forked from phoenix-oss/llama-stack-mirror
Auto-generate distro yamls + docs (#468)
# What does this PR do? Automatically generates - build.yaml - run.yaml - run-with-safety.yaml - parts of markdown docs for the distributions. ## Test Plan At this point, this only updates the YAMLs and the docs. Some testing (especially with ollama and vllm) has been performed but needs to be much more tested.
This commit is contained in:
Ashwin Bharambe
GitHub
parent
0784284ab5
commit
2a31163178
88 changed files with 3008 additions and 852 deletions
7
llama_stack/templates/remote-vllm/__init__.py
Normal file
7
llama_stack/templates/remote-vllm/__init__.py
Normal file
|
|
@ -0,0 +1,7 @@
|
|||
# Copyright (c) Meta Platforms, Inc. and affiliates.
|
||||
# All rights reserved.
|
||||
#
|
||||
# This source code is licensed under the terms described in the LICENSE file in
|
||||
# the root directory of this source tree.
|
||||
|
||||
from .vllm import get_distribution_template # noqa: F401
|
||||
|
|
@ -1,12 +1,19 @@
|
|||
version: '2'
|
||||
name: remote-vllm
|
||||
distribution_spec:
|
||||
description: Use (an external) vLLM server for running LLM inference
|
||||
docker_image: null
|
||||
providers:
|
||||
inference: remote::vllm
|
||||
inference:
|
||||
- remote::vllm
|
||||
memory:
|
||||
- inline::faiss
|
||||
- remote::chromadb
|
||||
- remote::pgvector
|
||||
safety: inline::llama-guard
|
||||
agents: inline::meta-reference
|
||||
telemetry: inline::meta-reference
|
||||
safety:
|
||||
- inline::llama-guard
|
||||
agents:
|
||||
- inline::meta-reference
|
||||
telemetry:
|
||||
- inline::meta-reference
|
||||
image_type: conda
|
||||
|
|
|
|||
119
llama_stack/templates/remote-vllm/doc_template.md
Normal file
119
llama_stack/templates/remote-vllm/doc_template.md
Normal file
|
|
@ -0,0 +1,119 @@
|
|||
# Remote vLLM Distribution
|
||||
|
||||
The `llamastack/distribution-{{ name }}` distribution consists of the following provider configurations:
|
||||
|
||||
{{ providers_table }}
|
||||
|
||||
You can use this distribution if you have GPUs and want to run an independent vLLM server container for running inference.
|
||||
|
||||
{% if run_config_env_vars %}
|
||||
### Environment Variables
|
||||
|
||||
The following environment variables can be configured:
|
||||
|
||||
{% for var, (default_value, description) in run_config_env_vars.items() %}
|
||||
- `{{ var }}`: {{ description }} (default: `{{ default_value }}`)
|
||||
{% endfor %}
|
||||
{% endif %}
|
||||
|
||||
|
||||
## Setting up vLLM server
|
||||
|
||||
Please check the [vLLM Documentation](https://docs.vllm.ai/en/v0.5.5/serving/deploying_with_docker.html) to get a vLLM endpoint. Here is a sample script to start a vLLM server locally via Docker:
|
||||
|
||||
```bash
|
||||
export INFERENCE_PORT=8000
|
||||
export INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct
|
||||
export CUDA_VISIBLE_DEVICES=0
|
||||
|
||||
docker run \
|
||||
--runtime nvidia \
|
||||
--gpus $CUDA_VISIBLE_DEVICES \
|
||||
-v ~/.cache/huggingface:/root/.cache/huggingface \
|
||||
--env "HUGGING_FACE_HUB_TOKEN=$HF_TOKEN" \
|
||||
-p $INFERENCE_PORT:$INFERENCE_PORT \
|
||||
--ipc=host \
|
||||
vllm/vllm-openai:latest \
|
||||
--model $INFERENCE_MODEL \
|
||||
--port $INFERENCE_PORT
|
||||
```
|
||||
|
||||
If you are using Llama Stack Safety / Shield APIs, then you will need to also run another instance of a vLLM with a corresponding safety model like `meta-llama/Llama-Guard-3-1B` using a script like:
|
||||
|
||||
```bash
|
||||
export SAFETY_PORT=8081
|
||||
export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
|
||||
export CUDA_VISIBLE_DEVICES=1
|
||||
|
||||
docker run \
|
||||
--runtime nvidia \
|
||||
--gpus $CUDA_VISIBLE_DEVICES \
|
||||
-v ~/.cache/huggingface:/root/.cache/huggingface \
|
||||
--env "HUGGING_FACE_HUB_TOKEN=$HF_TOKEN" \
|
||||
-p $SAFETY_PORT:$SAFETY_PORT \
|
||||
--ipc=host \
|
||||
vllm/vllm-openai:latest \
|
||||
--model $SAFETY_MODEL \
|
||||
--port $SAFETY_PORT
|
||||
```
|
||||
|
||||
## Running Llama Stack
|
||||
|
||||
Now you are ready to run Llama Stack with vLLM as the inference provider. You can do this via Conda (build code) or Docker which has a pre-built image.
