skypilot-nightly 1.0.0.dev2024053101__py3-none-any.whl → 1.0.0.dev2025022801__py3-none-any.whl

sky/clouds/service_catalog/data_fetchers/fetch_gcp.py

@@ -5,11 +5,11 @@ VMs, GPUs, and TPUs. The script takes about 1-2 minutes to run.
 """
 
 import argparse
-import functools
 import io
 import multiprocessing
 import os
 import textwrap
+import time
 import typing
 from typing import Any, Callable, Dict, List, Optional, Set
 
@@ -19,6 +19,8 @@ import numpy as np
 
 from sky.adaptors import common as adaptors_common
 from sky.adaptors import gcp
+from sky.utils import annotations
+from sky.utils import common_utils
 
 if typing.TYPE_CHECKING:
     import pandas as pd
@@ -38,6 +40,9 @@ TPU_SERVICE_ID = 'E000-3F24-B8AA'
 # The number of digits to round the price to.
 PRICE_ROUNDING = 5
 
+# The number of retries for the TPU API.
+TPU_RETRY_CNT = 3
+
 # This zone is only for TPU v4, and does not appear in the skus yet.
 TPU_V4_ZONES = ['us-central2-b']
 # TPU v3 pods are available in us-east1-d, but hidden in the skus.
@@ -54,6 +59,113 @@ HIDDEN_TPU_DF = pd.read_csv(
  ,tpu-v3-1024,1,,,tpu-v3-1024,1024.0,307.2,us-east1,us-east1-d
  ,tpu-v3-2048,1,,,tpu-v3-2048,2048.0,614.4,us-east1,us-east1-d
  """)))
+
+# TPU V6e price for the following regions is missing in the SKUs.
+TPU_V6E_MISSING_REGIONS = ['us-central2', 'southamerica-west1']
+
+# TPU V5 is not visible in specific zones. We hardcode the missing zones here.
+# NOTE(dev): Keep the zones and the df in sync.
+TPU_V5_MISSING_ZONES_DF = {
+    'europe-west4-b': pd.read_csv(
+        io.StringIO(
+            textwrap.dedent("""\
+ AcceleratorName,AcceleratorCount,Region,AvailabilityZone
+ tpu-v5p-8,1,europe-west4,europe-west4-b
+ tpu-v5p-16,1,europe-west4,europe-west4-b
+ tpu-v5p-32,1,europe-west4,europe-west4-b
+ tpu-v5p-64,1,europe-west4,europe-west4-b
+ tpu-v5p-128,1,europe-west4,europe-west4-b
+ tpu-v5p-256,1,europe-west4,europe-west4-b
+ tpu-v5p-384,1,europe-west4,europe-west4-b
+ tpu-v5p-512,1,europe-west4,europe-west4-b
+ tpu-v5p-640,1,europe-west4,europe-west4-b
+ tpu-v5p-768,1,europe-west4,europe-west4-b
+ tpu-v5p-896,1,europe-west4,europe-west4-b
+ tpu-v5p-1024,1,europe-west4,europe-west4-b
+ tpu-v5p-1152,1,europe-west4,europe-west4-b
+ tpu-v5p-1280,1,europe-west4,europe-west4-b
+ tpu-v5p-1408,1,europe-west4,europe-west4-b
+ tpu-v5p-1536,1,europe-west4,europe-west4-b
+ tpu-v5p-1664,1,europe-west4,europe-west4-b
+ tpu-v5p-1792,1,europe-west4,europe-west4-b
+ tpu-v5p-1920,1,europe-west4,europe-west4-b
+ tpu-v5p-2048,1,europe-west4,europe-west4-b
+ tpu-v5p-2176,1,europe-west4,europe-west4-b
+ tpu-v5p-2304,1,europe-west4,europe-west4-b
+ tpu-v5p-2432,1,europe-west4,europe-west4-b
+ tpu-v5p-2560,1,europe-west4,europe-west4-b
+ tpu-v5p-2688,1,europe-west4,europe-west4-b
+ tpu-v5p-2816,1,europe-west4,europe-west4-b
+ tpu-v5p-2944,1,europe-west4,europe-west4-b
+ tpu-v5p-3072,1,europe-west4,europe-west4-b
+ tpu-v5p-3200,1,europe-west4,europe-west4-b
+ tpu-v5p-3328,1,europe-west4,europe-west4-b
+ tpu-v5p-3456,1,europe-west4,europe-west4-b
+ tpu-v5p-3584,1,europe-west4,europe-west4-b
