llama general-runner

llama gguf menu picker

.gitignore

@@ -0,0 +1 @@
+/**/main*.log

bin/llama-menu.sh

@@ -0,0 +1,11 @@
+clear
+
+CHOICES=$(cd ~/.ai/models/llama && find . -name "*.gguf")
+SELECTED=$(echo "$CHOICES" | fzf --reverse --border --ansi)
+
+if [[ -z "$SELECTED" ]]; then
+    return
+fi
+
+echo "Selected: $SELECTED"
+llama.sh "$SELECTED" chatml 2048 0.8

bin/llama.sh

@@ -0,0 +1,161 @@
+#!/bin/bash
+
+# if not enough arguments
+if [[ $# -lt 4 ]]; then
+    echo "Usage: llama.sh <model> <template> <context-size> <temperature>"
+    exit 1
+fi
+
+# if conf does not exist, create it
+if [ ! -f "$HOME/.config/llama/llama-main.conf" ]; then
+    mkdir -p "$HOME/.config/llama"
+    cat <<EOF > "$HOME/.config/llama/llama-main.conf"
+LLAMA_TEMPERATURE=0.1
+LLAMA_CONTEXT_SIZE=4096
+LLAMA_REPETITION_PENALTY=1.15
+LLAMA_TOP_P=0.9
+LLAMA_TOP_K=20
+LLAMA_TEMPLATE=chatml
+LLAMA_MODEL_NAME=teknium/OpenHermes-2.5-Mistral-7B/openhermes-2.5-mistral-7b-f16.gguf
+EOF
+fi
+
+source $HOME/.config/llama/llama-main.conf
+
+function llama_interactive {
+    MODEL_NAME=$1
+    TEMPLATE=$2
+    CONTEXT_SIZE=$3
+    TEMPERATURE=$4
+
+    CMD=$HOME/.ai/bin/llama
+
+    $CMD \
+        --n-gpu-layers 1 \
+        --model "$(model_path $MODEL_NAME)" \
+        --prompt-cache "$(cache_path $MODEL_NAME)" \
+        --file "$(get_model_prompt $MODEL_NAME)" \
+        --in-prefix "$(get_model_prefix $TEMPLATE)" \
+        --in-suffix "$(get_model_suffix $TEMPLATE)" \
+        --reverse-prompt "$(get_model_prefix $TEMPLATE)" \
+        --reverse-prompt "<|im_end|>" \
+        --threads "7" \
+        --temp "$TEMPERATURE" \
+        --top-p "$LLAMA_TOP_P" \
+        --top-k "$LLAMA_TOP_K" \
+        --repeat-penalty "$LLAMA_REPETITION_PENALTY" \
+        --ctx-size "$CONTEXT_SIZE" \
+        --batch-size 1024 \
+        --n-predict -1 \
+        --keep -1 \
+        --instruct \
+        --no-mmap \
+        --color \
+        --escape
+}
+
+function model_path {
+    MODEL_NAME=$1
+    echo "$HOME/.ai/models/llama/${MODEL_NAME}"
+}
+
+function cache_path {
+    MODEL_NAME=$1
+    echo "$HOME/.ai/cache/menu-${MODEL_NAME//\//_}.cache"
+}
+
+function get_model_prefix {
+    TEMPLATE_NAME=$1
+
+    # if TEMPLATE_NAME contains string "guanaco"
+    if [[ $TEMPLATE_NAME == *"guanaco"* ]]; then
+        printf "### Human: "
+    elif [[ $TEMPLATE_NAME == *"alpaca"* ]]; then
+        printf "### Instruction: "
+    elif [[ $TEMPLATE_NAME == *"upstage"* ]]; then
+        printf "### Instruction: "
+    elif [[ $TEMPLATE_NAME == *"airoboros"* ]]; then
+        printf "### Instruction: "
+    elif [[ $TEMPLATE_NAME == *"hermes"* ]]; then
+        printf "### Instruction:"
+    elif [[ $TEMPLATE_NAME == *"vicuna"* ]]; then
+        printf "USER: "
+    elif [[ $TEMPLATE_NAME == *"based"* ]]; then
+        printf "Human: "
+    elif [[ $TEMPLATE_NAME == *"wizardlm"* ]]; then
+        printf "USER: "
+    elif [[ $TEMPLATE_NAME == *"orca"* ]]; then
+        printf "### User: "
