[Refactor] image data process in bench_serving #6879
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
- Modifies DatasetRow to include optional image data. - Updates the sampling logic for the MMMU dataset to populate the image data field. - Adjusts the benchmark function to pass the image data. - Handles potential tokenizer errors gracefully. - Ensures InternVL processor can handle image data as a list. Signed-off-by: Xinyuan Tong <justinning0323@outlook.com>
There was a problem hiding this comment.
Hello @JustinTong0323, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!
Summary of Changes
Hello team, gemini-code-assist here to provide a summary of this pull request. This PR focuses on refactoring how image data is handled within the bench_serving.py script, primarily to improve support for Vision-Language Models (VLMs) and address a specific bug related to the InternVL processor. The core change involves separating the image data from the prompt string itself within the benchmark dataset structure and request handling logic. This allows the benchmark to pass image data explicitly, which is a more robust approach for VLM benchmarking.
Highlights
- Data Structure Update: The
DatasetRowclass inbench_serving.pyis updated to include an optionalimage_datafield, allowing image information (specifically base64 encoded strings for MMMU) to be stored alongside the text prompt and lengths. - MMMU Data Loading Refinement: The
sample_mmmu_requestsfunction is modified to populate the newimage_datafield in theDatasetRowobjects. It also removes the previous approach of embedding the base64 image string directly into the prompt using<image>tags. - Benchmark Request Handling: The
limited_request_func(used for both warmup and main benchmark loops) is updated to retrieve the image data directly from theDatasetRowobject and pass it as a separate argument (image_data) to theRequestFuncInput, rather than parsing it out of the prompt string. - Tokenizer Compatibility Fallback: A
try...exceptblock is added insample_mmmu_requestswhen applying the chat template. This handles cases where a tokenizer might not support the standard list-based content format for multimodal inputs, falling back to the<image>tag format for compatibility (noted as a workaround for models like InternVL). - InternVL Processor Fix: The
InternVLMmProcessorininternvl.pyis updated to explicitly ensure that theimage_datait receives is a list, adding a check and conversion if it's a single string. This resolves a potential type issue when processing image data.
Changelog
Click here to see the changelog
- python/sglang/bench_serving.py
- Removed a commented-out print statement (Diff 1).
- Added
image_data: Optional[str] = Noneto theDatasetRowclass (Diff 2). - In
sample_mmmu_requests, changed variable nameimage_pathtoimage_data(Diff 3). - In
sample_mmmu_requests, removed embedding the image data string within<image>tags in the prompt (Diff 3). - In
sample_mmmu_requests, added atry...exceptblock fortokenizer.apply_chat_templateto handle tokenizers that don't support list content, falling back to the<image>tag format (Diff 3). - In
sample_mmmu_requests, updated prompt length calculation to only count text tokens, removing the estimated 512 tokens for the image (Diff 3). - In
sample_mmmu_requests, addedimage_data=image_datawhen creatingDatasetRowinstances (Diff 3). - In
limited_request_func(warmup and main loop), removed the logic that parsed the<image>tag from the prompt string to extract image data (Diff 4, Diff 5). - In
limited_request_func(warmup and main loop), updated the creation ofRequestFuncInputto passimage_datadirectly from theDatasetRowobject (Diff 4, Diff 5).
- python/sglang/srt/managers/multimodal_processors/internvl.py
- Added a check in
process_mm_data_asyncto ensureimage_datais a list, converting it if it's a single string, to fix compatibility with the InternVL processor (Diff 1).
- Added a check in
Using Gemini Code Assist
The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.
Invoking Gemini
You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| Feature | Command | Description |
|---|---|---|
| Code Review | /gemini review |
Performs a code review for the current pull request in its current state. |
| Pull Request Summary | /gemini summary |
Provides a summary of the current pull request in its current state. |
| Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in issue comments and review comments. |
| Help | /gemini help |
Displays a list of available commands. |
Customization
To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.
Limitations & Feedback
Gemini Code Assist is currently in preview and may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments to provide feedback.
You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.