+ tpu-v5p-3712,1,europe-west4,europe-west4-b
+ tpu-v5p-3840,1,europe-west4,europe-west4-b
+ tpu-v5p-3968,1,europe-west4,europe-west4-b
+ tpu-v5p-4096,1,europe-west4,europe-west4-b
+ tpu-v5p-4224,1,europe-west4,europe-west4-b
+ tpu-v5p-4352,1,europe-west4,europe-west4-b
+ tpu-v5p-4480,1,europe-west4,europe-west4-b
+ tpu-v5p-4608,1,europe-west4,europe-west4-b
+ tpu-v5p-4736,1,europe-west4,europe-west4-b
+ tpu-v5p-4864,1,europe-west4,europe-west4-b
+ tpu-v5p-4992,1,europe-west4,europe-west4-b
+ tpu-v5p-5120,1,europe-west4,europe-west4-b
+ tpu-v5p-5248,1,europe-west4,europe-west4-b
+ tpu-v5p-5376,1,europe-west4,europe-west4-b
+ tpu-v5p-5504,1,europe-west4,europe-west4-b
+ tpu-v5p-5632,1,europe-west4,europe-west4-b
+ tpu-v5p-5760,1,europe-west4,europe-west4-b
+ tpu-v5p-5888,1,europe-west4,europe-west4-b
+ tpu-v5p-6016,1,europe-west4,europe-west4-b
+ tpu-v5p-6144,1,europe-west4,europe-west4-b
+ tpu-v5p-6272,1,europe-west4,europe-west4-b
+ tpu-v5p-6400,1,europe-west4,europe-west4-b
+ tpu-v5p-6528,1,europe-west4,europe-west4-b
+ tpu-v5p-6656,1,europe-west4,europe-west4-b
+ tpu-v5p-6784,1,europe-west4,europe-west4-b
+ tpu-v5p-6912,1,europe-west4,europe-west4-b
+ tpu-v5p-7040,1,europe-west4,europe-west4-b
+ tpu-v5p-7168,1,europe-west4,europe-west4-b
+ tpu-v5p-7296,1,europe-west4,europe-west4-b
+ tpu-v5p-7424,1,europe-west4,europe-west4-b
+ tpu-v5p-7552,1,europe-west4,europe-west4-b
+ tpu-v5p-7680,1,europe-west4,europe-west4-b
+ tpu-v5p-7808,1,europe-west4,europe-west4-b
+ tpu-v5p-7936,1,europe-west4,europe-west4-b
+ tpu-v5p-8064,1,europe-west4,europe-west4-b
+ tpu-v5p-8192,1,europe-west4,europe-west4-b
+ tpu-v5p-8320,1,europe-west4,europe-west4-b
+ tpu-v5p-8448,1,europe-west4,europe-west4-b
+ tpu-v5p-8704,1,europe-west4,europe-west4-b
+ tpu-v5p-8832,1,europe-west4,europe-west4-b
+ tpu-v5p-8960,1,europe-west4,europe-west4-b
+ tpu-v5p-9216,1,europe-west4,europe-west4-b
+ tpu-v5p-9472,1,europe-west4,europe-west4-b
+ tpu-v5p-9600,1,europe-west4,europe-west4-b
+ tpu-v5p-9728,1,europe-west4,europe-west4-b
+ tpu-v5p-9856,1,europe-west4,europe-west4-b
+ tpu-v5p-9984,1,europe-west4,europe-west4-b
+ tpu-v5p-10240,1,europe-west4,europe-west4-b
+ tpu-v5p-10368,1,europe-west4,europe-west4-b
+ tpu-v5p-10496,1,europe-west4,europe-west4-b
+ tpu-v5p-10752,1,europe-west4,europe-west4-b
+ tpu-v5p-10880,1,europe-west4,europe-west4-b
+ tpu-v5p-11008,1,europe-west4,europe-west4-b
+ tpu-v5p-11136,1,europe-west4,europe-west4-b
+ tpu-v5p-11264,1,europe-west4,europe-west4-b
+ tpu-v5p-11520,1,europe-west4,europe-west4-b
+ tpu-v5p-11648,1,europe-west4,europe-west4-b
+ tpu-v5p-11776,1,europe-west4,europe-west4-b
+ tpu-v5p-11904,1,europe-west4,europe-west4-b
+ tpu-v5p-12032,1,europe-west4,europe-west4-b
+ tpu-v5p-12160,1,europe-west4,europe-west4-b
+ tpu-v5p-12288,1,europe-west4,europe-west4-b
+ """)))
+}
 # FIXME(woosuk): Remove this once the bug is fixed.
 # See https://github.com/skypilot-org/skypilot/issues/1759#issue-1619614345
 TPU_V4_HOST_DF = pd.read_csv(
@@ -169,7 +281,7 @@ def filter_zones(func: Callable[[], List[str]]) -> Callable[[], List[str]]:
 
