[recipe] fix: Update the grpo training script for gpt-oss models (verl-project#3836)

### What does this PR do?

> Add **concise** overview of what this PR aims to achieve or
accomplish. Reference related GitHub issues and PRs that help with the
review.
* remove customized python packages since they are already supported 
* Add reasoning_effort input 
* recommend a setup for batch size to avoid MOE instability. 

### 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
Test offline: run a training job
> 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).)

Co-authored-by: Hejian Sang <hsang@linkedin.com>

Author: HJSang

examples/grpo_trainer/run_gptoss_20b.sh

@@ -1,25 +1,5 @@
 #!/bin/bash
 
-# install flashinfer
-cd $HOME
-git clone https://github.com/flashinfer-ai/flashinfer.git --recursive
-cd flashinfer
-python -m pip install -v .
-
-# install sglang
-cd $HOME
-git fetch origin pull/9379/head:fix_weight_loading
-cd $HOME/sglang
-git checkout fix_weight_loading
-pip install --upgrade pip
-pip install -e "python[all]"
-
-pip install peft
-pip install transformers -U
-pip install https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.3/flash_attn-2.8.3+cu12torch2.8cxx11abiTRUE-cp311-cp311-linux_x86_64.whl
-pip install numpy==1.26.4
-
-
 cat > get_model.py << EOF
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, Mxfp4Config
@@ -47,19 +27,23 @@ tokenizer.save_pretrained(output_dir)
 EOF
 
 python get_model.py
-
+# or you can use lmsys/gpt-oss-20b-bf16
+# recommend to use same value for train_batch_size and ppo_mini_batch_size
+# to avoid MOE training instability
+# use large value for max_response_length if you want to use reasoning effort high.
 
 
 model_dir=$HOME/models/gpt-oss-20b-bf16
 python3 -m verl.trainer.main_ppo \
     algorithm.adv_estimator=grpo \
     data.train_files="$gsm8k_train_path" \
     data.val_files="$gsm8k_test_path" \
-    data.train_batch_size=1024 \
-    data.max_prompt_length=1024 \
-    data.max_response_length=1024 \
+    data.train_batch_size=256 \
+    data.max_prompt_length=512 \
+    data.max_response_length=8192 \
     data.filter_overlong_prompts=True \
     data.truncation='error' \
+    +data.apply_chat_template_kwargs.reasoning_effort=medium \
     actor_rollout_ref.model.path=${model_dir} \
     actor_rollout_ref.actor.optim.lr=1e-6 \
     actor_rollout_ref.model.use_remove_padding=True \
@@ -76,8 +60,9 @@ python3 -m verl.trainer.main_ppo \
     actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=32 \
     actor_rollout_ref.rollout.tensor_model_parallel_size=2 \
     actor_rollout_ref.rollout.name=sglang \
+    actor_rollout_ref.rollout.mode=sync \
     actor_rollout_ref.rollout.engine_kwargs.sglang.attention_backend=triton \
-    actor_rollout_ref.rollout.gpu_memory_utilization=0.6 \
+    actor_rollout_ref.rollout.gpu_memory_utilization=0.7 \
     actor_rollout_ref.rollout.n=5 \
     actor_rollout_ref.rollout.load_format=safetensors \
     actor_rollout_ref.ref.log_prob_micro_batch_size_per_gpu=32 \