|
||||
|
||||
### Via Docker
|
||||
|
||||
This method allows you to get started quickly without having to build the distribution code.
|
||||
|
||||
```bash
|
||||
LLAMA_STACK_PORT=5001
|
||||
docker run \
|
||||
-it \
|
||||
-p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
|
||||
-v ./run.yaml:/root/my-run.yaml \
|
||||
llamastack/distribution-{{ name }} \
|
||||
/root/my-run.yaml \
|
||||
--port $LLAMA_STACK_PORT \
|
||||
--env INFERENCE_MODEL=$INFERENCE_MODEL \
|
||||
--env VLLM_URL=http://host.docker.internal:$INFERENCE_PORT \
|
||||
```
|
||||
|
||||
If you are using Llama Stack Safety / Shield APIs, use:
|
||||
|
||||
```bash
|
||||
docker run \
|
||||
-it \
|
||||
-p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
|
||||
-v ./run-with-safety.yaml:/root/my-run.yaml \
|
||||
llamastack/distribution-{{ name }} \
|
||||
/root/my-run.yaml \
|
||||
--port $LLAMA_STACK_PORT \
|
||||
--env INFERENCE_MODEL=$INFERENCE_MODEL \
|
||||
--env VLLM_URL=http://host.docker.internal:$INFERENCE_PORT \
|
||||
--env SAFETY_MODEL=$SAFETY_MODEL \
|
||||
--env VLLM_SAFETY_URL=http://host.docker.internal:$SAFETY_PORT
|
||||
```
|
||||
|
||||
|
||||
### Via Conda
|
||||
|
||||
Make sure you have done `pip install llama-stack` and have the Llama Stack CLI available.
|
||||
|
||||
```bash
|
||||
llama stack build --template remote-vllm --image-type conda
|
||||
llama stack run ./run.yaml \
|
||||
--port 5001 \
|
||||
--env INFERENCE_MODEL=$INFERENCE_MODEL \
|
||||
--env VLLM_URL=http://127.0.0.1:$INFERENCE_PORT
|
||||
```
|
||||
|
||||
If you are using Llama Stack Safety / Shield APIs, use:
|
||||
|
||||
```bash
|
||||
llama stack run ./run-with-safety.yaml \
|
||||
--port 5001 \
|
||||
--env INFERENCE_MODEL=$INFERENCE_MODEL \
|
||||
--env VLLM_URL=http://127.0.0.1:$INFERENCE_PORT \
|
||||
--env SAFETY_MODEL=$SAFETY_MODEL \
|
||||
--env VLLM_SAFETY_URL=http://127.0.0.1:$SAFETY_PORT
|
||||
```
|
||||
70
llama_stack/templates/remote-vllm/run-with-safety.yaml
Normal file
70
llama_stack/templates/remote-vllm/run-with-safety.yaml
Normal file
|
|
@ -0,0 +1,70 @@
|
|||
version: '2'
|
||||
image_name: remote-vllm
|
||||
docker_image: null
|
||||
conda_env: null
|
||||
apis:
|
||||
- agents
|
||||
- inference
|
||||
- memory
|
||||
- safety
|
||||
- telemetry
|
||||
providers:
|
||||
inference:
|
||||
- provider_id: vllm-inference
|
||||
provider_type: remote::vllm
|
||||
config:
|
||||
url: ${env.VLLM_URL}
|
||||
max_tokens: ${env.VLLM_MAX_TOKENS:4096}
|
||||
api_token: ${env.VLLM_API_TOKEN:fake}
|
||||
- provider_id: vllm-safety
|
||||
provider_type: remote::vllm
|
||||
config:
|
||||
url: ${env.SAFETY_VLLM_URL}
|
||||
max_tokens: ${env.VLLM_MAX_TOKENS:4096}
|
||||
api_token: ${env.VLLM_API_TOKEN:fake}
|
||||
memory:
|
||||
- provider_id: faiss
|
||||
provider_type: inline::faiss
|
||||
config:
|
||||
kvstore:
|
||||
type: sqlite
|
||||
namespace: null
|
||||
