phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-08-07 02:58:21 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

update doc.md

Browse source
This commit is contained in:
Hardik Shah 2025-02-05 15:36:29 -08:00
parent 0d006bb341
commit aa6138e0ca
3 changed files with 162 additions and 81 deletions
Download patch file Download diff file Expand all files Collapse all files

121
docs/source/distributions/self_hosted_distro/dell.md
View file

@ -26,15 +26,15 @@ The `llamastack/distribution-dell` distribution consists of the following provid
| vector_io | `inline::faiss`, `remote::chromadb`, `remote::pgvector` | | vector_io | `inline::faiss`, `remote::chromadb`, `remote::pgvector` |
You can use this distribution if you have GPUs and want to run an independent TGI server container for running inference. You can use this distribution if you have GPUs and want to run an independent TGI or Dell Enterprise Hub container for running inference.
### Environment Variables ### Environment Variables
The following environment variables can be configured: The following environment variables can be configured:
- `DEH_URL`: URL for the Dell inference server (default: `http://0.0.0.0:8080`) - `DEH_URL`: URL for the Dell inference server (default: `http://0.0.0.0:8181`)
- `DEH_SAFETY_URL`: URL for the Dell safety inference server (default: `http://0.0.0.0:8081`) - `DEH_SAFETY_URL`: URL for the Dell safety inference server (default: `http://0.0.0.0:8282`)
- `CHROMA_URL`: URL for the Chroma server (default: `http://0.0.0.0:8000`) - `CHROMA_URL`: URL for the Chroma server (default: `http://localhost:6601`)
- `INFERENCE_MODEL`: Inference model loaded into the TGI server (default: `meta-llama/Llama-3.2-3B-Instruct`) - `INFERENCE_MODEL`: Inference model loaded into the TGI server (default: `meta-llama/Llama-3.2-3B-Instruct`)
- `SAFETY_MODEL`: Name of the safety (Llama-Guard) model to use (default: `meta-llama/Llama-Guard-3-1B`) - `SAFETY_MODEL`: Name of the safety (Llama-Guard) model to use (default: `meta-llama/Llama-Guard-3-1B`)
@ -44,40 +44,69 @@ The following environment variables can be configured:
Please check the [TGI Getting Started Guide](https://github.com/huggingface/text-generation-inference?tab=readme-ov-file#get-started) to get a TGI endpoint. Here is a sample script to start a TGI server locally via Docker: Please check the [TGI Getting Started Guide](https://github.com/huggingface/text-generation-inference?tab=readme-ov-file#get-started) to get a TGI endpoint. Here is a sample script to start a TGI server locally via Docker:
```bash ```bash
export INFERENCE_PORT=8080 export INFERENCE_PORT=8181
export INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct export DEH_URL=http://0.0.0.0:$INFERENCE_PORT
export INFERENCE_MODEL=meta-llama/Llama-3.1-8B-Instruct
export CHROMADB_HOST=localhost
export CHROMADB_PORT=6601
export CHROMA_URL=http://$CHROMADB_HOST:$CHROMADB_PORT
export CUDA_VISIBLE_DEVICES=0 export CUDA_VISIBLE_DEVICES=0
export LLAMA_STACK_PORT=8321
docker run --rm -it \ docker run --rm -it \
-v $HOME/.cache/huggingface:/data \ --network host \
-p $INFERENCE_PORT:$INFERENCE_PORT \ -v $HOME/.cache/huggingface:/data \
--gpus $CUDA_VISIBLE_DEVICES \ -e HF_TOKEN=$HF_TOKEN \
ghcr.io/huggingface/text-generation-inference:2.3.1 \ -p $INFERENCE_PORT:$INFERENCE_PORT \
--dtype bfloat16 \ --gpus $CUDA_VISIBLE_DEVICES \
