[DO NOT MERGE] Upstream codebase diff #470
Draft
Conversation
This file contains 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
Removes unnecessary mark step from MoE OP loop to speed up computation
Signed-off-by: Chendi.Xue <[email protected]>
Signed-off-by: Chendi.Xue <[email protected]>
Work around for allocation error while loading llama-405b.
This bugfix addresses incorrect lower boundary handling for bucketing Previous behavior: ``` INFO 09-03 19:36:28 habana_model_runner.py:564] Prompt bucket config (min, step, max_warmup) bs:[64, 32, 64], seq:[768, 128, 768] INFO 09-03 19:36:28 habana_model_runner.py:577] Generated 12 prompt buckets: [(32, 128), (32, 256), (32, 384), (32, 512), (32, 640), (32, 768), (64, 128), (64, 256), (64, 384), (64, 512), (64, 640), (64, 768)] INFO 09-03 19:36:28 habana_model_runner.py:582] Omitted 0 prompt buckets due to exceeded token budget (max_num_batched_tokens=131072) INFO 09-03 19:36:28 habana_model_runner.py:590] Decode bucket config (min, step, max_warmup) bs:[64, 128, 64], seq:[768, 128, 1024] INFO 09-03 19:36:28 habana_model_runner.py:601] Generated 8 decode buckets: [(64, 128), (64, 256), (64, 384), (64, 512), (64, 640), (64, 768), (64, 896), (64, 1024)] INFO 09-03 19:36:28 habana_model_runner.py:606] Omitted 0 decode buckets due to exceeded token budget (max_num_batched_tokens=131072) ``` Min seq len dimension is set to 768, but buckets with seq_len=128-768 are present Current behavior: ``` INFO 09-03 19:45:42 habana_model_runner.py:563] Prompt bucket config (min, step, max_warmup) bs:[64, 32, 64], seq:[768, 128, 768] INFO 09-03 19:45:42 habana_model_runner.py:576] Generated 1 prompt buckets: [(64, 768)] INFO 09-03 19:45:42 habana_model_runner.py:581] Omitted 0 prompt buckets due to exceeded token budget (max_num_batched_tokens=131072) INFO 09-03 19:45:42 habana_model_runner.py:589] Decode bucket config (min, step, max_warmup) bs:[64, 128, 64], seq:[768, 128, 1024] INFO 09-03 19:45:42 habana_model_runner.py:600] Generated 3 decode buckets: [(64, 768), (64, 896), (64, 1024)] INFO 09-03 19:45:42 habana_model_runner.py:605] Omitted 0 decode buckets due to exceeded token budget (max_num_batched_tokens=131072) ``` No bucket with seq_len < 768 is captured
Signed-off-by: Chendi.Xue <[email protected]>
This PR prevents max_num_batched_tokens from limiting decode buckets, as decode buckets should be limited by number of blocks, not by max_num_batched_tokens.
Ports #97 to habana_main
Refactors BGMV implementation from gather based to mask-based to optimize performance and reduce device memory usage.
Use all possible slot values for dummy blocks to avoid caching issues.
With PT_COMPILE_ONLY_MODE flag, graphs can be compiled without performing synLaunch. The flag has been added to the warmup phase to decrease its execution time.
This fixes a very silly issue where mismatching values of `warmup_mode` flag could cause graph recompilations and eventually memory leaks.
