fix lib

check_log.py

@@ -0,0 +1,29 @@
+# from huggingface_hub import HfApi
+
+# def main():
+#     api = HfApi()
+#     namespace = "uav-vts"
+#     job_id = "68c7eab1b8ca6fa7e35cf89b"
+
+#     print(f"🔍 Fetching logs for job {namespace}/{job_id} ...\n")
+#     for log in api.fetch_job_logs(namespace=namespace, job_id=job_id):
+#         print(log)
+
+# if __name__ == "__main__":
+#     main()
+
+
+from huggingface_hub import HfApi
+
+def main():
+    api = HfApi()
+    namespace = "uav-vts"
+    job_id = "68c7eab1b8ca6fa7e35cf89b"
+
+    job_info = api.inspect_job(job_id=job_id, namespace=namespace)
+    print("Job status:", job_info.status.stage)
+    print("Message:", job_info.status.message)
+    print("Full JobInfo:", job_info)
+
+if __name__ == "__main__":
+    main()

config.yaml

@@ -0,0 +1,10 @@
+name: training-sem
+image: pytorch/pytorch:2.6.0-cuda12.4-cudnn9-devel
+tasks:
+  - id: train
+    commands:
+      - pip install -r requirements.txt
+      - python runner/train.py
+resources:
+  gpu: A100
+repository: uav-vts/training_sem

requirements.txt

@@ -52,21 +52,16 @@ openmim==0.3.9
 openxlab==0.1.2
 ordered-set==4.1.0
 oss2==2.17.0
-packaging==24.2
+packaging==25.0
 pandas==2.0.2
 Pillow==10.0.0
 platformdirs==4.4.0
 polygon==1.1.0
 Polygon3==3.0.9.1
 pycocotools==2.0.10
-pycryptodome==3.23.0
-Pygments==2.19.2
-PyGObject==3.42.1
-PyJWT==2.3.0
 pyparsing==3.2.3
-python-apt==2.4.0+ubuntu4
 python-dateutil==2.9.0.post0
-pytz==2023.4
+pytz==2025.2
 PyYAML==6.0.2
 requests==2.28.2
 rich==13.4.2

train_hugg.py

@@ -0,0 +1,33 @@
+# from huggingface_hub import HfApi
+
+# api = HfApi()
+
+# job = api.create_job(
+#     repo_id="uav-vts/training_sem",
+#     docker_image="pytorch/pytorch:2.6.0-cuda12.4-cudnn9-devel",
+#     command="python runner/train.py",
+#     hardware="a100",   
+#     # token="<TOKEN>"
+# )
+
+# print("Job created:", job.job_id)
+
+
+from huggingface_hub import run_job
+
+image = "pytorch/pytorch:2.6.0-cuda12.4-cudnn9-devel"
+command = [
+    "python",
+    "-c",
+    "import torch; print(f'This code ran with the following GPU: {torch.cuda.get_device_name()}')"
+]
+
+job = run_job(
+    image=image,
+    command=command,
+    flavor="a100-large",   # hoặc "a100-large" nếu cần GPU mạnh hơn
+    namespace="uav-vts",   # để job thuộc account của bạn
+)
+
+print("Job submitted!")
+print("Track it here:", job.url)