Skip to content

feat: update nightly gsm8k eval #1304

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 3 commits into from
Sep 2, 2024
Merged

feat: update nightly gsm8k eval #1304

merged 3 commits into from
Sep 2, 2024

Conversation

zhyncs
Copy link
Member

@zhyncs zhyncs commented Sep 2, 2024

Motivation

Modifications

Checklist

  • Format your code according to the Contributor Guide.
  • Add unit tests as outlined in the Contributor Guide.
  • Update documentation as needed, including docstrings or example tutorials.

@zhyncs zhyncs added the wip label Sep 2, 2024
@zhyncs
Copy link
Member Author

zhyncs commented Sep 2, 2024 

# H100 TP 2, latest v0.2.15
python3 -m sglang.launch_server --model neuralmagic/Qwen2-72B-Instruct-FP8 --quantization fp8  --trust-remote-code --tp 2 --kv-cache-dtype fp8_e5m2
python3 -m sglang.bench_serving --backend sglang
Traceback (most recent call last):
[14:47:01 TP0] Exception in ModelTpServer:
Traceback (most recent call last):
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/managers/tp_worker.py", line 244, in exposed_step
    self.forward_step()
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/managers/tp_worker.py", line 260, in forward_step
    self.forward_prefill_batch(new_batch)
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/managers/tp_worker.py", line 507, in forward_prefill_batch
    sample_output, logits_output = self.model_runner.forward(
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/model_executor/model_runner.py", line 584, in forward
    return self.forward_extend(batch)
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/model_executor/model_runner.py", line 542, in forward_extend
    input_metadata = InputMetadata.from_schedule_batch(
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/model_executor/forward_batch_info.py", line 215, in from_schedule_batch
    ret.init_flashinfer_handlers(
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/model_executor/forward_batch_info.py", line 245, in init_flashinfer_handlers
    update_flashinfer_indices(
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/model_executor/forward_batch_info.py", line 374, in update_flashinfer_indices
    model_runner.flashinfer_prefill_wrapper_paged.begin_forward(
  File "/usr/local/lib/python3.10/dist-packages/flashinfer/prefill.py", line 832, in plan
    self._wrapper.plan(
RuntimeError: Failed to allocate memory for batch_prefill_tmp_v with size 599785472 and alignment 16 in AlignedAllocator

  File "/usr/lib/python3.10/multiprocessing/process.py", line 314, in _bootstrap
    self.run()
  File "/usr/lib/python3.10/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/managers/tp_worker.py", line 896, in run_tp_server
    model_server.exposed_step(recv_reqs)
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/managers/tp_worker.py", line 244, in exposed_step
    self.forward_step()
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/managers/tp_worker.py", line 260, in forward_step
    self.forward_prefill_batch(new_batch)
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/managers/tp_worker.py", line 507, in forward_prefill_batch
    sample_output, logits_output = self.model_runner.forward(
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/model_executor/model_runner.py", line 584, in forward
    return self.forward_extend(batch)
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/model_executor/model_runner.py", line 542, in forward_extend
    input_metadata = InputMetadata.from_schedule_batch(
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/model_executor/forward_batch_info.py", line 215, in from_schedule_batch
    ret.init_flashinfer_handlers(
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/model_executor/forward_batch_info.py", line 245, in init_flashinfer_handlers
    update_flashinfer_indices(
  File "/usr/local/lib/python3.10/dist-packages/sglang/srt/model_executor/forward_batch_info.py", line 374, in update_flashinfer_indices
    model_runner.flashinfer_prefill_wrapper_paged.begin_forward(
  File "/usr/local/lib/python3.10/dist-packages/flashinfer/prefill.py", line 832, in plan
    self._wrapper.plan(
RuntimeError: Failed to allocate memory for batch_prefill_tmp_v with size 599785472 and alignment 16 in AlignedAllocator

It works well without --kv-cache-dtype fp8_e5m2. @ispobock @yzh119 may help take a look

@zhyncs zhyncs removed the wip label Sep 2, 2024
zhyncs
zhyncs commented Sep 2, 2024
View reviewed changes
@zhyncs
Copy link
Member Author

zhyncs commented Sep 2, 2024

fix #1272

@zhyncs
Copy link
Member Author

zhyncs commented Sep 2, 2024

ref https://github.com/sgl-project/sglang/actions/runs/10669348477/job/29571069828

@zhyncs zhyncs self-assigned this Sep 2, 2024
@zhyncs zhyncs enabled auto-merge (squash) September 2, 2024 14:57
@zhyncs zhyncs disabled auto-merge September 2, 2024 15:18
@zhyncs zhyncs merged commit 2561ed0 into main Sep 2, 2024
10 checks passed
@zhyncs zhyncs deleted the night branch September 2, 2024 15:18
@zhyncs zhyncs mentioned this pull request Sep 2, 2024
Merged
3 tasks
@ispobock ispobock mentioned this pull request Sep 7, 2024
Closed
5 tasks
timethink pushed a commit to timethink/sglang that referenced this pull request Mar 9, 2025
@zhyncs
feat: update nightly gsm8k eval (sgl-project#1304)
f516f17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@Ying1123 Ying1123 Awaiting requested review from Ying1123

@yzh119 yzh119 Awaiting requested review from yzh119

@merrymercy merrymercy Awaiting requested review from merrymercy merrymercy is a code owner

@ispobock ispobock Awaiting requested review from ispobock

@hnyls2002 hnyls2002 Awaiting requested review from hnyls2002

Assignees

@zhyncs zhyncs

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
1 participant
@zhyncs