基础环境
8K100 64GB
2海贡 C86 7380 32 核
1TB 内存

使用镜像image.sourcefind.cn:5000/dcu/admin/base/pytorch 2.1.0-py3.10-dtk24.04.3-ubuntu20.04。运行ollama时，监测到了8张dcu，但是运行模型时时cpu在运行，dcu信息根本不占用
time=2025-02-13T14:18:59.735+08:00 level=INFO source=download.go:175 msg="downloading c7f3ea903b50 in 1 488 B part(s)"
[GIN] 2025/02/13 - 14:19:21 | 200 | 17m18s | 127.0.0.1 | POST "/api/pull"
[GIN] 2025/02/13 - 14:19:21 | 200 | 31.954341ms | 127.0.0.1 | POST "/api/show"
time=2025-02-13T14:19:21.639+08:00 level=INFO source=sched.go:714 msg="new model will fit in available VRAM in single GPU, loading" model=/root/.ollama/models/blobs/sha256-6150cb382311b69f09cc0f9a1b69fc029cbd742b66bb8ec531aa5ecf5c613e93 gpu=GPU-709ccd6899844861 parallel=4 available=68700557312 required="21.5 GiB"
time=2025-02-13T14:19:21.639+08:00 level=INFO source=server.go:104 msg="system memory" total="995.9 GiB" free="974.2 GiB" free_swap="0 B"
time=2025-02-13T14:19:21.641+08:00 level=INFO source=memory.go:356 msg="offload to rocm" layers.requested=-1 layers.model=65 layers.offload=65 layers.split="" memory.available="[64.0 GiB]" memory.gpu_overhead="0 B" memory.required.full="21.5 GiB" memory.required.partial="21.5 GiB" memory.required.kv="2.0 GiB" memory.required.allocations="[21.5 GiB]" memory.weights.total="19.5 GiB" memory.weights.repeating="18.9 GiB" memory.weights.nonrepeating="609.1 MiB" memory.graph.full="676.0 MiB" memory.graph.partial="916.1 MiB"
time=2025-02-13T14:19:21.641+08:00 level=INFO source=server.go:376 msg="starting llama server" cmd="/home/ollama/llama/build/linux-amd64/runners/cpu_avx2/ollama_llama_server runner --model /root/.ollama/models/blobs/sha256-6150cb382311b69f09cc0f9a1b69fc029cbd742b66bb8ec531aa5ecf5c613e93 --ctx-size 8192 --batch-size 512 --n-gpu-layers 65 --threads 64 --parallel 4 --port 40585"