Image data flows,
No longer tagged within text,
Benchmark runs clearer.
Footnotes
-
Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩
There was a problem hiding this comment.
Code Review
This pull request significantly refactors how image data is handled in bench_serving.py, making the process cleaner and more robust for Vision Language Models (VLMs). It also includes a direct fix for the InternVL processor. The changes are well-aligned with the PR's objectives.
Summary of Findings
- Change in
prompt_lenCalculation: The method for calculatingprompt_leninsample_mmmu_requestshas changed from including an estimated image token count to representing only the textual part of the prompt. Clarification on this new definition is suggested. - Pending Unit Tests and Documentation: The PR checklist indicates that unit tests and documentation updates are not yet completed. These would be valuable additions, especially given the changes to data structures and processing logic, to ensure correctness, prevent regressions, and aid understanding.
- Logging in
sample_mmmu_requests: Thesample_mmmu_requestsfunction usesprintfor reporting errors whentokenizer.apply_chat_templatefails. For library code or benchmark tools, using theloggingmodule is generally preferred for better control over message formatting and destination.
Merge Readiness
The pull request introduces valuable refactoring and fixes. However, before merging, it would be beneficial to address the clarification regarding prompt_len calculation and to complete the pending unit tests and documentation updates as indicated in the PR checklist. These additions will enhance the robustness and maintainability of the changes. I am unable to approve the pull request myself; please ensure these points are considered and the code is reviewed and approved by others before merging.
mickqian
reviewed
Jun 6, 2025
| @@ -175,6 +175,10 @@ async def process_mm_data_async( | |||
| if not image_data: | |||
| return None | |||
|
|
|||
| # Ensure image_data is a list | |||
| if isinstance(image_data, str): | |||
| image_data = [image_data] | |||
There was a problem hiding this comment.
Is it a common requirement for all models, or just this one?
There was a problem hiding this comment.
Good question, I think it's for all models. Would you think it's better to move the logic to tokenizer_manager? If so, I could do this in a separate PR to avoid expanding this PR's scope.
There was a problem hiding this comment.
By the way, an unrelated question: Some models require audio_data and video_data. I am contemplating refactoring the interface in the future, such as consolidating all into a single multimodal_data. Do you consider this to be a sound approach?
There was a problem hiding this comment.
Good question, I think it's for all models. Would you think it's better to move the logic to tokenizer_manager? If so, I could do this in a separate PR to avoid expanding this PR's scope.
Yes, we can use a new type MultimodalData ( choose a better name to be distinguishable from MultimodalDataItem ), representing the data from Req, validating the multimodal data types, then modify the parameter of load_mm_data.
mickqian
approved these changes
Jun 6, 2025
lifuhuang
reviewed
Jun 7, 2025
| except Exception as e: | ||
| # Note (Xinyuan): This is a workaround for an issue where some tokenizers do not support content as a list. (e.g. InternVL) | ||
| print(f"Error applying chat template: {e}, fallback to <image> tag") | ||
| prompt = f"<image>{prompt}" |
There was a problem hiding this comment.
(non-blocking) what if the model has a different image token than <image>?
