GitLab

Co-authored-by: mohammedabdulwahhab <furkhan324@berkeley.edu>

. New mistralrs and llamacpp version
. mistralrs: Handle Gemma 3 and Llama 4 as vision models
. Update the dynamo-run docs to use Qwen 3
. Our pre-processor now supports Llama 4's newer multi-modal `config.json`
. Upgrade minijinja to handle Qwen 3's prompt template

For Llama 4 we'll need to limit the max seq len. vllm says:
> To serve at least one request with the models's max seq len (10485760), (240.00 GiB KV cache is needed,...

I was able to run Llama 4 with llamacpp and a quantized GGUF, with Dynamo doing the pre-processing.

Signed-off-by: Anthony Casagrande <acasagrande@nvidia.com>

Signed-off-by: Graham King <graham@gkgk.org>
Co-authored-by: Ryan McCormick <rmccormick@nvidia.com>

Signed-off-by: Tanmay Verma <tanmay2592@gmail.com>
Co-authored-by: Ryan McCormick <rmccormick@nvidia.com>

Signed-off-by: Chasing1020 <chasing1020@gmail.com>

vllm and sglang are now the sub-process engines from #954

Also updated docs on doing vllm and sglang multi-gpu (tensor parallel) and multi-node (pipeline parallel).

New vllm and sglang engines that run in a sub-process. Will hopefully replace the existing embedded python engines.
    
Why?
    
  - Pure Python, does not require knowing Rust to work on it. Much simpler to maintain.
  - No embedded Python interpreter which avoids linking libpython and avoids the MacOS virtualenv issues.
  - Should have better performance as it's "native" vllm / sglang.
  - Works with any version of vllm (including v1!) and sglang. Less upgrade struggle.

Approved by OSRB in Slack.

Note we don't check for the closing delimiter to allow the longer copyright format.

Motivation is that it reduces the context usage by 12 lines for every file in the project. That helps things like Cursor and Claude Code fit more, go faster, and cost less.

Adding this to a Python script makes it register on the network so that `dynamo-run` can discover it and send it requests:
```
from dynamo.llm import register_llm

MODEL = "Qwen/Qwen2.5-0.5B-Instruct"
await register_llm(endpoint, MODEL, 3)
```

Full vllm example, with pre-processing in dynamo:
- `dynamo-run in=text out=dyn://dynamo.backend.generate`
- `cd lib/bindings/python/examples/hello_world`
- `python server_vllm.py`

This builds on top of the work to move pre-processor to ingress side. It means we can decouple Rust and Python using NATS as the bus.

The `register_llm` call does this:

- Download the model from HF if necessary
- Load the model deployment card from the HF folder or extract from GGUF
- Push the tokenizer config etc into NATS object store so ingress can access it from a different machine
- Publish the model deployment card to ETCD

Signed-off-by: Harrison Saturley-Hall <454891+saturley-hall@users.noreply.github.com>
Co-authored-by: Anant Sharma <anants@nvidia.com>

Part of https://github.com/ai-dynamo/dynamo/issues/743