[UX/Catalog] Add DEVICE_MEM info to GCP GPUs.

[concretevitamin]

Previously, DEVICE_MEM is missing in sky show-gpus GCP results. This was because GCP APIs didn't explicitly return such info.

Now:

» sky show-gpus L4:1 --cloud gcp
GPU  QTY  CLOUD  INSTANCE_TYPE  DEVICE_MEM  vCPUs  HOST_MEM  HOURLY_PRICE  HOURLY_SPOT_PRICE  REGION
L4   1    GCP    g2-standard-4  24GB        4      16GB      $ 0.705       $ 0.248            us-east4

» sky show-gpus A100:1 --cloud gcp
GPU   QTY  CLOUD  INSTANCE_TYPE  DEVICE_MEM  vCPUs  HOST_MEM  HOURLY_PRICE  HOURLY_SPOT_PRICE  REGION
A100  1    GCP    a2-highgpu-1g  40GB        12     85GB      $ 3.673       $ 1.469            us-central1

GPU        QTY  CLOUD  INSTANCE_TYPE   DEVICE_MEM  vCPUs  HOST_MEM  HOURLY_PRICE  HOURLY_SPOT_PRICE  REGION   
A100-80GB  1    GCP    a2-ultragpu-1g  80GB        12     170GB     $ 5.028       $ 2.011            us-central1

Tested (run the relevant ones):

• Code formatting: bash format.sh
• Any manual or new tests for this PR (please specify below)
• All smoke tests: pytest tests/test_smoke.py
• Relevant individual smoke tests: pytest tests/test_smoke.py::test_fill_in_the_name
• Backward compatibility tests: bash tests/backward_comaptibility_tests.sh

[romilbhardwaj]

Thanks @concretevitamin!

[romilbhardwaj]

Should we add P100 too?

[romilbhardwaj]

BTW, generating a catalog from this branch removed GCP P100 from my sky show-gpus P100:

# Master catalog
(base) ➜  ~ sky show-gpus P100
GPU   QTY  CLOUD  INSTANCE_TYPE       DEVICE_MEM  vCPUs  HOST_MEM  HOURLY_PRICE  HOURLY_SPOT_PRICE  REGION
P100  1    Azure  Standard_NC6s_v2    -           6      112GB     $ 2.070       $ 0.207            eastus
P100  2    Azure  Standard_NC12s_v2   -           12     224GB     $ 4.140       $ 0.414            eastus
P100  4    Azure  Standard_NC24rs_v2  -           24     448GB     $ 9.108       $ 0.911            eastus
P100  4    Azure  Standard_NC24s_v2   -           24     448GB     $ 8.280       $ 0.828            eastus
P100  1    GCP    n1-highmem-8        -           8      52GB      $ 1.933       $ 0.679            us-central1
P100  2    GCP    n1-highmem-16       -           16     104GB     $ 3.866       $ 1.357            us-central1
P100  4    GCP    n1-highmem-32       -           32     208GB     $ 7.733       $ 2.714            us-central1
P100  1    OCI    VM.GPU2.1           16GB        24     72GB      $ 1.275       -                  eu-amsterdam-1
P100  2    OCI    BM.GPU2.2           16GB        56     256GB     $ 2.550       -                  eu-amsterdam-1

# Catalog from this branch
(base) ➜  ~ sky show-gpus P100
GPU   QTY  CLOUD  INSTANCE_TYPE       DEVICE_MEM  vCPUs  HOST_MEM  HOURLY_PRICE  HOURLY_SPOT_PRICE  REGION
P100  1    Azure  Standard_NC6s_v2    -           6      112GB     $ 2.070       $ 0.207            eastus
P100  2    Azure  Standard_NC12s_v2   -           12     224GB     $ 4.140       $ 0.414            eastus
P100  4    Azure  Standard_NC24rs_v2  -           24     448GB     $ 9.108       $ 0.911            eastus
P100  4    Azure  Standard_NC24s_v2   -           24     448GB     $ 8.280       $ 0.828            eastus
P100  1    OCI    VM.GPU2.1           16GB        24     72GB      $ 1.275       -                  eu-amsterdam-1
P100  2    OCI    BM.GPU2.2           16GB        56     256GB     $ 2.550       -                  eu-amsterdam-1

More generally, does this change handle any new GPUs that may get added but are not updated in the name_to_gpu_memory_in_mib dict?

Here's the catalog that got generated for me for reference.

[concretevitamin]

Great catch! Fixed both P100 and missing-GPU-info-not-shown issue (tested with this repro).

[romilbhardwaj]

Thanks @concretevitamin! LGTM.