--usage-stats off \ ghcr.io/huggingface/text-generation-inference \
--sharded false \ --dtype bfloat16 \
--cuda-memory-fraction 0.7 \ --usage-stats off \
--model-id $INFERENCE_MODEL \ --sharded false \
--port $INFERENCE_PORT --cuda-memory-fraction 0.7 \
--model-id $INFERENCE_MODEL \
--port $INFERENCE_PORT --hostname 0.0.0.0
``` ```
If you are using Llama Stack Safety / Shield APIs, then you will need to also run another instance of a TGI with a corresponding safety model like `meta-llama/Llama-Guard-3-1B` using a script like: If you are using Llama Stack Safety / Shield APIs, then you will need to also run another instance of a TGI with a corresponding safety model like `meta-llama/Llama-Guard-3-1B` using a script like:
```bash ```bash
export SAFETY_PORT=8081 export SAFETY_INFERENCE_PORT=8282
export DEH_SAFETY_URL=http://0.0.0.0:$SAFETY_INFERENCE_PORT
export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
export CUDA_VISIBLE_DEVICES=1 export CUDA_VISIBLE_DEVICES=1
docker run --rm -it \ docker run --rm -it \
-v $HOME/.cache/huggingface:/data \ --network host \
-p $SAFETY_PORT:$SAFETY_PORT \ -v $HOME/.cache/huggingface:/data \
--gpus $CUDA_VISIBLE_DEVICES \ -e HF_TOKEN=$HF_TOKEN \
ghcr.io/huggingface/text-generation-inference:2.3.1 \ -p $SAFETY_PORT:$SAFETY_PORT \
--dtype bfloat16 \ --gpus $CUDA_VISIBLE_DEVICES \
--usage-stats off \ ghcr.io/huggingface/text-generation-inference \
--sharded false \ --dtype bfloat16 \
--model-id $SAFETY_MODEL \ --usage-stats off \
--port $SAFETY_PORT --sharded false \
--cuda-memory-fraction 0.7 \
--model-id $SAFETY_MODEL \
--hostname 0.0.0.0 \
--port $SAFETY_INFERENCE_PORT
```
## Dell distribution relies on ChromDB for vector database usage
You can start a chrom-db easily using docker.
```bash
# This is where the indices are persisted
mkdir -p chromadb
podman run --rm -it \
--network host \
--name chromadb \
-v ./chromadb:/chroma/chroma \
-e IS_PERSISTENT=TRUE \
chromadb/chroma:latest \
--port $CHROMADB_PORT \
--host $CHROMADB_HOST
``` ```
## Running Llama Stack ## Running Llama Stack
@ -89,14 +118,19 @@ Now you are ready to run Llama Stack with TGI as the inference provider. You can
This method allows you to get started quickly without having to build the distribution code. This method allows you to get started quickly without having to build the distribution code.
```bash ```bash
LLAMA_STACK_PORT=5001 docker run -it \
docker run \ --network host \
-it \ -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
-p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \ -v ~/.llama:/root/.llama \
llamastack/distribution-dell \ # NOTE: mount the llama-stack / llama-model directories if testing local changes else not needed
--port $LLAMA_STACK_PORT \ -v /home/hjshah/git/llama-stack:/app/llama-stack-source -v /home/hjshah/git/llama-models:/app/llama-models-source \
--env INFERENCE_MODEL=$INFERENCE_MODEL \ # localhost/distribution-dell:dev if building / testing locally
--env TGI_URL=http://host.docker.internal:$INFERENCE_PORT llamastack/distribution-dell\
--port $LLAMA_STACK_PORT \
--env INFERENCE_MODEL=$INFERENCE_MODEL \
--env DEH_URL=$DEH_URL \
--env CHROMA_URL=$CHROMA_URL
``` ```
If you are using Llama Stack Safety / Shield APIs, use: If you are using Llama Stack Safety / Shield APIs, use:
@ -106,6 +140,10 @@ If you are using Llama Stack Safety / Shield APIs, use:
