Support AWS G7e (RTXPRO6000) instances

Author: jvstme

src/dstack/_internal/core/backends/aws/compute.py

@@ -1177,6 +1177,7 @@ def _supported_instances(offer: InstanceOffer) -> bool:
         "p4d.",
         "p4de.",
         "p3.",
+        "g7e.",
         "g6.",
         "g6e.",
         "gr6.",

src/dstack/_internal/server/services/backends/provisioning.py

@@ -9,6 +9,7 @@
 from dstack._internal.server.schemas.runner import GPUDevice
 from dstack._internal.server.services.docker import apply_server_docker_defaults, parse_image_name
 
+# https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa.html#efa-instance-types
 _AWS_EFA_ENABLED_INSTANCE_TYPE_PATTERNS = [
     # TODO: p6-b200 isn't supported yet in gpuhunt
     r"^p6-b200\.(48xlarge)$",
@@ -17,6 +18,7 @@
     r"^p5en\.(48xlarge)$",
     r"^p4d\.(24xlarge)$",
     r"^p4de\.(24xlarge)$",
+    r"^g7e\.(8xlarge|12xlarge|24xlarge|48xlarge)$",
     r"^g6\.(8xlarge|12xlarge|16xlarge|24xlarge|48xlarge)$",
     r"^g6e\.(8xlarge|12xlarge|16xlarge|24xlarge|48xlarge)$",
     r"^gr6\.8xlarge$",