-
**Is your enhancement related to a problem? Please describe.**
No
**Describe the solution you'd like**
is there any plan to support rk3588/rk3576 etc. ?
**Describe alternatives you've cons…
elcky updated
3 weeks ago
-
rk3576量化完精度损失很严重是什么原因？
没有特殊层，都是常用的层
models:
name: wheel_keypoint_model # 模型输出名称
platform: onnx # 原始模型使用的框架
model_file_path: ./wheel_keypoint.onnx #原模型路径
q…
-
### Describe the issue
Env:
- RK3576: Debian 12 aarch64, CPU only
- PC : Ubuntu20.04, x86
Ram
- 8GB
Description:
Hello,
1. I am currently working with the onnxruntime-training demo,…
-
and is memory footprint equivalent to a normal model?
-
编译内核文件时报下面的错误，我确实也没找到该变量的申明.
drivers/rknpu/rknpu_devfreq.c:237:25: error: ‘rockchip_opp_set_low_length’ undeclared here (not in a function)
237 | .set_soc_info = rockchip_opp_set_low_l…
elcky updated
4 weeks ago
-
模型转换后在rk3576推理结果完全不对，根据文档进行精度排查，模拟器FP16和量化精度都没有问题，连板精度异常，请问怎么解决？
-
现在只能支持3588和3576吗？用target_platform='rk3576'导出的量化模型可以用在3566吗?
-
rga不支持NV24的缩放，请问是否有支持的计划？如果不提供支持，请问有什么替代方法吗？
-
Rockchip mpp uses functions with the same name as libvpx. This leads to a multiple definition error when using both, mpp and libvpx, in a project together. Here is the build output when trying to buil…
-
运行日志如下：
~/llm/demo/server_demo$ sudo python flask_server.py --target_platform rk3588 --rkllm_model_path ../../model/minicpm.rkllm
=========init....===========
rkllm-runtime version: 1.0.1, rknpu d…