This PR fixes crashes observed on older Synapse builds introduced with #227. Setting PT_COMPILE_ONLY_MODE is not supported in current or older public Synapse builds, but we should not crash because of it, rather we should advise user to use the latest build. Previous behavior: ``` ... INFO 09-06 17:08:37 habana_executor.py:85] # HPU blocks: 10761, # CPU blocks: 910 INFO 09-06 17:08:37 habana_worker.py:201] Initializing cache engine took 47.29 GiB of device memory (54.34 GiB/94.62 GiB used) and -159.6 MiB of host memory (414.9 GiB/1007 GiB used) [rank0]: Traceback (most recent call last): [rank0]: File "/software/users/kzawora/vllm-utils/vllm_hpu_simple_test.py", line 9, in <module> [rank0]: llm = LLM(model="facebook/opt-125m") [rank0]: File "/software/users/kzawora/vllm-fork/vllm/entrypoints/llm.py", line 155, in __init__ [rank0]: self.llm_engine = LLMEngine.from_engine_args( [rank0]: File "/software/users/kzawora/vllm-fork/vllm/engine/llm_engine.py", line 456, in from_engine_args [rank0]: engine = cls( [rank0]: File "/software/users/kzawora/vllm-fork/vllm/engine/llm_engine.py", line 266, in __init__ [rank0]: self._initialize_kv_caches() [rank0]: File "/software/users/kzawora/vllm-fork/vllm/engine/llm_engine.py", line 378, in _initialize_kv_caches [rank0]: self.model_executor.initialize_cache(num_gpu_blocks, num_cpu_blocks) [rank0]: File "/software/users/kzawora/vllm-fork/vllm/executor/habana_executor.py", line 89, in initialize_cache [rank0]: self.driver_worker.initialize_cache(num_gpu_blocks, num_cpu_blocks) [rank0]: File "/software/users/kzawora/vllm-fork/vllm/worker/habana_worker.py", line 202, in initialize_cache [rank0]: self._warm_up_model() [rank0]: File "/software/users/kzawora/vllm-fork/vllm/worker/habana_worker.py", line 220, in _warm_up_model [rank0]: self.model_runner.warmup_model(self.hpu_cache[0]) [rank0]: File "/usr/local/lib/python3.10/dist-packages/torch/utils/_contextlib.py", line 115, in decorate_context [rank0]: return func(*args, **kwargs) [rank0]: File "/software/users/kzawora/vllm-fork/vllm/worker/habana_model_runner.py", line 1412, in warmup_model [rank0]: with compile_only_mode_context(): [rank0]: File "/usr/lib/python3.10/contextlib.py", line 135, in __enter__ [rank0]: return next(self.gen) [rank0]: File "/usr/local/lib/python3.10/dist-packages/habana_frameworks/torch/internal/bridge_config.py", line 20, in env_setting [rank0]: get_func = globals()['get_' + var.lower()] [rank0]: KeyError: 'get_pt_compile_only_mode' inc shutdown inc shutdown inc shutdown inc shutdown ``` Current behavior: ``` ... INFO 09-06 17:06:42 habana_executor.py:85] # HPU blocks: 10761, # CPU blocks: 910 INFO 09-06 17:06:43 habana_worker.py:201] Initializing cache engine took 47.29 GiB of device memory (54.34 GiB/94.62 GiB used) and -143.7 MiB of host memory (415 GiB/1007 GiB used) WARNING 09-06 17:06:43 habana_model_runner.py:1419] Cannot use PT_COMPILE_ONLY_MODE. Warmup time will be negatively impacted. Please update Gaudi Software Suite. INFO 09-06 17:06:43 habana_model_runner.py:1336] [Warmup][Prompt][1/23] batch_size:2 seq_len:1024 free_mem:40.28 GiB ... ```
Fixes serving mode issue; due to error in fastapi
This PR contains mask based BGMV implementation for LoRA embedding instead of index-select of LoRA-B weights. Removing special handling in no LoRA case also.
Eliminate two graph breaks for torch.compile mode:
1. [__graph_breaks] torch._dynamo.exc.Unsupported: builtin: eq [<class
'torch._dynamo.variables.misc.GetAttrVariable'>, <class
'torch._dynamo.variables.constant.EnumVariable'>] False
2. [__graph_breaks] torch._dynamo.exc.Unsupported: Tensor.item
---
<details>
<!-- inside this <details> section, markdown rendering does not work, so
we use raw html here. -->
<summary><b> PR Checklist (Click to Expand) </b></summary>
<p>Thank you for your contribution to vLLM! Before submitting the pull
request, please ensure the PR meets the following criteria. This helps
vLLM maintain the code quality and improve the efficiency of the review
process.</p>
<h3>PR Title and Classification</h3>
<p>Only specific types of PRs will be reviewed. The PR title is prefixed
appropriately to indicate the type of change. Please use one of the
following:</p>
<ul>
<li><code>[Bugfix]</code> for bug fixes.</li>
<li><code>[CI/Build]</code> for build or continuous integration
improvements.</li>
<li><code>[Doc]</code> for documentation fixes and improvements.</li>
<li><code>[Model]</code> for adding a new model or improving an existing
model. Model name should appear in the title.</li>
<li><code>[Frontend]</code> For changes on the vLLM frontend (e.g.,
OpenAI API server, <code>LLM</code> class, etc.) </li>
<li><code>[Kernel]</code> for changes affecting CUDA kernels or other
compute kernels.</li>
<li><code>[Core]</code> for changes in the core vLLM logic (e.g.,
<code>LLMEngine</code>, <code>AsyncLLMEngine</code>,
<code>Scheduler</code>, etc.)</li>
<li><code>[Hardware][Vendor]</code> for hardware-specific changes.
