[recipe] feat: Add qwen2.5-7b DAPO NPU example script (#3501)

### What does this PR do?

#1858 support DAPO on Ascend NPU, but example `qwen2.5-7b-instruct`
training script is not added, which will be added through this PR.

The script in this PR is borrowed from
https://gitee.com/ascend/ModelZoo-PyTorch/blob/master/PyTorch/built-in/rl/VeRL_for_PyTorch/test/train_qwen2_5_7b_instruct_DAPO_full_16p.sh

### 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

Not related.

### API and Usage Example

Not related.

### Design & Code Changes

Not relaetd.

### 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`
- [x] Add / Update [the
documentation](https://github.com/volcengine/verl/tree/main/docs).
- [x] 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: ...
- [x] 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).)
This commit is contained in:
Zhen
2025-09-17 16:52:28 +08:00
committed by GitHub
parent 0665153b9a
commit ee8a7af8f4
2 changed files with 143 additions and 1 deletions

View File

@ -28,7 +28,6 @@ train_prompt_bsz=128
gen_prompt_bsz=$((train_prompt_bsz * 3)) gen_prompt_bsz=$((train_prompt_bsz * 3))
n_resp_per_prompt=16 n_resp_per_prompt=16
train_prompt_mini_bsz=32 train_prompt_mini_bsz=32
now=$(date +%Y%m%d_%H%M%S)
# Ray # Ray
PWD=./ PWD=./