+    elif [[ $TEMPLATE_NAME == *"samantha"* ]]; then
+        printf "USER: "
+    elif [[ $TEMPLATE_NAME == "chatml" ]]; then
+        printf "<|im_start|>user\\\n"
+    else
+        printf "Input: "
+    fi
+}
+
+#  USER: hello, who are you? ASSISTANT: 
+
+function get_model_suffix {
+    TEMPLATE_NAME=$1
+
+    # if TEMPLATE_NAME contains string "guanaco"
+    if [[ $TEMPLATE_NAME == *"guanaco"* ]]; then
+        printf "### Assistant: "
+    elif [[ $TEMPLATE_NAME == *"alpaca"* ]]; then
+        printf "### Response: "
+    elif [[ $TEMPLATE_NAME == *"airoboros"* ]]; then
+        printf "### Response: "
+    elif [[ $TEMPLATE_NAME == *"upstage"* ]]; then
+        printf "### Response: "
+    elif [[ $TEMPLATE_NAME == *"hermes"* ]]; then
+        printf "### Response: "
+    elif [[ $TEMPLATE_NAME == *"vicuna"* ]]; then
+        printf "ASSISTANT: "
+    elif [[ $TEMPLATE_NAME == *"samantha"* ]]; then
+        printf "ASSISTANT: "
+    elif [[ $TEMPLATE_NAME == *"based"* ]]; then
+        printf "Assistant: "
+    elif [[ $TEMPLATE_NAME == *"wizardlm"* ]]; then
+        printf "ASSISTANT: "
+    elif [[ $TEMPLATE_NAME == *"orca"* ]]; then
+        printf "### Response: "
+    elif [[ $TEMPLATE_NAME == "chatml" ]]; then
+        printf "<|im_end|>\n<|im_start|>assistant\\\n"
+    else
+        printf "Output: "
+    fi
+}
+
+function get_model_prompt {
+    MODEL_NAME=$1
+
+    if [[ $MODEL_NAME == *"guanaco"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/guanaco.txt"
+    elif [[ $MODEL_NAME == *"samantha"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/samantha.txt"
+    elif [[ $MODEL_NAME == *"openhermes-2-mistral-7b"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/hermes-mistral.txt"
+    elif [[ $MODEL_NAME == *"alpaca"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/alpaca.txt"
+    elif [[ $MODEL_NAME == *"upstage"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/alpaca.txt"
+    elif [[ $MODEL_NAME == *"airoboros"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/alpaca.txt"
+    elif [[ $MODEL_NAME == *"hermes"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/alpaca.txt"
+    elif [[ $MODEL_NAME == *"vicuna"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/vicuna-v11.txt"
+    elif [[ $MODEL_NAME == *"based"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/based.txt"
+    elif [[ $MODEL_NAME == *"wizardlm"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/wizardlm-30b.txt"
+    elif [[ $MODEL_NAME == *"orca"* ]]; then
+        echo "$HOME/.local/share/ai/prompts/orca.txt"
+    else
+        echo "$HOME/.local/share/ai/prompts/idm-gpt-lite.txt"
+    fi
+}
+
+llama_interactive "$1" "$2" "$3" "$4"

install.sh

@@ -7,5 +7,7 @@ mkdir -p ~/.local/bin
 install -C -v ./bin/llama-hf-to-q6_k.sh ~/.local/bin
 install -C -v ./bin/llama-update.sh ~/.local/bin
 install -C -v ./bin/llama-finetune.sh ~/.local/bin
+install -C -v ./bin/llama.sh ~/.local/bin
+install -C -v ./bin/llama-menu.sh ~/.local/bin
 
 echo "done"