Vendor name should appear in the prefix (e.g.,
<code>[Hardware][AMD]</code>).</li>
<li><code>[Misc]</code> for PRs that do not fit the above categories.
Please use this sparingly.</li>
</ul>
<p><strong>Note:</strong> If the PR spans more than one category, please
include all relevant prefixes.</p>
<h3>Code Quality</h3>
<p>The PR need to meet the following code quality standards:</p>
<ul>
<li>We adhere to <a
href="https://google.github.io/styleguide/pyguide.html">Google Python
style guide</a> and <a
href="https://google.github.io/styleguide/cppguide.html">Google C++
style guide</a>.</li>
<li>Pass all linter checks. Please use <a
href="https://github.com/vllm-project/vllm/blob/main/format.sh"><code>format.sh</code></a>
to format your code.</li>
<li>The code need to be well-documented to ensure future contributors
can easily understand the code.</li>
<li>Include sufficient tests to ensure the project to stay correct and
robust. This includes both unit tests and integration tests.</li>
<li>Please add documentation to <code>docs/source/</code> if the PR
modifies the user-facing behaviors of vLLM. It helps vLLM user
understand and utilize the new features or changes.</li>
</ul>
<h3>Notes for Large Changes</h3>
<p>Please keep the changes as concise as possible. For major
architectural changes (>500 LOC excluding kernel/data/config/test), we
would expect a GitHub issue (RFC) discussing the technical design and
justification. Otherwise, we will tag it with <code>rfc-required</code>
and might not go through the PR.</p>
<h3>What to Expect for the Reviews</h3>
<p>The goal of the vLLM team is to be a <i>transparent reviewing
machine</i>. We would like to make the review process transparent and
efficient and make sure no contributor feel confused or frustrated.
However, the vLLM team is small, so we need to prioritize some PRs over
others. Here is what you can expect from the review process: </p>
<ul>
<li> After the PR is submitted, the PR will be assigned to a reviewer.
Every reviewer will pick up the PRs based on their expertise and
availability.</li>
<li> After the PR is assigned, the reviewer will provide status update
every 2-3 days. If the PR is not reviewed within 7 days, please feel
free to ping the reviewer or the vLLM team.</li>
<li> After the review, the reviewer will put an <code>
action-required</code> label on the PR if there are changes required.
The contributor should address the comments and ping the reviewer to
re-review the PR.</li>
<li> Please respond to all comments within a reasonable time frame. If a
comment isn't clear or you disagree with a suggestion, feel free to ask
for clarification or discuss the suggestion.