 
 @filter_zones
-@functools.lru_cache(maxsize=None)
+@annotations.lru_cache(scope='global', maxsize=None)
 def _get_all_zones() -> List[str]:
     zones_request = gcp_client.zones().list(project=project_id)
     zones = []
@@ -225,7 +337,7 @@ def get_vm_df(skus: List[Dict[str, Any]], region_prefix: str) -> 'pd.DataFrame':
     df = df[~df['AvailabilityZone'].str.startswith(tuple(TPU_V4_ZONES))]
 
     # TODO(woosuk): Make this more efficient.
-    def get_vm_price(row: pd.Series, spot: bool) -> float:
+    def get_vm_price(row: pd.Series, spot: bool) -> Optional[float]:
         series = row['InstanceType'].split('-')[0].lower()
 
         ondemand_or_spot = 'OnDemand' if not spot else 'Preemptible'
@@ -276,12 +388,26 @@ def get_vm_df(skus: List[Dict[str, Any]], region_prefix: str) -> 'pd.DataFrame':
         if series in ['f1', 'g1']:
             memory_price = 0.0
 
-        assert cpu_price is not None, row
-        assert memory_price is not None, row
+        # TODO(tian): (2024/11/10) Some SKUs are missing in the SKUs API. We
+        # skip them in the catalog for now. We should investigate why they are
+        # missing and add them back.
+        if cpu_price is None or memory_price is None:
+            return None
         return cpu_price + memory_price
 
     df['Price'] = df.apply(lambda row: get_vm_price(row, spot=False), axis=1)
     df['SpotPrice'] = df.apply(lambda row: get_vm_price(row, spot=True), axis=1)
+    dropped_rows = df[df['Price'].isna() & df['SpotPrice'].isna()]
+    dropped_info = (dropped_rows[['InstanceType',
+                                  'AvailabilityZone']].drop_duplicates())
+    az2missing = dropped_info.groupby('AvailabilityZone').apply(
+        lambda x: x['InstanceType'].tolist())
+    print('Price not found for the following zones and instance types. '
+          'Dropping them.')
+    for az, instances in az2missing.items():
+        print('-' * 30, az, '-' * 30)
+        print(', '.join(instances))
+    df = df.dropna(subset=['Price', 'SpotPrice'], how='all')
     df = df.reset_index(drop=True)
     df = df.sort_values(['InstanceType', 'Region', 'AvailabilityZone'])
     return df
@@ -307,8 +433,10 @@ def _get_gpus_for_zone(zone: str) -> 'pd.DataFrame':
             gpu_name = gpu_name.upper()
             if 'H100-80GB' in gpu_name:
                 gpu_name = 'H100'
+            if 'H100-MEGA-80GB' in gpu_name:
+                gpu_name = 'H100-MEGA'
                 if count != 8:
-                    # H100 only has 8 cards.
+                    # H100-MEGA only has 8 cards.
                     continue
             if 'VWS' in gpu_name:
                 continue
@@ -338,6 +466,7 @@ def _gpu_info_from_name(name: str) -> Optional[Dict[str, List[Dict[str, Any]]]]:
         'A100-80GB': 80 * 1024,
         'A100': 40 * 1024,
         'H100': 80 * 1024,
+        'H100-MEGA': 80 * 1024,
         'P4': 8 * 1024,
         'T4': 16 * 1024,
         'V100': 16 * 1024,
@@ -382,12 +511,17 @@ def get_gpu_df(skus: List[Dict[str, Any]],
             if sku['category']['usageType'] != ondemand_or_spot:
                 continue
 
