[rollout,vllm] fix: Add LoRA Loading to Async vLLM (volcengine#3639)

### What does this PR do?

Currently, async vLLM with AgentWorkerLoop throws an error when
`update_weights` with LoRA weights. This expands support for
AgentWorkerLoop with LoRAs.

### Checklist Before Starting

- [x] Search for similar PRs. Paste at least one query link here: ...
- [x] Format the PR title as `[{modules}] {type}: {description}` (This
will be checked by the CI)
- `{modules}` include `fsdp`, `megatron`, `sglang`, `vllm`, `rollout`,
`trainer`, `ci`, `training_utils`, `recipe`, `hardware`, `deployment`,
`ray`, `worker`, `single_controller`, `misc`, `perf`, `model`, `algo`,
`env`, `tool`, `ckpt`, `doc`, `data`
- If this PR involves multiple modules, separate them with `,` like
`[megatron, fsdp, doc]`
  - `{type}` is in `feat`, `fix`, `refactor`, `chore`, `test`
- If this PR breaks any API (CLI arguments, config, function signature,
etc.), add `[BREAKING]` to the beginning of the title.
  - Example: `[BREAKING][fsdp, megatron] feat: dynamic batching`

### Test

> For changes that can not be tested by CI (e.g., algorithm
implementation, new model support), validate by experiment(s) and show
results like training curve plots, evaluation results, etc.

### API and Usage Example

> Demonstrate how the API changes if any, and provide usage example(s)
if possible.

```python
# Add code snippet or script demonstrating how to use this
```

### Design & Code Changes

> Demonstrate the high-level design if this PR is complex, and list the
specific changes.

### Checklist Before Submitting

> [!IMPORTANT]
> Please check all the following items before requesting a review,
otherwise the reviewer might deprioritize this PR for review.

- [x] Read the [Contribute
Guide](https://github.com/volcengine/verl/blob/main/CONTRIBUTING.md).
- [x] Apply [pre-commit
checks](https://github.com/volcengine/verl/blob/main/CONTRIBUTING.md#code-linting-and-formatting):
`pre-commit install && pre-commit run --all-files --show-diff-on-failure
--color=always`
- [ ] Add / Update [the
documentation](https://github.com/volcengine/verl/tree/main/docs).
- [ ] Add unit or end-to-end test(s) to [the CI
workflow](https://github.com/volcengine/verl/tree/main/.github/workflows)
to cover all the code. If not feasible, explain why: ...
- [ ] Once your PR is ready for CI, send a message in [the `ci-request`
channel](https://verl-project.slack.com/archives/C091TCESWB1) in [the
`verl` Slack
workspace](https://join.slack.com/t/verl-project/shared_invite/zt-3855yhg8g-CTkqXu~hKojPCmo7k_yXTQ).
(If not accessible, please try [the Feishu group
(飞书群)](https://applink.larkoffice.com/client/chat/chatter/add_by_link?link_token=772jd4f1-cd91-441e-a820-498c6614126a).)

Author: kfallah

verl/workers/rollout/vllm_rollout/vllm_rollout_spmd.py

@@ -50,7 +50,7 @@
 from tensordict import TensorDict
 from torch.distributed.device_mesh import DeviceMesh
 from vllm import LLM, SamplingParams
-from vllm.config import CompilationConfig, CompilationLevel
+from vllm.config import CompilationConfig, CompilationLevel, LoRAConfig
 from vllm.lora.request import LoRARequest
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.worker.worker_base import WorkerWrapperBase
@@ -479,10 +479,12 @@ def __init__(
         device_mesh: DeviceMesh,
     ):
         super().__init__(config, model_config, device_mesh)
-
         self.tokenizer = model_config.tokenizer
         self.inference_engine: WorkerWrapperBase = None
         self.address = self._init_zeromq()
+        self.lora_config = (
+            {"max_loras": 1, "max_lora_rank": model_config.lora_rank} if model_config.lora_rank > 0 else {}
+        )
 
         # https://github.com/vllm-project/vllm/issues/25171
         if config.layered_summon or config.expert_parallel_size > 1:
@@ -536,7 +538,6 @@ def _init_worker(self, all_kwargs: list[dict[str, Any]]):
         """Initialize worker engine."""
         if not torch.distributed.is_initialized():
             initialize_global_process_group_ray()
-
         all_kwargs[0]["rank"] = int(os.environ["RANK"])
         device_name = "NPU" if is_npu_available else "GPU"
         all_kwargs[0]["local_rank"] = (
@@ -545,6 +546,8 @@ def _init_worker(self, all_kwargs: list[dict[str, Any]]):
             else int(ray.get_runtime_context().get_accelerator_ids()[device_name][0])
         )
         self.vllm_config = all_kwargs[0]["vllm_config"]
+        if self.lora_config:
+            self.vllm_config.lora_config = LoRAConfig(**self.lora_config)
         self.inference_engine = WorkerWrapperBase(vllm_config=self.vllm_config)
         self.inference_engine.init_worker(all_kwargs)
 
@@ -582,11 +585,24 @@ async def update_weights(self, weights: Generator[tuple[str, torch.Tensor], None
         Args:
             weights: A generator that yields the name of the weight tensor and the tensor itself.
         """
-        from verl.utils.vllm.patch import patch_vllm_moe_model_weight_loader
+        peft_config, base_sync_done = kwargs.get("peft_config", None), kwargs.get("base_sync_done", False)
+        if peft_config and base_sync_done:
+            lora_int_id = int(time.time_ns() % 0x7FFFFFFF)
+            lora_reqest = TensorLoRARequest(
+                lora_name=f"{lora_int_id}",
+                lora_int_id=lora_int_id,
+                lora_path="simon_lora_path",
+                peft_config=asdict(peft_config),
+                lora_tensors=weights,
+            )
+            self.inference_engine.worker.add_lora(lora_reqest)
+            logger.info(f"vLLM load weights, loaded_params: {len(weights)}")
+        else:
+            from verl.utils.vllm.patch import patch_vllm_moe_model_weight_loader
 
-        model = self.inference_engine.worker.model_runner.model
-        patch_vllm_moe_model_weight_loader(model)
-        model.load_weights(weights)
+            model = self.inference_engine.worker.model_runner.model
+            patch_vllm_moe_model_weight_loader(model)
+            model.load_weights(weights)
 
     def generate_sequences(self, prompts: DataProto) -> DataProto:
         """Batch generate sequences in sync mode."""