}s | %${col_thr}s | %${col_gpu}s | %${col_cpu}s | %${col_spd}s | %${col_tim}s |\n" \
+		"Model" "HDD" "Context" "RAM" "Threads" "iGPU" "Processor" "Speed (t/s)" "Time (s)"
+
+	printf "| %-${col_mdl}s | %${col_hdd}s | %${col_ctx}s | %${col_ram}s | %${col_thr}s | %${col_gpu}s | %${col_cpu}s | %${col_spd}s | %${col_tim}s |\n" \
+		":------------------------" "-----:" "-------:" "-----:" "---:" "------:" "--------:" "----------:" "-------:"
+
+	for model in $models; do
+		for context in $contexts; do
+			for thread in $threads; do
+				for igpu in $igpus; do
+					(killall ollama >/dev/null 2>&1)
+					sleep 5
+					export OLLAMA_IGPU_ENABLE=$igpu
+					export OLLAMA_VULKAN=$igpu
 					(ollama serve >$log_file 2>&1 &)
 					wait_for_ollama
-	echo " Done!"
-
-	echo -e "\n Starting main loop... \n"
-
-	printf "| %-${col_mdl}s | %${col_hdd}s | %${col_ctx}s | %${col_ram}s | %${col_thr}s | %${col_cpu}s | %${col_spd}s | %${col_tim}s |\n" \
-		"Model @ CTX=$context" "HDD" "Context" "RAM" "Threads" "Processor" "Speed (t/s)" "Time (s)"
-
-	printf "| %-${col_mdl}s | %${col_hdd}s | %${col_ctx}s | %${col_ram}s | %${col_thr}s | %${col_cpu}s | %${col_spd}s | %${col_tim}s |\n" \
-		":------------------------" "-----:" "-------:" "-----:" "------:" "--------:" "----------:" "-------:"
-
-	model_focus="$model"
-	unset model
-	for model in `ollama ls | grep -v NAME | awk '{print $1}'`; do
-		# If a model was specified then only focus on it.
-		if [[ -n "$model_focus" ]]; then
-			if [[ "$model" != "$model_focus" ]]; then
-				continue;
-			fi
-		fi
 
 					hdd="`ollama ls | grep $model | awk '{print $3 $4}'`"
-
 					printf "| %-${col_mdl}s | %${col_hdd}s | " "$model" "$hdd"
 
 					# Start the model and store the response.
 					start=$SECONDS
 					speed_json=$(curl -s http://localhost:11434/api/generate \
-			-d '{"model":"'$model'","prompt":"Hello","stream":false,"keep_alive":"1m","options":{"num_ctx":'$context',"num_thread":'$threads'}}' \
+						-d '{"model":"'$model'","prompt":"Hello","stream":false,"keep_alive":"1m","options":{"num_ctx":'$context',"num_thread":'$thread'}}' \
 						2>/dev/null
 					)
 					status_ollama=$?
@@ -1775,8 +1773,8 @@ function ollama_benchmark() {
 						fi
 					fi
 
-		printf "%${col_ctx}s | %${col_ram}s | %${col_thr}s | %${col_cpu}s | %${col_spd}.${col_spd_round}f | %${col_tim}s |\n" \
+					printf "%${col_ctx}s | %${col_ram}s | %${col_thr}s | %${col_gpu}s | %${col_cpu}s | %${col_spd}.${col_spd_round}f | %${col_tim}s |\n" \
-			"$ctx" "$ram" "$threads" "$cpu" "$speed" "$time"
+						"$ctx" "$ram" "$thread" "$igpu" "$cpu" "$speed" "$time"
 
 					ollama stop $model >/dev/null 2>&1
 					while ollama ps | grep "$model"; do
@@ -1784,27 +1782,30 @@ function ollama_benchmark() {
 						sleep 1
 					done
 				done
+			done
+		done
+	done
 
-	echo -e "\n Done!! :D \n"
+	[[ -z "$QUIET" ]] && echo -e "\n Done!! :D \n"
-	cat <<- EOF
+	[[ -z "$QUIET" ]] && cat <<- EOF
 		Speed Champion: $champ_speed_model with a tokens/seconds of ${champ_speed_num}!
 		Time Champion: $champ_time_model with a time of ${champ_time_num}!
 	EOF
 
 	# Finish
 
-	echo -e "\n Killing local ollama serve..."
+	[[ -z "$QUIET" ]] && echo -e "\n Killing local ollama serve..."
 	killall ollama
-	echo " Done!"
+	[[ -z "$QUIET" ]] && echo " Done!"
 
 	if [[ ! -d /sdcard && -d /etc/sv ]]; then
-		echo -e "\n Restarting runit service..."
+		[[ -z "$QUIET" ]] && echo -e "\n Restarting runit service..."
 		sudo sv enable ollama
 		sudo sv start ollama
-		echo " Done!"
+		[[ -z "$QUIET" ]] && echo " Done!"
 	fi
 
-	echo -e "\n Done!!! :)"
+	[[ -z "$QUIET" ]] && echo -e "\n Done!!! :)"
 }
 alias ollama-benchmark="ollama_benchmark "
 alias ollama-test="ollama_benchmark "