-            gpu_name = row['AcceleratorName']
-            if gpu_name == 'A100-80GB':
-                gpu_name = 'A100 80GB'
-            if gpu_name == 'H100':
-                gpu_name = 'H100 80GB'
-            if f'{gpu_name} GPU' not in sku['description']:
+            gpu_names = [row['AcceleratorName']]
+            if gpu_names[0] == 'A100-80GB':
+                gpu_names = ['A100 80GB']
+            if gpu_names[0] == 'H100':
+                gpu_names = ['H100 80GB']
+            if gpu_names[0] == 'H100-MEGA':
+                # Seems that H100-MEGA has two different descriptions in SKUs in
+                # different regions: 'H100 80GB Mega' and 'H100 80GB Plus'.
+                gpu_names = ['H100 80GB Mega', 'H100 80GB Plus']
+            if not any(f'{gpu_name} GPU' in sku['description']
+                       for gpu_name in gpu_names):
                 continue
 
             unit_price = _get_unit_price(sku)
@@ -414,34 +548,55 @@ def get_gpu_df(skus: List[Dict[str, Any]],
     return df
 
 
+def _get_tpu_response_for_zone(zone: str) -> list:
+    parent = f'projects/{project_id}/locations/{zone}'
+    # Sometimes the response is empty ({}) even for enabled zones. Here we
+    # retry the request for a few times.
+    backoff = common_utils.Backoff(initial_backoff=1)
+    for _ in range(TPU_RETRY_CNT):
+        tpus_request = (
+            tpu_client.projects().locations().acceleratorTypes().list(
+                parent=parent))
+        try:
+            tpus_response = tpus_request.execute()
+            if 'acceleratorTypes' in tpus_response:
+                return tpus_response['acceleratorTypes']
+        except gcp.http_error_exception() as error:
+            if error.resp.status == 403:
+                print('  TPU API is not enabled or you don\'t have TPU access '
+                      f'to zone: {zone!r}.')
+            else:
+                print(f'  An error occurred: {error}')
+            # If error happens, fail early.
+            return []
+        time_to_sleep = backoff.current_backoff()
+        print(f'  Retry zone {zone!r} in {time_to_sleep} seconds...')
+        time.sleep(time_to_sleep)
+    print(f'ERROR: Failed to fetch TPUs for zone {zone!r}.')
+    return []
+
+
 def _get_tpu_for_zone(zone: str) -> 'pd.DataFrame':
+    # Use hardcoded TPU V5 data as it is invisible in some zones.
+    missing_tpus_df = pd.DataFrame(columns=[
+        'AcceleratorName', 'AcceleratorCount', 'Region', 'AvailabilityZone'
+    ])
+    if zone in TPU_V5_MISSING_ZONES_DF:
+        missing_tpus_df = TPU_V5_MISSING_ZONES_DF[zone]
     tpus = []
-    parent = f'projects/{project_id}/locations/{zone}'
-    tpus_request = tpu_client.projects().locations().acceleratorTypes().list(
-        parent=parent)
-    try:
-        tpus_response = tpus_request.execute()
-        for tpu in tpus_response['acceleratorTypes']:
-            tpus.append(tpu)
-    except gcp.http_error_exception() as error:
-        if error.resp.status == 403:
-            print('  TPU API is not enabled or you don\'t have TPU access '
-                  f'to zone: {zone!r}.')
-        else:
-            print(f'  An error occurred: {error}')
+    for tpu in _get_tpu_response_for_zone(zone):
+        tpus.append(tpu)
     new_tpus = []
     for tpu in tpus:
         tpu_name = tpu['type']
-        # skip tpu v5 as we currently don't support it
-        if 'v5' in tpu_name:
-            continue
         new_tpus.append({
             'AcceleratorName': f'tpu-{tpu_name}',
             'AcceleratorCount': 1,
             'Region': zone.rpartition('-')[0],
             'AvailabilityZone': zone,
         })
-    return pd.DataFrame(new_tpus).reset_index(drop=True)
+    new_tpu_df = pd.DataFrame(new_tpus).reset_index(drop=True)
+    return pd.concat([new_tpu_df, missing_tpus_df])
 