</li>
</ul>
<h3>Thank You</h3>
<p> Finally, thank you for taking the time to read these guidelines and
for your interest in contributing to vLLM. Your contributions make vLLM
a great tool for everyone! </p>
</details>
---------
Signed-off-by: yuwenzho <[email protected]>
FILL IN THE PR DESCRIPTION HERE
FIX #xxxx (*link existing issues this PR will resolve*)
**BEFORE SUBMITTING, PLEASE READ THE CHECKLIST BELOW AND FILL IN THE
DESCRIPTION ABOVE**
---
<details>
<!-- inside this <details> section, markdown rendering does not work, so
we use raw html here. -->
<summary><b> PR Checklist (Click to Expand) </b></summary>
<p>Thank you for your contribution to vLLM! Before submitting the pull
request, please ensure the PR meets the following criteria. This helps
vLLM maintain the code quality and improve the efficiency of the review
process.</p>
<h3>PR Title and Classification</h3>
<p>Only specific types of PRs will be reviewed. The PR title is prefixed
appropriately to indicate the type of change. Please use one of the
following:</p>
<ul>
<li><code>[Bugfix]</code> for bug fixes.</li>
<li><code>[CI/Build]</code> for build or continuous integration
improvements.</li>
<li><code>[Doc]</code> for documentation fixes and improvements.</li>
<li><code>[Model]</code> for adding a new model or improving an existing
model. Model name should appear in the title.</li>
<li><code>[Frontend]</code> For changes on the vLLM frontend (e.g.,
OpenAI API server, <code>LLM</code> class, etc.) </li>
<li><code>[Kernel]</code> for changes affecting CUDA kernels or other
compute kernels.</li>
<li><code>[Core]</code> for changes in the core vLLM logic (e.g.,
<code>LLMEngine</code>, <code>AsyncLLMEngine</code>,
<code>Scheduler</code>, etc.)</li>
<li><code>[Hardware][Vendor]</code> for hardware-specific changes.
Vendor name should appear in the prefix (e.g.,
<code>[Hardware][AMD]</code>).</li>
<li><code>[Misc]</code> for PRs that do not fit the above categories.
Please use this sparingly.</li>
</ul>
<p><strong>Note:</strong> If the PR spans more than one category, please
include all relevant prefixes.</p>
<h3>Code Quality</h3>
<p>The PR need to meet the following code quality standards:</p>
<ul>
<li>We adhere to <a
href="https://google.github.io/styleguide/pyguide.html">Google Python
style guide</a> and <a
href="https://google.github.io/styleguide/cppguide.html">Google C++
style guide</a>.</li>
<li>Pass all linter checks. Please use <a
href="https://github.com/vllm-project/vllm/blob/main/format.sh"><code>format.sh</code></a>
to format your code.</li>
<li>The code need to be well-documented to ensure future contributors
can easily understand the code.</li>
<li>Include sufficient tests to ensure the project to stay correct and
robust. This includes both unit tests and integration tests.</li>
<li>Please add documentation to <code>docs/source/</code> if the PR
modifies the user-facing behaviors of vLLM. It helps vLLM user
understand and utilize the new features or changes.</li>
</ul>
<h3>Notes for Large Changes</h3>
<p>Please keep the changes as concise as possible. For major
architectural changes (>500 LOC excluding kernel/data/config/test), we
would expect a GitHub issue (RFC) discussing the technical design and
justification. Otherwise, we will tag it with <code>rfc-required</code>
and might not go through the PR.</p>
<h3>What to Expect for the Reviews</h3>
<p>The goal of the vLLM team is to be a <i>transparent reviewing
machine</i>. We would like to make the review process transparent and
efficient and make sure no contributor feel confused or frustrated.
However, the vLLM team is small, so we need to prioritize some PRs over
others. Here is what you can expect from the review process: </p>
<ul>
<li> After the PR is submitted, the PR will be assigned to a reviewer.
Every reviewer will pick up the PRs based on their expertise and
availability.</li>
<li> After the PR is assigned, the reviewer will provide status update
every 2-3 days. If the PR is not reviewed within 7 days, please feel
free to ping the reviewer or the vLLM team.</li>
<li> After the review, the reviewer will put an <code>
action-required</code> label on the PR if there are changes required.
The contributor should address the comments and ping the reviewer to
re-review the PR.</li>
<li> Please respond to all comments within a reasonable time frame. If a
comment isn't clear or you disagree with a suggestion, feel free to ask
for clarification or discuss the suggestion.
</li>
</ul>
<h3>Thank You</h3>
<p> Finally, thank you for taking the time to read these guidelines and
for your interest in contributing to vLLM. Your contributions make vLLM
a great tool for everyone! </p>
</details>
---------
Co-authored-by: Michal Adamczyk <[email protected]>
Co-authored-by: barak goldberg <[email protected]>
Co-authored-by: Michal Szutenberg <[email protected]>
Co-authored-by: Jan Kaniecki <[email protected]>
RuntimeErrors are not observed anymore on habana_main when disable_tensor_cache is used. This PR enables disable_tensor_cache.
