# What does this PR do?
This PR adds two methods to the Inference API:
- `batch_completion`
- `batch_chat_completion`
The motivation is for evaluations targeting a local inference engine
(like meta-reference or vllm) where batch APIs provide for a substantial
amount of acceleration.
Why did I not add this to `Api.batch_inference` though? That just
resulted in a _lot_ more book-keeping given the structure of Llama
Stack. Had I done that, I would have needed to create a notion of a
"batch model" resource, setup routing based on that, etc. This does not
sound ideal.
So what's the future of the batch inference API? I am not sure. Maybe we
can keep it for true _asynchronous_ execution. So you can submit
requests, and it can return a Job instance, etc.
## Test Plan
Run meta-reference-gpu using:
```bash
export INFERENCE_MODEL=meta-llama/Llama-4-Scout-17B-16E-Instruct
export INFERENCE_CHECKPOINT_DIR=../checkpoints/Llama-4-Scout-17B-16E-Instruct-20250331210000
export MODEL_PARALLEL_SIZE=4
export MAX_BATCH_SIZE=32
export MAX_SEQ_LEN=6144
LLAMA_MODELS_DEBUG=1 llama stack run meta-reference-gpu
```
Then run the batch inference test case.
Mirror to https://github.com/meta-llama/llama-models/pull/324 with some
clean up
```
with-proxy pip install -e .
export INFERENCE_MODEL=meta-llama/Llama-4-Scout-17B-16E-Instruct
export INFERENCE_CHECKPOINT_DIR=../checkpoints/Llama-4-Scout-17B-16E-Instruct
export QUANTIZATION_TYPE=int4_mixed
with-proxy llama stack build --run --template meta-reference-gpu
```
# What does this PR do?
[Provide a short summary of what this PR does and why. Link to relevant
issues if applicable.]
[//]: # (If resolving an issue, uncomment and update the line below)
[//]: # (Closes #[issue-number])
## Test Plan
[Describe the tests you ran to verify your changes with result
summaries. *Provide clear instructions so the plan can be easily
re-executed.*]
[//]: # (## Documentation)
# What does this PR do?
Move around bits. This makes the copies from llama-models _much_ easier
to maintain and ensures we don't entangle meta-reference specific
tidbits into llama-models code even by accident.
Also, kills the meta-reference-quantized-gpu distro and rolls
quantization deps into meta-reference-gpu.
## Test Plan
```
LLAMA_MODELS_DEBUG=1 \
with-proxy llama stack run meta-reference-gpu \
--env INFERENCE_MODEL=meta-llama/Llama-4-Scout-17B-16E-Instruct \
--env INFERENCE_CHECKPOINT_DIR=<DIR> \
--env MODEL_PARALLEL_SIZE=4 \
--env QUANTIZATION_TYPE=fp8_mixed
```
Start a server with and without quantization. Point integration tests to
it using:
```
pytest -s -v tests/integration/inference/test_text_inference.py \
--stack-config http://localhost:8321 --text-model meta-llama/Llama-4-Scout-17B-16E-Instruct
```
Running full Tool Calling required some updates to work e2e.
- Remove `python_start` and `python_end` tags
- Tool Call messages and Tool Resposne messages should end with
`<|eom|>`
- System prompt needed updates
```
You are a helpful assisant who can can answer general questions or invoke tools when necessary.
In addition to tool calls, you should also augment your responses by using the tool outputs.
```
### Test Plan
- Start server with meta-reference
```
LLAMA_STACK_DISABLE_VERSION_CHECK=1 LLAMA_MODELS_DEBUG=1 INFERENCE_MODEL=meta-llama/$MODEL llama stack run meta-reference-gpu
```
- Added **NEW** tests with 5 test cases for multi-turn tool calls
```
pytest -s -v --stack-config http://localhost:8321 tests/integration/inference/test_text_inference.py --text-model meta-llama/Llama-4-Scout-17B-16E-Instruct
```
- Also verified all vision and agent tests pass
# What does this PR do?
closes#1584
This should be a rather innocuous change.
## Test Plan
Verify that there's no more tool call parsing error for example in issue
<img width="1216" alt="image"
src="https://github.com/user-attachments/assets/a5a6f4e8-2093-4ca2-bc06-794b707a0429"
/>
LLAMA_STACK_CONFIG=fireworks pytest -s -v
tests/integration/agents/test_agents.py --safety-shield
meta-llama/Llama-Guard-3-8B --text-model
meta-llama/Llama-3.1-8B-Instruct
# What does this PR do?
Don't set type variables from register_schema().
