feat: combine ProviderSpec datatypes

currently `RemoteProviderSpec` has an `AdapterSpec` embedded in it. Remove `AdapterSpec`, and put its leftover fields into `RemoteProviderSpec`.

Additionally, many of the fields were duplicated between `InlineProviderSpec` and `RemoteProviderSpec`. Move these to `ProviderSpec` so they are shared.

Fixup the distro codegen to use `RemoteProviderSpec` directly rather than `remote_provider_spec` which took an AdapterSpec and returned a full provider spec

Signed-off-by: Charlie Doern <cdoern@redhat.com>
This commit is contained in:
Charlie Doern 2025-09-08 15:51:53 -04:00
parent e66103c09d
commit 686f87d138
15 changed files with 381 additions and 503 deletions

View file

@ -7,7 +7,7 @@
from typing import cast
from llama_stack.providers.datatypes import AdapterSpec, Api, InlineProviderSpec, ProviderSpec, remote_provider_spec
from llama_stack.providers.datatypes import Api, InlineProviderSpec, ProviderSpec, RemoteProviderSpec
# We provide two versions of these providers so that distributions can package the appropriate version of torch.
# The CPU version is used for distributions that don't have GPU support -- they result in smaller container images.
@ -57,14 +57,13 @@ def available_providers() -> list[ProviderSpec]:
],
description="HuggingFace-based post-training provider for fine-tuning models using the HuggingFace ecosystem.",
),
remote_provider_spec(
RemoteProviderSpec(
api=Api.post_training,
adapter=AdapterSpec(
adapter_type="nvidia",
pip_packages=["requests", "aiohttp"],
module="llama_stack.providers.remote.post_training.nvidia",
config_class="llama_stack.providers.remote.post_training.nvidia.NvidiaPostTrainingConfig",
description="NVIDIA's post-training provider for fine-tuning models on NVIDIA's platform.",
),
adapter_type="nvidia",
provider_type="remote::nvidia",
pip_packages=["requests", "aiohttp"],
module="llama_stack.providers.remote.post_training.nvidia",
config_class="llama_stack.providers.remote.post_training.nvidia.NvidiaPostTrainingConfig",
description="NVIDIA's post-training provider for fine-tuning models on NVIDIA's platform.",
),
]