 
 def _get_tpus() -> 'pd.DataFrame':
@@ -458,11 +613,24 @@ def _get_tpus() -> 'pd.DataFrame':
 
 
 # TODO: the TPUs fetched fails to contain us-east1
-def get_tpu_df(skus: List[Dict[str, Any]]) -> 'pd.DataFrame':
+def get_tpu_df(gce_skus: List[Dict[str, Any]],
+               tpu_skus: List[Dict[str, Any]]) -> 'pd.DataFrame':
     df = _get_tpus()
     if df.empty:
         return df
 
+    def _get_tpu_description_str(tpu_version: str) -> str:
+        # TPU V5 has a different naming convention since it is contained in
+        # the GCE SKUs. v5p -> TpuV5p, v5litepod -> TpuV5e.
+        if tpu_version.startswith('v5'):
+            if tpu_version == 'v5p':
+                return 'TpuV5p'
+            assert tpu_version == 'v5litepod', tpu_version
+            return 'TpuV5e'
+        if tpu_version.startswith('v6e'):
+            return 'TpuV6e'
+        return f'Tpu-{tpu_version}'
+
     def get_tpu_price(row: pd.Series, spot: bool) -> Optional[float]:
         assert row['AcceleratorCount'] == 1, row
         tpu_price = None
@@ -475,9 +643,12 @@ def get_tpu_df(skus: List[Dict[str, Any]]) -> 'pd.DataFrame':
         # whether the TPU is a single device or a pod.
         # For TPU-v4, the pricing is uniform, and thus the pricing API
         # only provides the price of TPU-v4 pods.
-        is_pod = num_cores > 8 or tpu_version == 'v4'
+        # The price shown for v5 & v6e TPU is per chip hour, so there is
+        # no 'Pod' keyword in the description.
+        is_pod = ((num_cores > 8 or tpu_version == 'v4') and
+                  not tpu_version.startswith('v5') and tpu_version != 'v6e')
 
-        for sku in skus:
+        for sku in gce_skus + tpu_skus:
             if tpu_region not in sku['serviceRegions']:
                 continue
             description = sku['description']
@@ -489,7 +660,7 @@ def get_tpu_df(skus: List[Dict[str, Any]]) -> 'pd.DataFrame':
                 if 'Preemptible' in description:
                     continue
 
-            if f'Tpu-{tpu_version}' not in description:
+            if _get_tpu_description_str(tpu_version) not in description:
                 continue
             if is_pod:
                 if 'Pod' not in description:
@@ -500,7 +671,17 @@ def get_tpu_df(skus: List[Dict[str, Any]]) -> 'pd.DataFrame':
 
             unit_price = _get_unit_price(sku)
             tpu_device_price = unit_price
-            tpu_core_price = tpu_device_price / 8
+            # v5p naming convention is v$VERSION_NUMBERp-$CORES_COUNT, while
+            # v5e is v$VERSION_NUMBER-$CHIP_COUNT. In the same time, V5 price
+            # is shown as per chip price, which is 2 cores for v5p and 1 core
+            # for v5e. Reference here:
+            # https://cloud.google.com/tpu/docs/v5p#using-accelerator-type
+            # https://cloud.google.com/tpu/docs/v5e#tpu-v5e-config
+            # v6e is also per chip price. Reference here:
+            # https://cloud.google.com/tpu/docs/v6e#configurations
+            core_per_sku = (1 if tpu_version in ['v5litepod', 'v6e'] else
+                            2 if tpu_version == 'v5p' else 8)
+            tpu_core_price = tpu_device_price / core_per_sku
             tpu_price = num_cores * tpu_core_price
             break
 
