RoversX's picture
Update config.yml
96f5367
raw
history blame contribute delete
395 Bytes
---
hub:
repo_id: RoversX/llama-2-7b-chat-hf-Qlora-Samantha-V2-ggml
filename: ggml-model-q4_0.bin
llama_cpp:
n_ctx: 2048
n_gpu_layers: 40 # llama 13b has 40 layers
chat:
stop:
- "</s>"
- "<unk>"
- "### human:"
- "human:"
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