jianan-gu
pushed a commit
to jianan-gu/sglang
that referenced
this pull request
Jun 12, 2025
Signed-off-by: Xinyuan Tong <justinning0323@outlook.com>
walker-ai
pushed a commit
to walker-ai/sglang
that referenced
this pull request
Jul 8, 2025
Merge branch 'sgl_20250610_sync_tag047 of git@code.alipay.com:Theta/SGLang.git into main https://code.alipay.com/Theta/SGLang/pull_requests/52 Reviewed-by: 剑川 <jianchuan.gys@antgroup.com> * [Bugfix] Fix slice operation when chunk size mismatch (sgl-project#6697) * [Bugfix] Fix ChatCompletion endpoint of mini_lb when stream is set (sgl-project#6703) * [CI] Fix setup of disaggregation with different tp (sgl-project#6706) * [PD] Remove Unnecessary Exception Handling for FastQueue.get() (sgl-project#6712) * Fuse routed_scaling_factor in DeepSeek (sgl-project#6710) * Overlap two kernels in DeepSeek with communication (sgl-project#6711) * Minor refactor two-batch overlap (sgl-project#6682) * Speed up when having padding tokens two-batch overlap (sgl-project#6668) * [Feature] Support Flashinfer fp8 blockwise GEMM kernel on Blackwell (sgl-project#6479) * Fix LoRA bench (sgl-project#6719) * temp * Fix PP for Qwen3 MoE (sgl-project#6709) * [feat] triton kernel for get_last_loc (sgl-project#6676) * [fix] more mem for draft_extend cuda_graph (sgl-project#6726) * [PD] bug fix: Update status if nixl receiver send a a dummy req. (sgl-project#6720) * Tune memory arguments on B200 (sgl-project#6718) * Add DeepSeek-R1-0528 function call chat template (sgl-project#6725) * refactor(tool call): Fix BaseFormatDetector tool_index issue and refactor `parse_streaming_increment` (sgl-project#6715) * Add draft extend CUDA graph for Triton backend (sgl-project#6705) * refactor apply_w8a8_block_fp8_linear in fp (sgl-project#6545) * [PD] Support completion endpoint (sgl-project#6729) * PD Rust LB (PO2) (sgl-project#6437) * Super tiny enable sole usage of expert distribution metrics and update doc (sgl-project#6680) * Support picking variants of EPLB algorithms (sgl-project#6728) * Support tuning DeepEP configs (sgl-project#6742) * [test] add ut and bm for get_last_loc (sgl-project#6746) * Fix mem_fraction_static for AMD CI (sgl-project#6748) * [fix][RL] Fix DeepSeekV3ForCausalLM.post_load_weights for multiple update weight (sgl-project#6265) * Improve EPLB logical to physical dispatch map (sgl-project#6727) * Update DeepSeek-R1-0528 function call chat template (sgl-project#6765) * [PD] Optimize time out logic and add env var doc for mooncake (sgl-project#6761) * Fix aiohttp 'Chunk too big' in bench_serving (sgl-project#6737) * Support sliding window in triton backend (sgl-project#6509) * Fix shared experts fusion error (sgl-project#6289) * Fix one bug in the grouped-gemm triton kernel (sgl-project#6772) * update llama4 chat template and pythonic parser (sgl-project#6679) * feat(tool call): Enhance Llama32Detector for improved JSON parsing in non-stream (sgl-project#6784) * Support token-level quantization for EP MoE (sgl-project#6782) * Temporarily lower mmlu threshold for triton sliding window backend (sgl-project#6785) * ci: relax test_function_call_required (sgl-project#6786) * Add intel_amx backend for Radix Attention for CPU (sgl-project#6408) * Fix incorrect LoRA weight loading for fused gate_up_proj (sgl-project#6734) * fix(PD-disaggregation): Can not get local ip (sgl-project#6792) * [FIX] mmmu bench serving result display error (sgl-project#6525) (sgl-project#6791) * Bump torch to 2.7.0 (sgl-project#6788) * chore: bump sgl-kernel v0.1.5 (sgl-project#6794) * Improve profiler and integrate profiler in bench_one_batch_server (sgl-project#6787) * chore: upgrade sgl-kernel v0.1.5 (sgl-project#6795) * [Minor] Always append newline after image token when parsing chat message (sgl-project#6797) * Update CI tests for Llama4 models (sgl-project#6421) * [Feat] Enable PDL