@@ -518,7 +699,13 @@ def get_tpu_df(skus: List[Dict[str, Any]]) -> 'pd.DataFrame':
             spot_str = 'spot ' if spot else ''
             print(f'The {spot_str}price of {tpu_name} in {tpu_region} is '
                   'not found in SKUs or hidden TPU price DF.')
-        assert spot or tpu_price is not None, (row, hidden_tpu, HIDDEN_TPU_DF)
+        if (tpu_name.startswith('tpu-v6e') and
+                tpu_region in TPU_V6E_MISSING_REGIONS):
+            if not spot:
+                tpu_price = 0.0
+        else:
+            assert spot or tpu_price is not None, (row, hidden_tpu,
+                                                   HIDDEN_TPU_DF)
         return tpu_price
 
     df['Price'] = df.apply(lambda row: get_tpu_price(row, spot=False), axis=1)
@@ -546,7 +733,8 @@ def get_catalog_df(region_prefix: str) -> 'pd.DataFrame':
         region_prefix)] if not gpu_df.empty else gpu_df
 
     gcp_tpu_skus = get_skus(TPU_SERVICE_ID)
-    tpu_df = get_tpu_df(gcp_tpu_skus)
+    # TPU V5 SKU is not included in the TPU SKUs but in the GCE SKUs.
+    tpu_df = get_tpu_df(gcp_skus, gcp_tpu_skus)
 
     # Merge the dataframes.
     df = pd.concat([vm_df, gpu_df, tpu_df, TPU_V4_HOST_DF])

sky/clouds/service_catalog/data_fetchers/fetch_lambda_cloud.py

@@ -11,6 +11,7 @@ import argparse
 import csv
 import json
 import os
+from typing import Optional, Tuple
 
 import requests
 
@@ -19,17 +20,21 @@ DEFAULT_LAMBDA_KEYS_PATH = os.path.expanduser('~/.lambda_cloud/lambda_keys')
 
 # List of all possible regions.
 REGIONS = [
-    'australia-southeast-1',
     'europe-central-1',
     'asia-south-1',
     'me-west-1',
     'europe-south-1',
     'asia-northeast-1',
     'asia-northeast-2',
+    'australia-east-1',
     'us-east-1',
+    'us-east-2',
+    'us-east-3',
     'us-west-2',
     'us-west-1',
     'us-south-1',
+    'us-south-2',
+    'us-south-3',
     'us-west-3',
     'us-midwest-1',
 ]
@@ -43,18 +48,25 @@ GPU_TO_MEMORY = {
     'RTX6000': 24576,
     'V100': 16384,
     'H100': 81920,
+    'GH200': 98304,
+    'GENERAL': None
 }
 
 
-def name_to_gpu(name: str) -> str:
+def name_to_gpu_and_cnt(name: str) -> Optional[Tuple[str, int]]:
+    """Extract GPU and count from instance type name.
+
+    The instance type name is in the format:
+      'gpu_{gpu_count}x_{gpu_name}_<suffix>'.
+    """
     # Edge case
     if name == 'gpu_8x_a100_80gb_sxm4':
-        return 'A100-80GB'
-    return name.split('_')[2].upper()
-
-
-def name_to_gpu_cnt(name: str) -> int:
-    return int(name.split('_')[1].replace('x', ''))
+        return 'A100-80GB', 8
+    gpu = name.split('_')[2].upper()
+    if gpu == 'GENERAL':
+        return None
+    gpu_cnt = int(name.split('_')[1].replace('x', ''))
+    return gpu, gpu_cnt
 
