[sglang] fix: Qwen VLM Baseline

examples/grpo_trainer/run_qwen2_5_vl-7b-sglang.sh

@@ -0,0 +1,48 @@
+set -x
+
+# python examples/data_preprocess/geo3k.py --local_dir ~/data/geo3k
+
+python3 -m verl.trainer.main_ppo \
+    algorithm.adv_estimator=grpo \
+    data.train_files=$HOME/data/geo3k/train.parquet \
+    data.val_files=$HOME/data/geo3k/test.parquet \
+    data.train_batch_size=512 \
+    data.max_prompt_length=1024 \
+    data.max_response_length=2048 \
+    data.filter_overlong_prompts=True \
+    data.truncation='error' \
+    data.image_key=images \
+    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-7B-Instruct \
+    actor_rollout_ref.actor.optim.lr=1e-6 \
+    actor_rollout_ref.model.use_remove_padding=True \
+    actor_rollout_ref.actor.ppo_mini_batch_size=128 \
+    actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=10 \
+    actor_rollout_ref.actor.use_kl_loss=True \
+    actor_rollout_ref.actor.kl_loss_coef=0.01 \
+    actor_rollout_ref.actor.kl_loss_type=low_var_kl \
+    actor_rollout_ref.actor.entropy_coeff=0 \
+    actor_rollout_ref.model.enable_gradient_checkpointing=True \
+    actor_rollout_ref.actor.fsdp_config.param_offload=False \
+    actor_rollout_ref.actor.fsdp_config.optimizer_offload=False \
+    actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=20 \
+    actor_rollout_ref.rollout.tensor_model_parallel_size=1 \
+    actor_rollout_ref.rollout.name=sglang \
+    +actor_rollout_ref.rollout.engine_kwargs.vllm.disable_mm_preprocessor_cache=True \
+    actor_rollout_ref.rollout.gpu_memory_utilization=0.85 \
+    actor_rollout_ref.rollout.multi_stage_wake_up=True \
+    actor_rollout_ref.rollout.enable_chunked_prefill=False \
+    actor_rollout_ref.rollout.enforce_eager=False \
+    actor_rollout_ref.rollout.free_cache_engine=True \
+    actor_rollout_ref.rollout.n=5 \
+    actor_rollout_ref.ref.log_prob_micro_batch_size_per_gpu=20 \
+    actor_rollout_ref.ref.fsdp_config.param_offload=True \
+    algorithm.use_kl_in_reward=False \
+    trainer.critic_warmup=0 \
+    trainer.logger='["console","wandb"]' \
+    trainer.project_name='verl_grpo_example_geo3k' \
+    trainer.experiment_name='qwen2_5_vl_7b_function_rm' \
+    trainer.n_gpus_per_node=8 \
+    trainer.nnodes=1 \
+    trainer.save_freq=20 \
+    trainer.test_freq=5 \
+    trainer.total_epochs=15 $@

verl/models/transformers/qwen2_vl.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 
 import inspect
+import logging
 import os
 from dataclasses import dataclass
 from typing import Optional
@@ -36,13 +37,18 @@
     validate_ulysses_config,
 )
 
+logger = logging.getLogger(__file__)
+logger.setLevel(os.getenv("VERL_LOGGING_LEVEL", "WARN"))
+
 try:
     from transformers.modeling_flash_attention_utils import flash_attn_func, flash_attn_varlen_func
 
     _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters)
 except ImportError:
-    flash_attn_varlen_func = None
+    # Fallback: try to import from flash_attn package directly
+    from flash_attn import flash_attn_varlen_func
 
+    flash_attn_func = None
     _flash_supports_window_size = None
 
 
@@ -193,7 +199,12 @@ def flash_attention_forward(
             deterministic = os.environ.get("FLASH_ATTENTION_DETERMINISTIC", "0") == "1"
         flash_kwargs["deterministic"] = deterministic
 
-    if position_ids is not None and query_length != 1 and not (torch.diff(position_ids[0], dim=-1) >= 0).all():
+    if (
+        flash_attn_varlen_func is not None
+        and position_ids is not None
+        and query_length != 1
+        and not (torch.diff(position_ids[0], dim=-1) >= 0).all()
+    ):
         batch_size = query_states.size(0)
         query_states, key_states, value_states, _, cu_seq_lens, max_seq_lens = prepare_fa2_from_position_ids(
             query_states, key_states, value_states, position_ids[0]
@@ -215,6 +226,16 @@ def flash_attention_forward(
         )
         attn_output = attn_output.view(batch_size, -1, attn_output.size(-2), attn_output.size(-1))
     else:
+        if (
+            flash_attn_varlen_func is None
+            and position_ids is not None
+            and query_length != 1
+            and not (torch.diff(position_ids[0], dim=-1) >= 0).all()
+        ):
+            logger.warning_once(
+                "flash_attn_varlen_func is not available; falling back to _flash_attention_forward."
+                "This may be suboptimal for non-monotonic position_ids in VLM mRoPE."
+            )
         attn_output = _flash_attention_forward(
             query_states,
             key_states,
@@ -226,7 +247,7 @@ def flash_attention_forward(
             use_top_left_mask=flash_attn_supports_top_left_mask(),
             deterministic=deterministic,
             **kwargs,
-        )  # do not pass position_ids to old flash_attention_forward
+        )
 
     return attn_output