View File

@ -0,0 +1,143 @@
#!/usr/bin/env bash
set -xeuo pipefail
project_name='DAPO-Qwen2.5-7B-Instruct'
exp_name='DAPO-Qwen2.5-7B-Instruct'
adv_estimator=grpo
use_kl_in_reward=False
kl_coef=0.0
use_kl_loss=False
kl_loss_coef=0.0
clip_ratio_low=0.2
clip_ratio_high=0.28
max_prompt_length=$((1024 * 2))
max_response_length=$((1024 * 20))
enable_overlong_buffer=True
overlong_buffer_len=$((1024 * 4))
overlong_penalty_factor=1.0
loss_agg_mode="token-mean"
enable_filter_groups=True
filter_groups_metric=acc
max_num_gen_batches=10
NNODES=1
train_prompt_bsz=16
gen_prompt_bsz=$((train_prompt_bsz * 3))
n_resp_per_prompt=16
train_prompt_mini_bsz=1
# Ray
PWD=./
RAY_ADDRESS=${RAY_ADDRESS:-"http://localhost:8265"}
WORKING_DIR=${WORKING_DIR:-"${PWD}"}
RUNTIME_ENV=${RUNTIME_ENV:-"${WORKING_DIR}/verl/trainer/runtime_env.yaml"}
# Paths
RAY_DATA_HOME=${RAY_DATA_HOME:-"${HOME}/verl"}
MODEL_PATH=${MODEL_PATH:-"${RAY_DATA_HOME}/models/Qwen2.5-7B-Instruct"}
CKPTS_DIR=${CKPTS_DIR:-"${RAY_DATA_HOME}/ckpts/${project_name}/${exp_name}"}
TRAIN_FILE=${TRAIN_FILE:-"${RAY_DATA_HOME}/data/dapo-math-17k.parquet"}
TEST_FILE=${TEST_FILE:-"${RAY_DATA_HOME}/data/aime-2024.parquet"}
# Algorithm
temperature=1.0
top_p=1.0
top_k=-1 # 0 for HF rollout, -1 for vLLM rollout
# Performance Related Parameter
sp_size=4
use_dynamic_bsz=True
actor_ppo_max_token_len=$(((max_prompt_length + max_response_length) / sp_size))
infer_ppo_max_token_len=$(((max_prompt_length + max_response_length) / sp_size))
offload=True
gen_tp=1
ray job submit --no-wait --runtime-env="${RUNTIME_ENV}" \
--working-dir "${WORKING_DIR}" \
--address "${RAY_ADDRESS}" \
-- python3 -m recipe.dapo.main_dapo \
data.train_files="${TRAIN_FILE}" \
data.val_files="${TEST_FILE}" \
data.prompt_key=prompt \
data.truncation='left' \
data.max_prompt_length=${max_prompt_length} \
data.max_response_length=${max_response_length} \
data.gen_batch_size=${gen_prompt_bsz} \
data.train_batch_size=${train_prompt_bsz} \
actor_rollout_ref.rollout.n=${n_resp_per_prompt} \
algorithm.adv_estimator=${adv_estimator} \
algorithm.use_kl_in_reward=${use_kl_in_reward} \
algorithm.kl_ctrl.kl_coef=${kl_coef} \
actor_rollout_ref.actor.use_kl_loss=${use_kl_loss} \
actor_rollout_ref.actor.kl_loss_coef=${kl_loss_coef} \
actor_rollout_ref.actor.clip_ratio_low=${clip_ratio_low} \
actor_rollout_ref.actor.clip_ratio_high=${clip_ratio_high} \
actor_rollout_ref.actor.clip_ratio_c=10.0 \
algorithm.filter_groups.enable=${enable_filter_groups} \
algorithm.filter_groups.max_num_gen_batches=${max_num_gen_batches} \
algorithm.filter_groups.metric=${filter_groups_metric} \
actor_rollout_ref.actor.use_torch_compile=False \
actor_rollout_ref.ref.use_torch_compile=False \
actor_rollout_ref.model.use_remove_padding=True \
actor_rollout_ref.actor.use_dynamic_bsz=${use_dynamic_bsz} \
actor_rollout_ref.ref.log_prob_use_dynamic_bsz=${use_dynamic_bsz} \
actor_rollout_ref.rollout.log_prob_use_dynamic_bsz=${use_dynamic_bsz} \
actor_rollout_ref.actor.ppo_max_token_len_per_gpu=${actor_ppo_max_token_len} \
actor_rollout_ref.ref.log_prob_max_token_len_per_gpu=${infer_ppo_max_token_len} \
actor_rollout_ref.rollout.log_prob_max_token_len_per_gpu=${infer_ppo_max_token_len} \
actor_rollout_ref.rollout.name=vllm \
actor_rollout_ref.model.path="${MODEL_PATH}" \
+actor_rollout_ref.model.override_config.attention_dropout=0. \
+actor_rollout_ref.model.override_config.embd_pdrop=0. \
+actor_rollout_ref.model.override_config.resid_pdrop=0. \
actor_rollout_ref.model.enable_gradient_checkpointing=True \
actor_rollout_ref.actor.optim.lr=1e-6 \
actor_rollout_ref.actor.optim.lr_warmup_steps=10 \
actor_rollout_ref.actor.optim.weight_decay=0.1 \
actor_rollout_ref.actor.ppo_mini_batch_size=${train_prompt_mini_bsz} \
actor_rollout_ref.actor.fsdp_config.param_offload=${offload} \
actor_rollout_ref.actor.fsdp_config.optimizer_offload=${offload} \
actor_rollout_ref.actor.entropy_coeff=0 \
actor_rollout_ref.actor.grad_clip=1.0 \
actor_rollout_ref.actor.loss_agg_mode=${loss_agg_mode} \
actor_rollout_ref.actor.ulysses_sequence_parallel_size=${sp_size} \
actor_rollout_ref.rollout.gpu_memory_utilization=0.50 \
actor_rollout_ref.rollout.tensor_model_parallel_size=${gen_tp} \
actor_rollout_ref.rollout.enable_chunked_prefill=True \
actor_rollout_ref.rollout.max_num_batched_tokens=$((max_prompt_length + max_response_length)) \
actor_rollout_ref.rollout.temperature=${temperature} \
actor_rollout_ref.rollout.top_p=${top_p} \
actor_rollout_ref.rollout.top_k="${top_k}" \
actor_rollout_ref.rollout.val_kwargs.temperature=${temperature} \
actor_rollout_ref.rollout.val_kwargs.top_p=${top_p} \
actor_rollout_ref.rollout.val_kwargs.top_k=${top_k} \
actor_rollout_ref.rollout.val_kwargs.do_sample=True \
actor_rollout_ref.rollout.val_kwargs.n=1 \
actor_rollout_ref.ref.fsdp_config.param_offload=${offload} \
actor_rollout_ref.ref.ulysses_sequence_parallel_size=${sp_size} \
actor_rollout_ref.actor.fsdp_config.fsdp_size=-1 \
reward_model.reward_manager=dapo \
reward_model.overlong_buffer.enable=${enable_overlong_buffer} \
reward_model.overlong_buffer.len=${overlong_buffer_len} \
reward_model.overlong_buffer.penalty_factor=${overlong_penalty_factor} \
trainer.logger="['console']" \
trainer.project_name="${project_name}" \
trainer.experiment_name="${exp_name}" \
trainer.n_gpus_per_node=16 \
trainer.nnodes="${NNODES}" \
trainer.val_before_train=True \
trainer.test_freq=5 \
trainer.save_freq=20 \
trainer.total_epochs=1 \
trainer.default_local_dir="${CKPTS_DIR}" \
trainer.device=npu \
trainer.resume_mode=auto \
actor_rollout_ref.actor.entropy_checkpointing=True \
actor_rollout_ref.ref.entropy_checkpointing=True \
actor_rollout_ref.actor.fsdp_config.forward_prefetch=True \
actor_rollout_ref.ref.fsdp_config.forward_prefetch=True \
actor_rollout_ref.actor.entropy_from_logits_with_chunking=True \
actor_rollout_ref.ref.entropy_from_logits_with_chunking=True