[Feature] Support Ernie FP8 on sm100 ( the fixed version)

[zoooo0820]

Motivation

💡 If this PR is a Cherry Pick, the PR title needs to follow the format by adding the [Cherry-Pick] label at the very beginning and appending the original PR ID at the end. For example, [Cherry-Pick][CI] Add check trigger and logic(#5191)

💡 如若此PR是Cherry Pick，PR标题需遵循格式，在最开始加上[Cherry-Pick]标签，以及最后面加上原PR ID，例如[Cherry-Pick][CI] Add check trigger and logic(#5191)

This PR is the fixed version of #5593 . The problem is ep_moe_expert_dispatch_fp8 was changed to support column-major input scale. However, the Deepep now in Paddle requires row-major, contiguous scale.

Modifications

1. Because the old version of DeepGEMM does not support sm100, we use the newer version DeepGEMM for sm100. And this dependency will be switched to PaddleFleet in the future.
2. FP8 gemm / group_gemm on sm100 will use ue8m0 scale on sm100.
3. For EP mode, FP8 DeepEP low_latency_dispatch will use ue8m0 scale on sm100.
4. For activation quantization, we use paddle.incubate.nn.functional.fp8_quant_blockwise instead of per_token_quant / per_token_quant_padding. Add support ue8m0 for masked_per_token_quant, using in EP Decoding phase.

Usage or Command

requires paddlepaddle develop after(include) 20260119

For DeepGEMM installtion: https://github.com/PFCCLab/DeepGEMM
For DeepEP installtion: https://github.com/PFCCLab/DeepEP

Accuracy Tests

Checklist

• Add at least a tag in the PR title.
  • Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
  • You can add new tags based on the PR content, but the semantics must be clear.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[paddle-bot]

Thanks for your contribution!

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
3 out of 4 committers have signed the CLA.

✅ ckl117
✅ zoooo0820
✅ fxyfxy777
❌ K11OntheBoat

---

K11OntheBoat seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[codecov-commenter]

Codecov Report

❌ Patch coverage is 28.78788% with 94 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@4760835). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
...oy/model_executor/layers/quantization/fp8_utils.py	19.14%	36 Missing and 2 partials ⚠️
...el_executor/layers/moe/fused_moe_triton_backend.py	4.16%	23 Missing ⚠️
..._executor/layers/moe/fused_moe_deepgemm_backend.py	25.00%	19 Missing and 2 partials ⚠️
...del_executor/layers/quantization/block_wise_fp8.py	47.82%	9 Missing and 3 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #6304   +/-   ##
==========================================
  Coverage           ?   67.20%           
==========================================
  Files              ?      386           
  Lines              ?    51320           
  Branches           ?     8003           
==========================================
  Hits               ?    34490           
  Misses             ?    14351           
  Partials           ?     2479           

Flag	Coverage Δ
GPU	67.20% <28.78%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[RichardWooSJTU]

LGTM

[EmmonsCurse]

LGTM for skip coverage~