| repo: TheBloke/Manticore-13B-GGML | |
| file: Manticore-13B.ggmlv2.q5_1.bin | |
| llama_cpp: | |
| n_ctx: 2048 | |
| n_gpu_layers: 40 # llama 13b has 40 layers | |
| chat: | |
| stop: | |
| - "</s>" | |
| - "<unk>" | |
| - "### USER:" | |
| - "USER:" | |
| queue: | |
| max_size: 16 | |
| concurrency_count: 1 # leave this at 1, llama-cpp-python doesn't handle concurrent requests and will crash the entire app | |