Signed-off-by: Max de Bayser <[email protected]> Signed-off-by: Max de Bayser <[email protected]> Signed-off-by: Joe Runde <[email protected]> Signed-off-by: Russell Bryant <[email protected]> Signed-off-by: Thomas Parnell <[email protected]> Signed-off-by: Russell Bryant <[email protected]> Signed-off-by: Varad Ahirwadkar <[email protected]> Signed-off-by: Wallas Santos <[email protected]> Signed-off-by: Travis Johnson <[email protected]> Signed-off-by: Rafael Vasquez <[email protected]> Signed-off-by: Yuan Zhou <[email protected]> Signed-off-by: luka <[email protected]> Signed-off-by: Alex-Brooks <[email protected]> Signed-off-by: youkaichao <[email protected]> Signed-off-by: Tyler Michael Smith <[email protected]> Signed-off-by: mgoin <[email protected]> Signed-off-by: Vinay Damodaran <[email protected]> Signed-off-by: Woosuk Kwon <[email protected]> Signed-off-by: Jee Jee Li <[email protected]> Signed-off-by: Harry Mellor <[email protected]> Signed-off-by: charlifu <[email protected]> Signed-off-by: Sam Stoelinga <[email protected]> Signed-off-by: Vasily Alexeev <[email protected]> Signed-off-by: Kevin-Yang <[email protected]> Signed-off-by: Abatom <[email protected]> Signed-off-by: Bill Nell <[email protected]> Signed-off-by: wangshuai09 <[email protected]> Signed-off-by: Qishuai [email protected] Signed-off-by: yuze.zyz <[email protected]> Signed-off-by: Yannick Schnider <[email protected]> Signed-off-by: Kunjan Patel <[email protected]> Signed-off-by: simon-mo <[email protected]> Signed-off-by: kevin <[email protected]> Signed-off-by: YiSheng5 <[email protected]> Signed-off-by: yan ma <[email protected]> Signed-off-by: Went-Liang <[email protected]> Signed-off-by: Roger Wang <[email protected]> Signed-off-by: sasha0552 <[email protected]> Signed-off-by: mzusman <[email protected]> Signed-off-by: Prashant Gupta <[email protected]> Signed-off-by: André Jonasson <[email protected]> Signed-off-by: Gene Su <[email protected]> Signed-off-by: dependabot[bot] <[email protected]> Signed-off-by: Peter Salas <[email protected]> Signed-off-by: Nick Hill <[email protected]> Signed-off-by: Nick Hill <[email protected]> Signed-off-by: Michael Green <[email protected]> Signed-off-by: Shanshan Wang <[email protected]> Signed-off-by: Gregory Shtrasberg <[email protected]> Signed-off-by: daitran2k1 <[email protected]> Signed-off-by: MengqingCao <[email protected]> Signed-off-by: chaunceyjiang <[email protected]> Signed-off-by: Robert Shaw <[email protected]> Signed-off-by: Hissu Hyvarinen <[email protected]> Signed-off-by: [email protected] <[email protected]> Signed-off-by: Linkun Chen <[email protected]> Signed-off-by: Tomer Asida <[email protected]> Signed-off-by: DarkLight1337 <[email protected]> Co-authored-by: sasha0552 <[email protected]> Co-authored-by: Woosuk Kwon <[email protected]> Co-authored-by: Li, Jiang <[email protected]> Co-authored-by: Kuntai Du <[email protected]> Co-authored-by: Daniele <[email protected]> Co-authored-by: Cyrus Leung <[email protected]> Co-authored-by: Luka Govedič <[email protected]> Co-authored-by: bnellnm <[email protected]> Co-authored-by: Kai Wu <[email protected]> Co-authored-by: Isotr0py <[email protected]> Co-authored-by: Shashwat Srijan <[email protected]> Co-authored-by: Robert Shaw <[email protected]> Co-authored-by: Andrew Feldman <[email protected]> Co-authored-by: afeldman-nm <[email protected]> Co-authored-by: laishzh <[email protected]> Co-authored-by: Max de Bayser <[email protected]> Co-authored-by: Max de Bayser <[email protected]> Co-authored-by: Dipika Sikka <[email protected]> Co-authored-by: Joe Runde <[email protected]> Co-authored-by: Haoyu Wang <[email protected]> Co-authored-by: Russell Bryant <[email protected]> Co-authored-by: Nick Hill <[email protected]> Co-authored-by: tomeras91 <[email protected]> Co-authored-by: Tyler Michael Smith <[email protected]> Co-authored-by: Michael Goin <[email protected]> Co-authored-by: Kunjan <[email protected]> Co-authored-by: Kunjan Patel <kunjanp_google_com@vllm.us-central1-a.c.kunjanp-gke-dev-2.internal> Co-authored-by: Cody Yu <[email protected]> Co-authored-by: Thomas Parnell <[email protected]> Co-authored-by: Chih-Chieh Yang <[email protected]> Co-authored-by: Yue Zhang <[email protected]> Co-authored-by: Chen Zhang <[email protected]> Co-authored-by: Andy Dai <[email protected]> Co-authored-by: Dhia Eddine Rhaiem <[email protected]> Co-authored-by: yudian0504 <[email protected]> Co-authored-by: Varad Ahirwadkar <[email protected]> Co-authored-by: youkaichao <[email protected]> Co-authored-by: Baoyuan Qi <[email protected]> Co-authored-by: Wallas Henrique <[email protected]> Co-authored-by: Travis Johnson <[email protected]> Co-authored-by: Cyrus Leung <[email protected]> Co-authored-by: ngrozae <[email protected]> Co-authored-by: Falko1 <[email protected]> Co-authored-by: Rafael Vasquez <[email protected]> Co-authored-by: chenqianfzh <[email protected]> Co-authored-by: wangshuai09 <[email protected]> Co-authored-by: Jee Jee Li <[email protected]> Co-authored-by: xendo <[email protected]> Co-authored-by: Jerzy Zagorski <[email protected]> Co-authored-by: gopalsarda <[email protected]> Co-authored-by: Yuan <[email protected]> Co-authored-by: Gubrud, Aaron D <[email protected]> Co-authored-by: adgubrud <[email protected]> Co-authored-by: Yuhong Guo <[email protected]> Co-authored-by: Yuhong Guo <[email protected]> Co-authored-by: Ronen Schaffer <[email protected]> Co-authored-by: Aurick Qiao <[email protected]> Co-authored-by: Jeremy Arnold <[email protected]> Co-authored-by: Lucas Wilkinson <[email protected]> Co-authored-by: yulei <[email protected]> Co-authored-by: Seth Kimmel <[email protected]> Co-authored-by: Kaunil Dhruv <[email protected]> Co-authored-by: Flex Wang <[email protected]> Co-authored-by: Mengqing Cao <[email protected]> Co-authored-by: Alex Brooks <[email protected]> Co-authored-by: Yongzao <[email protected]> Co-authored-by: Yunfei Chu <[email protected]> Co-authored-by: Vinay R Damodaran <[email protected]> Co-authored-by: Yan Ma <[email protected]> Co-authored-by: Zhuohan Li <[email protected]> Co-authored-by: litianjian <[email protected]> Co-authored-by: Harry Mellor <[email protected]> Co-authored-by: Charlie Fu <[email protected]> Co-authored-by: Kevin H. Luu <[email protected]> Co-authored-by: Will Johnson <[email protected]> Co-authored-by: pavlo-ruban <[email protected]> Co-authored-by: Sam Stoelinga <[email protected]> Co-authored-by: ErkinSagiroglu <[email protected]> Co-authored-by: Vasiliy Alekseev <[email protected]> Co-authored-by: kakao-kevin-us <[email protected]> Co-authored-by: Kevin-Yang <[email protected]> Co-authored-by: 科英 <[email protected]> Co-authored-by: madt2709 <[email protected]> Co-authored-by: litianjian <[email protected]> Co-authored-by: Zhong Qishuai <[email protected]> Co-authored-by: tastelikefeet <[email protected]> Co-authored-by: Sven Seeberg <[email protected]> Co-authored-by: yannicks1 <[email protected]> Co-authored-by: Junichi Sato <[email protected]> Co-authored-by: Kunjan <[email protected]> Co-authored-by: Will Eaton <[email protected]> Co-authored-by: Simon Mo <[email protected]> Co-authored-by: Lily Liu <[email protected]> Co-authored-by: YiSheng5 <[email protected]> Co-authored-by: Went-Liang <[email protected]> Co-authored-by: Elfie Guo <[email protected]> Co-authored-by: Harsha vardhan manoj Bikki <[email protected]> Co-authored-by: Guillaume Calmettes <[email protected]> Co-authored-by: Roger Wang <[email protected]> Co-authored-by: Alexei-V-Ivanov-AMD <[email protected]> Co-authored-by: Mor