support kunlun r200

Dockerfile

@@ -0,0 +1,17 @@
+#FROM git.modelhub.org.cn:980/enginex-kunlunxin/xmlir/r200-8f_xmlir-ubuntu_2004_x86_64:v0.27
+
+FROM diffuser:r200-8f
+
+ENV HF_ENDPOINT=https://hf-mirror.com
+
+RUN /root/miniconda/envs/python38_torch201_cuda/bin/python3 -m pip install transformers==4.46.3 uvicorn\[standard\] fastapi -i https://pypi.tuna.tsinghua.edu.cn/simple
+
+WORKDIR /app
+
+COPY ./ /app
+
+RUN sed -i 's/(inv_freq_expanded\.float() @ position_ids_expanded\.float())/(inv_freq_expanded.float().clone() @ position_ids_expanded.float().clone())/g' /root/miniconda/envs/python38_torch201_cuda/lib/python3.8/site-packages/transformers/models/llama/modeling_llama.py
+
+EXPOSE 8000
+CMD ["sh", "-c", "/root/miniconda/envs/python38_torch201_cuda/bin/python3 server.py"]
+