automatically on Hopper architecture (sgl-project#5981) * chore: update blackwell docker (sgl-project#6800) * misc: cache is_hopper_arch (sgl-project#6799) * Remove contiguous before Flashinfer groupwise fp8 gemm (sgl-project#6804) * Correctly abort the failed grammar requests & Improve the handling of abort (sgl-project#6803) * [EP] Add cuda kernel for moe_ep_pre_reorder (sgl-project#6699) * Add draft extend CUDA graph for flashinfer backend (sgl-project#6805) * Refactor CustomOp to avoid confusing bugs (sgl-project#5382) * Tiny log prefill time (sgl-project#6780) * Tiny fix EPLB assertion about rebalancing period and recorder window size (sgl-project#6813) * Add simple utility to dump tensors for debugging (sgl-project#6815) * Fix profiles do not have consistent names (sgl-project#6811) * Speed up rebalancing when using non-static dispatch algorithms (sgl-project#6812) * [1/2] Add Kernel support for Cutlass based Fused FP4 MoE (sgl-project#6093) * [Router] Fix k8s Service Discovery (sgl-project#6766) * Add CPU optimized kernels for topk and rope fusions (sgl-project#6456) * fix new_page_count_next_decode (sgl-project#6671) * Fix wrong weight reference in dynamic EPLB (sgl-project#6818) * Minor add metrics to expert location updater (sgl-project#6816) * [Refactor] Rename `n_share_experts_fusion` as `num_fused_shared_experts` (sgl-project#6735) * [FEAT] Add transformers backend support (sgl-project#5929) * [fix] recover auto-dispatch for rmsnorm and rope (sgl-project#6745) * fix ep_moe_reorder kernel bugs (sgl-project#6858) * [Refactor] Multimodal data processing for VLM (sgl-project#6659) * Decoder-only Scoring API (sgl-project#6460) * feat: add dp-rank to KV events (sgl-project#6852) * Set `num_fused_shared_experts` as `num_shared_experts` when shared_experts fusion is not disabled (sgl-project#6736) * Fix one missing arg in DeepEP (sgl-project#6878) * Support LoRA in TestOpenAIVisionServer and fix fused kv_proj loading bug. (sgl-project#6861) * support 1 shot allreduce in 1-node and 2-node using mscclpp (sgl-project#6277) * Fix Qwen3MoE missing token padding optimization (sgl-project#6820) * Tiny update error hints (sgl-project#6846) * Support layerwise rebalancing experts (sgl-project#6851) * Tiny allow profiler API to auto create directory (sgl-project#6865) * Support Blackwell DeepEP docker images (sgl-project#6868) * [EP] Add cuda kernel for moe_ep_post_reorder (sgl-project#6837) * [theta]merge 0605 * oai: fix openAI client error with single request via batch api (sgl-project#6170) * [PD] Fix potential perf spike caused by tracker gc and optimize doc (sgl-project#6764) * Use deepgemm instead of triton for fused_qkv_a_proj_with_mqa (sgl-project#6890) * [CUTLASS-FP4-MOE] Introduce CutlassMoEParams class for easy initialization of Cutlass Grouped Gems Metadata (sgl-project#6887) * bugfix(OAI): Fix image_data processing for jinja chat templates (sgl-project#6877) * [CPU] enable CI for PRs, add Dockerfile and auto build task (sgl-project#6458) * AITER backend extension and workload optimizations (sgl-project#6838) * [theta]merge * [theta]merge * [Feature] Support Flashinfer fmha on Blackwell (sgl-project#6930) * Fix a bug in abort & Improve docstrings for abort (sgl-project#6931) * Tiny support customize DeepEP max dispatch tokens per rank (sgl-project#6934) * Sync the changes on cuda graph runners (sgl-project#6932) * [PD] Optimize transfer queue forward logic for dummy rank (sgl-project#6922) * [Refactor] image data process in bench_serving (sgl-project#6879) * [fix] logical_to_all_physical_map index 256 is out of bounds in EP parallel. (sgl-project#6767) * Add triton fused moe kernel config for E=257 on B200 (sgl-project#6939) * [sgl-kernel] update deepgemm (sgl-project#6942) * chore: bump sgl-kernel v0.1.6 (sgl-project#6943) * Minor compile fused topk (sgl-project#6944) * [Bugfix] pipeline