# git clone https://github.com/meta-llama/llama-stack.git # git clone https://github.com/meta-llama/llama-stack.git
cd /path/to/llama-stack cd /path/to/llama-stack
export SAFETY_INFERENCE_PORT=8282
export DEH_SAFETY_URL=http://0.0.0.0:$SAFETY_INFERENCE_PORT
export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
docker run \ docker run \
-it \ -it \
-p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \ -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
@ -115,9 +153,10 @@ docker run \
--yaml-config /root/my-run.yaml \ --yaml-config /root/my-run.yaml \
--port $LLAMA_STACK_PORT \ --port $LLAMA_STACK_PORT \
--env INFERENCE_MODEL=$INFERENCE_MODEL \ --env INFERENCE_MODEL=$INFERENCE_MODEL \
--env TGI_URL=http://host.docker.internal:$INFERENCE_PORT \ --env DEH_URL=$DEH_URL \
--env SAFETY_MODEL=$SAFETY_MODEL \ --env SAFETY_MODEL=$SAFETY_MODEL \
--env TGI_SAFETY_URL=http://host.docker.internal:$SAFETY_PORT --env DEH_SAFETY_URL=$DEH_SAFETY_URL \
--env CHROMA_URL=$CHROMA_URL
``` ```
### Via Conda ### Via Conda
@ -126,10 +165,11 @@ Make sure you have done `pip install llama-stack` and have the Llama Stack CLI a
```bash ```bash
llama stack build --template dell --image-type conda llama stack build --template dell --image-type conda
llama stack run ./run.yaml llama stack run dell
--port $LLAMA_STACK_PORT \ --port $LLAMA_STACK_PORT \
--env INFERENCE_MODEL=$INFERENCE_MODEL \ --env INFERENCE_MODEL=$INFERENCE_MODEL \
--env TGI_URL=http://127.0.0.1:$INFERENCE_PORT --env DEH_URL=$DEH_URL \
--env CHROMA_URL=$CHROMA_URL
``` ```
If you are using Llama Stack Safety / Shield APIs, use: If you are using Llama Stack Safety / Shield APIs, use:
@ -138,7 +178,8 @@ If you are using Llama Stack Safety / Shield APIs, use:
llama stack run ./run-with-safety.yaml \ llama stack run ./run-with-safety.yaml \
--port $LLAMA_STACK_PORT \ --port $LLAMA_STACK_PORT \
--env INFERENCE_MODEL=$INFERENCE_MODEL \ --env INFERENCE_MODEL=$INFERENCE_MODEL \
--env TGI_URL=http://127.0.0.1:$INFERENCE_PORT \ --env DEH_URL=$DEH_URL \
--env SAFETY_MODEL=$SAFETY_MODEL \ --env SAFETY_MODEL=$SAFETY_MODEL \
--env TGI_SAFETY_URL=http://127.0.0.1:$SAFETY_PORT --env DEH_SAFETY_URL=$DEH_SAFETY_URL \
--env CHROMA_URL=$CHROMA_URL
``` ```

7
llama_stack/templates/dell/dell.py
View file

@ -128,16 +128,15 @@ def get_distribution_template() -> DistributionTemplate:
}, },
run_config_env_vars={ run_config_env_vars={
"DEH_URL": ( "DEH_URL": (
"http://0.0.0.0:8080", "http://0.0.0.0:8181",
"URL for the Dell inference server", "URL for the Dell inference server",
), ),
"DEH_SAFETY_URL": ( "DEH_SAFETY_URL": (
"http://0.0.0.0:8081", "http://0.0.0.0:8282",
"URL for the Dell safety inference server", "URL for the Dell safety inference server",
), ),
"CHROMA_URL": ( "CHROMA_URL": (
# http://host.containers.internal:8000 if running via docker "http://localhost:6601",
"http://0.0.0.0:8000",
"URL for the Chroma server", "URL for the Chroma server",
), ),
"INFERENCE_MODEL": ( "INFERENCE_MODEL": (

115
llama_stack/templates/dell/doc_template.md
View file

@ -15,7 +15,7 @@ The `llamastack/distribution-{{ name }}` distribution consists of the following
{{ providers_table }} {{ providers_table }}
You can use this distribution if you have GPUs and want to run an independent TGI server container for running inference. You can use this distribution if you have GPUs and want to run an independent TGI or Dell Enterprise Hub container for running inference.
