yimuu

[Feature] LLaVA 1.5 WINT 8 量化

1

### Motivation 能否支持LLaVA 1.5 WINT 8 量化 ### Related resources _No response_ ### Additional context _No response_

Change the command to `CMAKE_ARGS="-DGGML_CUDA=on -DLLAVA_BUILD=off" pip install -U llama-cpp-python --force-reinstall --no-cache-dir` solved the problem.

1

Change the command to `CMAKE_ARGS="-DGGML_CUDA=on -DLLAVA_BUILD=off" pip install -U llama-cpp-python --force-reinstall --no-cache-dir` solved the problem. _Originally posted by @blx0102 in https://github.com/abetlen/llama-cpp-python/issues/1639#issuecomment-2257757873_ can not find libllava.so

llava-onevision-qwen2-7b-ov finetune based on LoRA

2

When I fine-tune using Lora, the model's convergence effect is not good. The hyperparameters are set as follows: --lora_enable True \ --deepspeed scripts/zero3.json \ --model_name_or_path ${MODEL} \ --version ${PROMPT_VERSION} \...