 
 def create_catalog(api_key: str, output_path: str) -> None:
@@ -71,24 +83,32 @@ def create_catalog(api_key: str, output_path: str) -> None:
         # We parse info.keys() in reverse order so gpu_1x_a100_sxm4 comes before
         # gpu_1x_a100 in the catalog (gpu_1x_a100_sxm4 has more availability).
         for vm in reversed(list(info.keys())):
-            gpu = name_to_gpu(vm)
-            gpu_cnt = float(name_to_gpu_cnt(vm))
+            gpu_and_cnt = name_to_gpu_and_cnt(vm)
+            gpu: Optional[str]
+            gpu_cnt: Optional[float]
+            if gpu_and_cnt is None:
+                gpu, gpu_cnt = None, None
+            else:
+                gpu = gpu_and_cnt[0]
+                gpu_cnt = float(gpu_and_cnt[1])
             vcpus = float(info[vm]['instance_type']['specs']['vcpus'])
             mem = float(info[vm]['instance_type']['specs']['memory_gib'])
-            price = float(info[vm]['instance_type']\
-                    ['price_cents_per_hour']) / 100
-            gpuinfo = {
-                'Gpus': [{
-                    'Name': gpu,
-                    'Manufacturer': 'NVIDIA',
-                    'Count': gpu_cnt,
-                    'MemoryInfo': {
-                        'SizeInMiB': GPU_TO_MEMORY[gpu]
-                    },
-                }],
-                'TotalGpuMemoryInMiB': GPU_TO_MEMORY[gpu]
-            }
-            gpuinfo = json.dumps(gpuinfo).replace('"', "'")  # pylint: disable=invalid-string-quote
+            price = (float(info[vm]['instance_type']['price_cents_per_hour']) /
+                     100)
+            gpuinfo: Optional[str] = None
+            if gpu is not None:
+                gpuinfo_dict = {
+                    'Gpus': [{
+                        'Name': gpu,
+                        'Manufacturer': 'NVIDIA',
+                        'Count': gpu_cnt,
+                        'MemoryInfo': {
+                            'SizeInMiB': GPU_TO_MEMORY[gpu]
+                        },
+                    }],
+                    'TotalGpuMemoryInMiB': GPU_TO_MEMORY[gpu]
+                }
+                gpuinfo = json.dumps(gpuinfo_dict).replace('"', "'")  # pylint: disable=invalid-string-quote
             for r in REGIONS:
                 writer.writerow(
                     [vm, gpu, gpu_cnt, vcpus, mem, price, r, gpuinfo, ''])