{% if run_config_env_vars %} {% if run_config_env_vars %}
### Environment Variables ### Environment Variables
@ -33,40 +33,69 @@ The following environment variables can be configured:
Please check the [TGI Getting Started Guide](https://github.com/huggingface/text-generation-inference?tab=readme-ov-file#get-started) to get a TGI endpoint. Here is a sample script to start a TGI server locally via Docker: Please check the [TGI Getting Started Guide](https://github.com/huggingface/text-generation-inference?tab=readme-ov-file#get-started) to get a TGI endpoint. Here is a sample script to start a TGI server locally via Docker:
```bash ```bash
export INFERENCE_PORT=8080 export INFERENCE_PORT=8181
export INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct export DEH_URL=http://0.0.0.0:$INFERENCE_PORT
export INFERENCE_MODEL=meta-llama/Llama-3.1-8B-Instruct
export CHROMADB_HOST=localhost
export CHROMADB_PORT=6601
export CHROMA_URL=http://$CHROMADB_HOST:$CHROMADB_PORT
export CUDA_VISIBLE_DEVICES=0 export CUDA_VISIBLE_DEVICES=0
export LLAMA_STACK_PORT=8321
docker run --rm -it \ docker run --rm -it \
-v $HOME/.cache/huggingface:/data \ --network host \
-p $INFERENCE_PORT:$INFERENCE_PORT \ -v $HOME/.cache/huggingface:/data \
--gpus $CUDA_VISIBLE_DEVICES \ -e HF_TOKEN=$HF_TOKEN \
ghcr.io/huggingface/text-generation-inference:2.3.1 \ -p $INFERENCE_PORT:$INFERENCE_PORT \
--dtype bfloat16 \ --gpus $CUDA_VISIBLE_DEVICES \
--usage-stats off \ ghcr.io/huggingface/text-generation-inference \
--sharded false \ --dtype bfloat16 \
--cuda-memory-fraction 0.7 \ --usage-stats off \
--model-id $INFERENCE_MODEL \ --sharded false \
--port $INFERENCE_PORT --cuda-memory-fraction 0.7 \
--model-id $INFERENCE_MODEL \
--port $INFERENCE_PORT --hostname 0.0.0.0
``` ```
If you are using Llama Stack Safety / Shield APIs, then you will need to also run another instance of a TGI with a corresponding safety model like `meta-llama/Llama-Guard-3-1B` using a script like: If you are using Llama Stack Safety / Shield APIs, then you will need to also run another instance of a TGI with a corresponding safety model like `meta-llama/Llama-Guard-3-1B` using a script like:
```bash ```bash
export SAFETY_PORT=8081 export SAFETY_INFERENCE_PORT=8282
export DEH_SAFETY_URL=http://0.0.0.0:$SAFETY_INFERENCE_PORT
export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
export CUDA_VISIBLE_DEVICES=1 export CUDA_VISIBLE_DEVICES=1
docker run --rm -it \ docker run --rm -it \
-v $HOME/.cache/huggingface:/data \ --network host \
-p $SAFETY_PORT:$SAFETY_PORT \ -v $HOME/.cache/huggingface:/data \
--gpus $CUDA_VISIBLE_DEVICES \ -e HF_TOKEN=$HF_TOKEN \
ghcr.io/huggingface/text-generation-inference:2.3.1 \ -p $SAFETY_PORT:$SAFETY_PORT \
--dtype bfloat16 \ --gpus $CUDA_VISIBLE_DEVICES \
--usage-stats off \ ghcr.io/huggingface/text-generation-inference \
--sharded false \ --dtype bfloat16 \
--model-id $SAFETY_MODEL \ --usage-stats off \
--port $SAFETY_PORT --sharded false \
--cuda-memory-fraction 0.7 \
--model-id $SAFETY_MODEL \
--hostname 0.0.0.0 \
--port $SAFETY_INFERENCE_PORT
```
## Dell distribution relies on ChromDB for vector database usage
You can start a chrom-db easily using docker.
```bash
# This is where the indices are persisted
mkdir -p chromadb
podman run --rm -it \
--network host \
--name chromadb \
-v ./chromadb:/chroma/chroma \
-e IS_PERSISTENT=TRUE \
chromadb/chroma:latest \
--port $CHROMADB_PORT \
--host $CHROMADB_HOST
``` ```
## Running Llama Stack ## Running Llama Stack
@ -78,14 +107,19 @@ Now you are ready to run Llama Stack with TGI as the inference provider. You can
This method allows you to get started quickly without having to build the distribution code. This method allows you to get started quickly without having to build the distribution code.
