mirror of
https://github.com/NicolasBohn/NexQuant.git
synced 2026-07-27 15:37:44 +00:00
TPTBusiness
362cff291c
Added to rdagent CLI: - rdagent start_llama: Start llama.cpp server (replaces start_llama.sh) - rdagent start_loop: Start strategy generator loop (replaces start_strategy_loop.sh) Features: - start_llama: --model, --port, --gpu-layers, --ctx-size, --reasoning options - start_loop: --target, --max-wait options with auto-restart on crash - Both commands have full help (--help) and examples Moved .sh scripts to scripts/: - start_llama.sh → scripts/ (kept as reference) - start_strategy_loop.sh → scripts/ (kept as reference) Usage: rdagent start_llama # Start LLM server rdagent start_llama --gpu-layers 40 # Custom GPU layers rdagent start_loop # Start strategy loop rdagent start_loop --target 5 # Generate 5 strategies per run
32 lines
827 B
Bash
Executable File
32 lines
827 B
Bash
Executable File
#!/bin/bash
|
|
# Start llama.cpp server with Qwen3.5-35B for strategy generation
|
|
# Usage: ./start_llama.sh
|
|
|
|
MODEL_PATH="$HOME/models/qwen3.5/Qwen3.5-35B-A3B-Q3_K_M.gguf"
|
|
PORT=8081
|
|
|
|
# GPU: RTX 5060 Ti (16GB VRAM)
|
|
# Modell braucht ~15.7GB bei voller GPU-Nutzung
|
|
# Mit 14GB free → reduzieren wir GPU-Layers + Context
|
|
GPU_LAYERS=30 # Weniger Layers für VRAM
|
|
CTX_SIZE=4096 # 4K Context (reicht für Strategien)
|
|
|
|
echo "🚀 Starting llama.cpp server..."
|
|
echo " Model: $(basename $MODEL_PATH)"
|
|
echo " Port: $PORT"
|
|
echo " GPU Layers: $GPU_LAYERS"
|
|
echo " Context: $CTX_SIZE"
|
|
echo ""
|
|
|
|
exec ~/llama.cpp/build/bin/llama-server \
|
|
--model "$MODEL_PATH" \
|
|
--n-gpu-layers $GPU_LAYERS \
|
|
--ctx-size $CTX_SIZE \
|
|
--port $PORT \
|
|
--threads 8 \
|
|
--threads-batch 8 \
|
|
--parallel 1 \
|
|
--flash-attn \
|
|
--jinja \
|
|
--host 0.0.0.0
|