db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/faiss_store.db
|
||||
safety:
|
||||
- provider_id: llama-guard
|
||||
provider_type: inline::llama-guard
|
||||
config: {}
|
||||
agents:
|
||||
- provider_id: meta-reference
|
||||
provider_type: inline::meta-reference
|
||||
config:
|
||||
persistence_store:
|
||||
type: sqlite
|
||||
namespace: null
|
||||
db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/agents_store.db
|
||||
telemetry:
|
||||
- provider_id: meta-reference
|
||||
provider_type: inline::meta-reference
|
||||
config: {}
|
||||
metadata_store:
|
||||
namespace: null
|
||||
type: sqlite
|
||||
db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/registry.db
|
||||
models:
|
||||
- metadata: {}
|
||||
model_id: ${env.INFERENCE_MODEL}
|
||||
provider_id: vllm-inference
|
||||
provider_model_id: null
|
||||
- metadata: {}
|
||||
model_id: ${env.SAFETY_MODEL}
|
||||
provider_id: vllm-safety
|
||||
provider_model_id: null
|
||||
shields:
|
||||
- params: null
|
||||
shield_id: ${env.SAFETY_MODEL}
|
||||
provider_id: null
|
||||
provider_shield_id: null
|
||||
memory_banks: []
|
||||
datasets: []
|
||||
scoring_fns: []
|
||||
eval_tasks: []
|
||||
56
llama_stack/templates/remote-vllm/run.yaml
Normal file
56
llama_stack/templates/remote-vllm/run.yaml
Normal file
|
|
@ -0,0 +1,56 @@
|
|||
version: '2'
|
||||
image_name: remote-vllm
|
||||
docker_image: null
|
||||
conda_env: null
|
||||
apis:
|
||||
- agents
|
||||
- inference
|
||||
- memory
|
||||
- safety
|
||||
- telemetry
|
||||
providers:
|
||||
inference:
|
||||
- provider_id: vllm-inference
|
||||
provider_type: remote::vllm
|
||||
config:
|
||||
url: ${env.VLLM_URL}
|
||||
max_tokens: ${env.VLLM_MAX_TOKENS:4096}
|
||||
api_token: ${env.VLLM_API_TOKEN:fake}
|
||||
memory:
|
||||
- provider_id: faiss
|
||||
provider_type: inline::faiss
|
||||
config:
|
||||
kvstore:
|
||||
type: sqlite
|
||||
namespace: null
|
||||
db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/faiss_store.db
|
||||
safety:
|
||||
- provider_id: llama-guard
|
||||
provider_type: inline::llama-guard
|
||||
config: {}
|
||||
agents:
|
||||
- provider_id: meta-reference
|
||||
provider_type: inline::meta-reference
|
||||
config:
|
||||
persistence_store:
|
||||
type: sqlite
|
||||
namespace: null
|
||||
db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/agents_store.db
|
||||
telemetry:
|
||||
- provider_id: meta-reference
|
||||
provider_type: inline::meta-reference
|
||||
config: {}
|
||||
metadata_store:
|
||||
namespace: null
|
||||
type: sqlite
|
||||
db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/remote-vllm}/registry.db
|
||||
models:
|
||||
- metadata: {}
|
||||
model_id: ${env.INFERENCE_MODEL}
|
||||
provider_id: vllm-inference
|
||||
provider_model_id: null
|
||||
shields: []
|
||||
memory_banks: []
|
||||
datasets: []
|
||||
scoring_fns: []
|
||||
eval_tasks: []