```bash ```bash
LLAMA_STACK_PORT=5001 docker run -it \
docker run \ --network host \
-it \ -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
-p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \ -v ~/.llama:/root/.llama \
llamastack/distribution-{{ name }} \ # NOTE: mount the llama-stack / llama-model directories if testing local changes else not needed
--port $LLAMA_STACK_PORT \ -v /home/hjshah/git/llama-stack:/app/llama-stack-source -v /home/hjshah/git/llama-models:/app/llama-models-source \
--env INFERENCE_MODEL=$INFERENCE_MODEL \ # localhost/distribution-dell:dev if building / testing locally
--env TGI_URL=http://host.docker.internal:$INFERENCE_PORT llamastack/distribution-{{ name }}\
--port $LLAMA_STACK_PORT \
--env INFERENCE_MODEL=$INFERENCE_MODEL \
--env DEH_URL=$DEH_URL \
--env CHROMA_URL=$CHROMA_URL
``` ```
If you are using Llama Stack Safety / Shield APIs, use: If you are using Llama Stack Safety / Shield APIs, use:
@ -95,6 +129,10 @@ If you are using Llama Stack Safety / Shield APIs, use:
# git clone https://github.com/meta-llama/llama-stack.git # git clone https://github.com/meta-llama/llama-stack.git
cd /path/to/llama-stack cd /path/to/llama-stack
export SAFETY_INFERENCE_PORT=8282
export DEH_SAFETY_URL=http://0.0.0.0:$SAFETY_INFERENCE_PORT
export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
docker run \ docker run \
-it \ -it \
-p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \ -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
@ -104,9 +142,10 @@ docker run \
--yaml-config /root/my-run.yaml \ --yaml-config /root/my-run.yaml \
--port $LLAMA_STACK_PORT \ --port $LLAMA_STACK_PORT \
--env INFERENCE_MODEL=$INFERENCE_MODEL \ --env INFERENCE_MODEL=$INFERENCE_MODEL \
--env TGI_URL=http://host.docker.internal:$INFERENCE_PORT \ --env DEH_URL=$DEH_URL \
--env SAFETY_MODEL=$SAFETY_MODEL \ --env SAFETY_MODEL=$SAFETY_MODEL \
--env TGI_SAFETY_URL=http://host.docker.internal:$SAFETY_PORT --env DEH_SAFETY_URL=$DEH_SAFETY_URL \
--env CHROMA_URL=$CHROMA_URL
``` ```
### Via Conda ### Via Conda
@ -115,10 +154,11 @@ Make sure you have done `pip install llama-stack` and have the Llama Stack CLI a
```bash ```bash
llama stack build --template {{ name }} --image-type conda llama stack build --template {{ name }} --image-type conda
llama stack run ./run.yaml llama stack run {{ name }}
--port $LLAMA_STACK_PORT \ --port $LLAMA_STACK_PORT \
--env INFERENCE_MODEL=$INFERENCE_MODEL \ --env INFERENCE_MODEL=$INFERENCE_MODEL \
--env TGI_URL=http://127.0.0.1:$INFERENCE_PORT --env DEH_URL=$DEH_URL \
--env CHROMA_URL=$CHROMA_URL
``` ```
If you are using Llama Stack Safety / Shield APIs, use: If you are using Llama Stack Safety / Shield APIs, use:
@ -127,7 +167,8 @@ If you are using Llama Stack Safety / Shield APIs, use:
llama stack run ./run-with-safety.yaml \ llama stack run ./run-with-safety.yaml \
--port $LLAMA_STACK_PORT \ --port $LLAMA_STACK_PORT \
--env INFERENCE_MODEL=$INFERENCE_MODEL \ --env INFERENCE_MODEL=$INFERENCE_MODEL \
--env TGI_URL=http://127.0.0.1:$INFERENCE_PORT \ --env DEH_URL=$DEH_URL \
--env SAFETY_MODEL=$SAFETY_MODEL \ --env SAFETY_MODEL=$SAFETY_MODEL \
--env TGI_SAFETY_URL=http://127.0.0.1:$SAFETY_PORT --env DEH_SAFETY_URL=$DEH_SAFETY_URL \
--env CHROMA_URL=$CHROMA_URL
``` ```