`mypy` is not happy about it since type variables are calculated at
runtime and hence the typing hints are not available during static
analysis.
Good news is there is no good reason to set the variables from the
return type.
Signed-off-by: Ihar Hrachyshka <ihar.hrachyshka@gmail.com>
Signed-off-by: Ihar Hrachyshka <ihar.hrachyshka@gmail.com>
### What does this PR do?
Currently, `ToolCall.arguments` is a `Dict[str, RecursiveType]`.
However, on the client SDK side -- the `RecursiveType` gets deserialized
into a number ( both int and float get collapsed ) and hence when params
are `int` they get converted to float which might break client side
tools that might be doing type checking.
Closes: https://github.com/meta-llama/llama-stack/issues/1683
### Test Plan
Stainless changes --
https://github.com/meta-llama/llama-stack-client-python/pull/204
```
pytest -s -v --stack-config=fireworks tests/integration/agents/test_agents.py --text-model meta-llama/Llama-3.1-8B-Instruct
```
# What does this PR do?
Updated all instances of datetime.now() to use timezone.utc for
consistency in handling time across different systems. This ensures that
timestamps are always in Coordinated Universal Time (UTC), avoiding
issues with time zone discrepancies and promoting uniformity in
time-related data.
Signed-off-by: Sébastien Han <seb@redhat.com>
# Summary:
The current prompt doesn't work well and tend to overindex on tool
calling. This PR is not perfect, but should be an improvement over the
current prompt. We can keep iterating.
# Test Plan:
Ran on a (small) eval with 20 HotpotQA examples.
With current prompt:
https://gist.github.com/ehhuang/9f967e62751907165eb13781ea968f5c
{
│ 'basic::equality': {'accuracy': {'accuracy': 0.2, 'num_correct': 4.0,
'num_total': 20}},
│ 'F1ScoringFn': {
│ │ 'f1_average': 0.25333333333333335,
│ │ 'precision_average': 0.23301767676767676,
│ │ 'recall_average': 0.375
│ }
}
num_tool_calls=[5, 5, 5, 5, 5, 5, 2, 5, 5, 5, 5, 5, 2, 2, 1, 1, 2, 1, 2,
2]
num_examples_with_tool_call=20
num_examples_with_pythontag=0
#########################################################
With new prompt:
https://gist.github.com/ehhuang/6e4a8ecf54db68922c2be8700056f962
{
│ 'basic::equality': {'accuracy': {'accuracy': 0.25, 'num_correct': 5.0,
'num_total': 20}},
│ 'F1ScoringFn': {
│ │ 'f1_average': 0.35579260478321006,
│ │ 'precision_average': 0.32030238933180105,
│ │ 'recall_average': 0.6091666666666666
│ }
}
num_tool_calls=[2, 1, 1, 5, 5, 5, 1, 2, 1, 1, 2, 1, 1, 1, 1, 1, 2, 1, 3,
2]
num_examples_with_tool_call=20
num_examples_with_pythontag=0
The answers have higher recall, and make fewer tool calls. Note that
these were run with max_infer_iter=5, so the current prompt hits this
limit more often, and without the limit, someitmes goes into infinite
tool calling loop.
The data here is with 3.3-70B. Results are equally poor with either
prompt with 3.2-3B ~30 recall.
llama-models should have extremely minimal cruft. Its sole purpose
should be didactic -- show the simplest implementation of the llama
models and document the prompt formats, etc.
This PR is the complement to
https://github.com/meta-llama/llama-models/pull/279
## Test Plan
Ensure all `llama` CLI `model` sub-commands work:
```bash
llama model list
llama model download --model-id ...
llama model prompt-format -m ...
```
Ran tests:
```bash
cd tests/client-sdk
LLAMA_STACK_CONFIG=fireworks pytest -s -v inference/
LLAMA_STACK_CONFIG=fireworks pytest -s -v vector_io/
LLAMA_STACK_CONFIG=fireworks pytest -s -v agents/
```
Create a fresh venv `uv venv && source .venv/bin/activate` and run
`llama stack build --template fireworks --image-type venv` followed by
`llama stack run together --image-type venv` <-- the server runs
Also checked that the OpenAPI generator can run and there is no change
in the generated files as a result.
```bash
cd docs/openapi_generator
sh run_openapi_generator.sh
```