Zusman <[email protected]> Co-authored-by: Prashant Gupta <[email protected]> Co-authored-by: Patrick von Platen <[email protected]> Co-authored-by: André Jonasson <[email protected]> Co-authored-by: Pavani Majety <[email protected]> Co-authored-by: Gene Der Su <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Peter Salas <[email protected]> Co-authored-by: sroy745 <[email protected]> Co-authored-by: Michael Green <[email protected]> Co-authored-by: Nick Hill <[email protected]> Co-authored-by: Nikita Furin <[email protected]> Co-authored-by: shanshan wang <[email protected]> Co-authored-by: Roger Wang <[email protected]> Co-authored-by: Gregory Shtrasberg <[email protected]> Co-authored-by: Yang Zheng <[email protected]> Co-authored-by: Yang Zheng(SW)(Alex) <[email protected]> Co-authored-by: Tran Quang Dai <[email protected]> Co-authored-by: Chauncey <[email protected]> Co-authored-by: hissu-hyvarinen <[email protected]> Co-authored-by: lkchen <[email protected]> Co-authored-by: Linkun Chen <[email protected]> Co-authored-by: Linkun Chen <[email protected]> Co-authored-by: Gene Der Su <[email protected]>
Reverts #439
This PR adds all commits before vllm-project#6143 without vllm-project#6143.
vllm-project#6143 got merged, but it's based on an older revision of HPU components. This PR aligns the two.
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
| @@ -0,0 +1,35 @@ | |||
| name: cpu-test | |||
Check failure
Code scanning / Scorecard
Token-Permissions High
Accuracy fix for multi-step scheduling. This code solves the problem of a wrong second token when HPU Graphs are used. --------- Co-authored-by: Libin Tang <[email protected]>
Add multi step scheduling scenario to jenkins CI
Req - https://jira.habana-labs.com/browse/REQ-289 => target for 1.19 TODO: - There remains one hardcode to HPUWorker, need to remove Next Steps: - 1. submit necessary codes change to vllm-upstream branch => WIP - 2. support all 3 draft_model_types - mlp_speculator, medusa and others
Certain links on the gaudi-installation page which were pointing to docs.habana.ai are broken due to recent re-structuring of docs.habana.ai. This PR fixes those.
Current instructions for setup using standalone docker (not using
Dockerfile) is missing the `pip install -r requirements-hpu.txt`
instruction. A new user using this method for setup will encounter :
```
File "/root/vllm/setup.py", line 15, in <module>
from setuptools_scm import get_version
ModuleNotFoundError: No module named 'setuptools_scm'
```
This PR fixes that.
Current implementation of optimized topp/topk calculations for scalar case is handling the duplicates that are outside of kth border. Unfortunately, to analyze duplicates it is necessary to make a synchronization with CPU, what makes multi-step scheduling useless together with topp/topk. This PR adds option to skip duplicates handling with `VLLM_HANDLE_TOPK_DUPLICATES` (default `True`). When this variable is set, handling duplicates will be skipped and we will avoid synchronization with CPU. It also removes the synchronization which was done earlier in Sampler, by saving scalar value of `top_k` and `top_p`. It should give performance gain for all benchmarks with these sampling parameters, especially together with multi-step scheduling. While disabling the duplicates handling may cause small accuracy differences, the best solution will be to handle duplicates without synchronization with CPU. However, this is not a trivial problem, so I will try to provide such solution later.
This PR changes the view to `offset` tensor to (batch_size, -1) for enabling broadcasting.
Spec Decoder PR2 - enable Medusa, MLP This PR is add on to #375 => Do not merge until PR375 merged
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.
Scope of changes:
mark_steps)