parameterize enable_prefix_caching

[ji-huazhong]

What does this PR do?

Fixes #2798

Tested using the configuration (for functional verification only) of open-r1 with minor modifications:

# Model arguments
model_name_or_path: Qwen/Qwen2.5-1.5B-Instruct
model_revision: main
torch_dtype: float16
attn_implementation: eager

# Data training arguments
dataset_name: open-r1/OpenR1-Math-220k
dataset_configs:
- default
system_prompt: "You are a helpful AI Assistant that provides well-reasoned and detailed responses. You first think about the reasoning process as an internal monologue and then provide the user with the answer. Respond in the following format: <think>\n...\n</think>\n<answer>\n...\n</answer>"

# GRPO trainer config
bf16: true
use_vllm: true
vllm_device: auto
vllm_dtype: half
vllm_gpu_memory_utilization: 0.7
vllm_enable_prefix_caching: true
do_eval: false
gradient_accumulation_steps: 4
gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
# hub_model_id: Qwen2.5-1.5B-Open-R1-GRPO
# hub_strategy: every_save
learning_rate: 2.0e-05
log_completions: true
log_level: info
logging_first_step: false
logging_steps: 100
logging_strategy: steps
lr_scheduler_type: cosine
max_prompt_length: 512
max_completion_length: 1024
max_steps: -1
num_generations: 7
num_train_epochs: 1
output_dir: data/Qwen2.5-1.5B-Open-R1-GRPO
overwrite_output_dir: true
per_device_eval_batch_size: 1
per_device_train_batch_size: 1
push_to_hub: false
report_to:
- none
reward_funcs:
- accuracy
- format
reward_weights:
- 1.0
- 1.0
save_strategy: "epoch"
save_total_limit: 1
seed: 42
warmup_ratio: 0.1

with vllm_enable_prefix_cache: true (default):

with vllm_enable_prefix_cache: false:

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a GitHub issue? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

cc @qgallouedec

[qgallouedec]

Thanks for the contribution. Can you just please update the doc and we're good to merge

[ji-huazhong]

@qgallouedec Made the suggested change. :)