sky/clouds/service_catalog/data_fetchers/fetch_vast.py

@@ -0,0 +1,147 @@
+"""A script that generates the Vast Cloud catalog. """
+
+#
+# Due to the design of the sdk, pylint has a false
+# positive for the fnctions.
+#
+# pylint: disable=assignment-from-no-return
+import collections
+import csv
+import json
+import math
+import re
+import sys
+from typing import Any, Dict, List
+
+from sky.adaptors import vast
+
+_map = {
+    'TeslaV100': 'V100',
+    'TeslaT4': 'T4',
+    'TeslaP100': 'P100',
+    'QRTX6000': 'RTX6000',
+    'QRTX8000': 'RTX8000'
+}
+
+
+def create_instance_type(obj: Dict[str, Any]) -> str:
+    stubify = lambda x: re.sub(r'\s', '_', x)
+    return '{}x-{}-{}-{}'.format(obj['num_gpus'], stubify(obj['gpu_name']),
+                                 obj['cpu_cores'], obj['cpu_ram'])
+
+
+def dot_get(d: dict, key: str) -> Any:
+    for k in key.split('.'):
+        d = d[k]
+    return d
+
+
+if __name__ == '__main__':
+    seen = set()
+    # InstanceType and gpuInfo are basically just stubs
+    # so that the dictwriter is happy without weird
+    # code.
+    mapped_keys = (('gpu_name', 'InstanceType'), ('gpu_name',
+                                                  'AcceleratorName'),
+                   ('num_gpus', 'AcceleratorCount'), ('cpu_cores', 'vCPUs'),
+                   ('cpu_ram', 'MemoryGiB'), ('gpu_name', 'GpuInfo'),
+                   ('search.totalHour', 'Price'), ('min_bid', 'SpotPrice'),
+                   ('geolocation', 'Region'))
+    writer = csv.DictWriter(sys.stdout, fieldnames=[x[1] for x in mapped_keys])
+    writer.writeheader()
+
+    # Vast has a wide variety of machines, some of
+    # which will have less diskspace and network
+    # bandwidth than others.
+    #
+    # The machine normally have high specificity
+    # in the vast catalog - this is fairly unique
+    # to Vast and can make bucketing them into
+    # instance types difficult.
+    #
+    # The flags
+    #
+    #   * georegion consolidates geographic areas
+    #
+    #   * chunked rounds down specifications (such
+    #     as 1025GB to 1024GB disk) in order to
+    #     make machine specifications look more
+    #     consistent
+    #
+    #   * inet_down makes sure that only machines
+    #     with "reasonable" downlink speed are
+    #     considered
+    #
+    #   * disk_space sets a lower limit of how
+    #     much space is availble to be allocated
+    #     in order to ensure that machines with
+    #     small disk pools aren't listed
+    #
+    offerList = vast.vast().search_offers(
+        query=('georegion = true chunked = true '
+               'inet_down >= 100 disk_space >= 80'),
+        limit=10000)
+
+    priceMap: Dict[str, List] = collections.defaultdict(list)
+    for offer in offerList:
+        entry = {}
+        for ours, theirs in mapped_keys:
+            field = dot_get(offer, ours)
+            entry[theirs] = field
+
+        instance_type = create_instance_type(offer)
+        entry['InstanceType'] = instance_type
+
+        # the documentation says
+        # "{'gpus': [{
+        #   'name': 'v100',
+        #   'manufacturer': 'nvidia',
+        #   'count': 8.0,
+        #   'memoryinfo': {'sizeinmib': 16384}
+        #   }],
+        #   'totalgpumemoryinmib': 16384}",
+        # we can do that.
+        entry['MemoryGiB'] /= 1024
+
+        gpu = re.sub('Ada', '-Ada', re.sub(r'\s', '', offer['gpu_name']))
+        gpu = re.sub(r'(Ti|PCIE|SXM4|SXM|NVL)$', '', gpu)
+        gpu = re.sub(r'(RTX\d0\d0)(S|D)$', r'\1', gpu)
+
+        if gpu in _map:
+            gpu = _map[gpu]
+
+        entry['AcceleratorName'] = gpu
+        entry['GpuInfo'] = json.dumps({
+            'Gpus': [{
+                'Name': gpu,
+                'Count': offer['num_gpus'],
+                'MemoryInfo': {
+                    'SizeInMiB': offer['gpu_total_ram']
+                }
+            }],
+            'TotalGpuMemoryInMiB': offer['gpu_total_ram']
+        }).replace('"', '\'')
+
+        priceMap[instance_type].append(entry)
+
+    for instanceList in priceMap.values():
+        priceList = sorted([x['Price'] for x in instanceList])
+        index = math.ceil(0.5 * len(priceList)) - 1
+        priceTarget = priceList[index]
+        toList: List = []
+        for instance in instanceList:
+            if instance['Price'] <= priceTarget:
+                instance['Price'] = '{:.2f}'.format(priceTarget)
+                toList.append(instance)
+
+        maxBid = max([x.get('SpotPrice') for x in toList])
+        for instance in toList:
+            stub = f'{instance["InstanceType"]} {instance["Region"][-2:]}'
+            if stub in seen:
+                printstub = f'{stub}#print'
+                if printstub not in seen:
+                    instance['SpotPrice'] = f'{maxBid:.2f}'
+                    writer.writerow(instance)
+                    seen.add(printstub)
+            else:
+                seen.add(stub)

sky/clouds/service_catalog/data_fetchers/fetch_vsphere.py

@@ -534,7 +534,7 @@ def initialize_images_csv(csv_saving_path: str, vc_object,
                         gpu_name = tag_name.split('-')[1]
                         if gpu_name not in gpu_tags:
                             gpu_tags.append(gpu_name)
-                if len(gpu_tags) > 0:
+                if gpu_tags:
                     gpu_tags_str = str(gpu_tags).replace('\'', '\"')
                     f.write(f'{item.id},{vcenter_name},{item_cpu},{item_memory}'
                             f',,,\'{gpu_tags_str}\'\n')