diff --git a/sky/clouds/service_catalog/__init__.py b/sky/clouds/service_catalog/__init__.py index 4deab8ac204..d28b530ff06 100644 --- a/sky/clouds/service_catalog/__init__.py +++ b/sky/clouds/service_catalog/__init__.py @@ -324,9 +324,8 @@ def get_common_gpus() -> List[str]: 'A100', 'A100-80GB', 'H100', - 'K80', 'L4', - 'M60', + 'L40S', 'P100', 'T4', 'V100', @@ -337,13 +336,13 @@ def get_common_gpus() -> List[str]: def get_tpus() -> List[str]: """Returns a list of TPU names.""" # TODO(wei-lin): refactor below hard-coded list. - # There are many TPU configurations available, we show the three smallest - # and the largest configuration for the latest gen TPUs. + # There are many TPU configurations available, we show the some smallest + # ones for each generation, and people should find larger ones with + # sky show-gpus tpu. return [ - 'tpu-v2-512', 'tpu-v3-2048', 'tpu-v4-8', 'tpu-v4-16', 'tpu-v4-32', - 'tpu-v4-3968', 'tpu-v5litepod-1', 'tpu-v5litepod-4', 'tpu-v5litepod-8', - 'tpu-v5litepod-256', 'tpu-v5p-8', 'tpu-v5p-32', 'tpu-v5p-128', - 'tpu-v5p-12288' + 'tpu-v2-8', 'tpu-v3-8', 'tpu-v4-8', 'tpu-v4-16', 'tpu-v4-32', + 'tpu-v5litepod-1', 'tpu-v5litepod-4', 'tpu-v5litepod-8', 'tpu-v5p-8', + 'tpu-v5p-16', 'tpu-v5p-32', 'tpu-v6e-1', 'tpu-v6e-4', 'tpu-v6e-8' ] diff --git a/sky/clouds/service_catalog/data_fetchers/fetch_azure.py b/sky/clouds/service_catalog/data_fetchers/fetch_azure.py index f646cac339a..4aef41f9c90 100644 --- a/sky/clouds/service_catalog/data_fetchers/fetch_azure.py +++ b/sky/clouds/service_catalog/data_fetchers/fetch_azure.py @@ -64,7 +64,7 @@ 'standardNVSv2Family': 'M60', 'standardNVSv3Family': 'M60', 'standardNVPromoFamily': 'M60', - 'standardNVSv4Family': 'Radeon MI25', + 'standardNVSv4Family': 'MI25', 'standardNDSFamily': 'P40', 'StandardNVADSA10v5Family': 'A10', 'StandardNCadsH100v5Family': 'H100', diff --git a/sky/provision/kubernetes/utils.py b/sky/provision/kubernetes/utils.py index 4bd6a7fadb7..9a1de4e9036 100644 --- a/sky/provision/kubernetes/utils.py +++ b/sky/provision/kubernetes/utils.py @@ -1136,7 +1136,11 @@ def name(self) -> str: name = (f'{common_utils.format_float(self.cpus)}CPU--' f'{common_utils.format_float(self.memory)}GB') if self.accelerator_count: - name += f'--{self.accelerator_count}{self.accelerator_type}' + # Replace spaces with underscores in accelerator type to make it a + # valid logical instance type name. + assert self.accelerator_type is not None, self.accelerator_count + acc_name = self.accelerator_type.replace(' ', '_') + name += f'--{self.accelerator_count}{acc_name}' return name @staticmethod @@ -1167,7 +1171,9 @@ def _parse_instance_type( accelerator_type = match.group('accelerator_type') if accelerator_count: accelerator_count = int(accelerator_count) - accelerator_type = str(accelerator_type) + # This is to revert the accelerator types with spaces back to + # the original format. + accelerator_type = str(accelerator_type).replace('_', ' ') else: accelerator_count = None accelerator_type = None