parallelism and Eagle Qwen2 (sgl-project#6910) * Tiny re-introduce profile id logging (sgl-project#6912) * Add triton version as a fused_moe_triton config search key to avoid performace decrease in different Triton version (sgl-project#5955) * reduce torch.zeros overhead in moe align block size kernel (sgl-project#6369) * chore: upgrade sgl-kernel v0.1.6 (sgl-project#6945) * add fbgemm moe grouped gemm kernel benchmark (sgl-project#6924) * [Docker] Add docker file for SGL Router (sgl-project#6915) * Disabling mixed chunked prefill when eagle is enabled (sgl-project#6874) * Add canary for EPLB rebalancing (sgl-project#6895) * Refactor global_server_args_dict (sgl-project#6866) * Fuse routed scaling factor in topk_reduce kernel (sgl-project#6220) * Update server timeout time in AMD CI. (sgl-project#6953) * [misc] add is_cpu() (sgl-project#6950) * Add H20 fused MoE kernel tuning configs for DeepSeek-R1/V3 (sgl-project#6885) * Add a CUDA kernel for fusing mapping and weighted sum for MoE. (sgl-project#6916) * chore: bump sgl-kernel v0.1.6.post1 (sgl-project#6955) * chore: upgrade sgl-kernel v0.1.6.post1 (sgl-project#6957) * [DeepseekR1-FP4] Add Support for nvidia/DeepSeekR1-FP4 model (sgl-project#6853) * Revert "Fuse routed scaling factor in topk_reduce kernel (sgl-project#6220)" (sgl-project#6968) * [AMD] Add more tests to per-commit-amd (sgl-project#6926) * chore: bump sgl-kernel v0.1.7 (sgl-project#6963) * Slightly improve the sampler to skip unnecessary steps (sgl-project#6956) * rebase h20 fused_moe config (sgl-project#6966) * Fix CI and triton moe Configs (sgl-project#6974) * Remove unnecessary kernels of num_token_non_padded (sgl-project#6965) * Extend cuda graph capture bs for B200 (sgl-project#6937) * Fuse routed scaling factor in deepseek (sgl-project#6970) * Sync cuda graph runners (sgl-project#6976) * Fix draft extend ut stability with flush cache (sgl-project#6979) * Fix triton sliding window test case (sgl-project#6981) * Fix expert distribution dumping causes OOM (sgl-project#6967) * Minor remove one kernel for DeepSeek (sgl-project#6977) * [perf][sgl-kernel] extend cutlass_mla_decode to support num_head < 128 (sgl-project#6929) * Enable more unit tests for AMD CI. (sgl-project#6983) * Use torch.compile to fuse flash attention decode metadata preparation (sgl-project#6973) * Eliminate stream sync to speed up LoRA batch init (sgl-project#6960) * support qwen3 emebedding (sgl-project#6990) * Fix torch profiler bugs for bench_offline_throughput.py (sgl-project#6557) * chore: upgrade flashinfer v0.2.6.post1 jit (sgl-project#6958) * cleanup tmp dir (sgl-project#7007) * chore: update pr test xeon (sgl-project#7008) * Fix cutlass MLA gets almost zero accuracy (sgl-project#6998) * Update amd nightly models CI. (sgl-project#6992) * feat: add direct routing strategy to DP worker (sgl-project#6884) * Fallback to lower triton version for unfound fused moe configs (sgl-project#7013) * Fix torchvision version for Blackwell (sgl-project#7015) * Simplify prepare_extend_after_decode (sgl-project#6987) * Migrate to assertEqual (sgl-project#6741) * Fix torch version in blackwell dockerfile (sgl-project#7017) * chore: update pr test xeon (sgl-project#7018) * Update default settings for blackwell (sgl-project#7023) * Support both approximate and exact expert distribution collection (sgl-project#6964) * Add decode req pool (sgl-project#6980) * [theta]merge 0610 * [theta]merge 0610 * [CI] Add CI workflow for sgl-router docker build (sgl-project#7027) * Fix fused_moe triton configs (sgl-project#7029) * CPU: map changes from developing branch in sgl-kernel (sgl-project#6833) * chore: bump v0.4.7 (sgl-project#7038) * Update README.md (sgl-project#7040)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Motivation
Enhance bench_serving for VLM.
Also fix a bug in InternVL
Modifications
Checklist