|
||||
100
llama_stack/templates/remote-vllm/vllm.py
Normal file
100
llama_stack/templates/remote-vllm/vllm.py
Normal file
|
|
@ -0,0 +1,100 @@
|
|||
# Copyright (c) Meta Platforms, Inc. and affiliates.
|
||||
# All rights reserved.
|
||||
#
|
||||
# This source code is licensed under the terms described in the LICENSE file in
|
||||
# the root directory of this source tree.
|
||||
|
||||
from pathlib import Path
|
||||
|
||||
from llama_stack.distribution.datatypes import ModelInput, Provider, ShieldInput
|
||||
from llama_stack.providers.remote.inference.vllm import VLLMInferenceAdapterConfig
|
||||
from llama_stack.templates.template import DistributionTemplate, RunConfigSettings
|
||||
|
||||
|
||||
def get_distribution_template() -> DistributionTemplate:
|
||||
providers = {
|
||||
"inference": ["remote::vllm"],
|
||||
"memory": ["inline::faiss", "remote::chromadb", "remote::pgvector"],
|
||||
"safety": ["inline::llama-guard"],
|
||||
"agents": ["inline::meta-reference"],
|
||||
"telemetry": ["inline::meta-reference"],
|
||||
}
|
||||
|
||||
inference_provider = Provider(
|
||||
provider_id="vllm-inference",
|
||||
provider_type="remote::vllm",
|
||||
config=VLLMInferenceAdapterConfig.sample_run_config(
|
||||
url="${env.VLLM_URL}",
|
||||
),
|
||||
)
|
||||
|
||||
inference_model = ModelInput(
|
||||
model_id="${env.INFERENCE_MODEL}",
|
||||
provider_id="vllm-inference",
|
||||
)
|
||||
safety_model = ModelInput(
|
||||
model_id="${env.SAFETY_MODEL}",
|
||||
provider_id="vllm-safety",
|
||||
)
|
||||
|
||||
return DistributionTemplate(
|
||||
name="remote-vllm",
|
||||
distro_type="self_hosted",
|
||||
description="Use (an external) vLLM server for running LLM inference",
|
||||
template_path=Path(__file__).parent / "doc_template.md",
|
||||
providers=providers,
|
||||
default_models=[inference_model, safety_model],
|
||||
run_configs={
|
||||
"run.yaml": RunConfigSettings(
|
||||
provider_overrides={
|
||||
"inference": [inference_provider],
|
||||
},
|
||||
default_models=[inference_model],
|
||||
),
|
||||
"run-with-safety.yaml": RunConfigSettings(
|
||||
provider_overrides={
|
||||
"inference": [
|
||||
inference_provider,
|
||||
Provider(
|
||||
provider_id="vllm-safety",
|
||||
provider_type="remote::vllm",
|
||||
config=VLLMInferenceAdapterConfig.sample_run_config(
|
||||
url="${env.SAFETY_VLLM_URL}",
|
||||
),
|
||||
),
|
||||
],
|
||||
},
|
||||
default_models=[
|
||||
inference_model,
|
||||
safety_model,
|
||||
],
|
||||
default_shields=[ShieldInput(shield_id="${env.SAFETY_MODEL}")],
|
||||
),
|
||||
},
|
||||
docker_compose_env_vars={
|
||||
"LLAMASTACK_PORT": (
|
||||
"5001",
|
||||
"Port for the Llama Stack distribution server",
|
||||
),
|
||||
"INFERENCE_MODEL": (
|
||||
"meta-llama/Llama-3.2-3B-Instruct",
|
||||
"Inference model loaded into the vLLM server",
|
||||
),
|
||||
"VLLM_URL": (
|
||||
"http://host.docker.internal:5100}/v1",
|
||||
"URL of the vLLM server with the main inference model",
|
||||
),
|
||||
"MAX_TOKENS": (
|
||||
"4096",
|
||||
"Maximum number of tokens for generation",
|
||||
),
|
||||
"SAFETY_VLLM_URL": (
|
||||
"http://host.docker.internal:5101/v1",
|
||||
"URL of the vLLM server with the safety model",
|
||||
),
|
||||
"SAFETY_MODEL": (
|
||||
"meta-llama/Llama-Guard-3-1B",
|
||||
"Name of the safety (Llama-Guard) model to use",
|
||||
),
|
||||
},
|
||||
)
|
||||
Loading…
Add table
Add a link
Reference in a new issue