[BugFix] Fix model loading error for 300B FP8 EP parallel test case

[Echo-Nie]

Motivation

Fix model loading error in 300B FP8 EP parallel test case: the error occurred when transposing weights (dimension mismatch due to missing weight adaptation before loading).

Modifications

1. Added weight_adapter function to handle weight renaming.
2. Replaced original weights_iterator with adapted_iterator (processed by weight_adapter) in both cache and direct loading paths.
3. Fixed parameter count mismatch in get_padding_offset call.

Usage or Command

model_path=ERNIE-4.5-300B-A47B-FP8-Paddle
config_yaml=yaml/eb45-8k-fp8-tp1-dp8_ep.yaml
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
ep_engine_ports="5378,5379,5380,5381,5382,5383,5384,5385"
metrics_ports="5110,5111,5112,5113,5114,5115,5116,5117"
ports="8188,8189,8190,8191,8192,8193,8194,8195"
cache_ports="9320,9321,9322,9323,9324,9325,9326,9327"
server_num=8
export FD_ENABLE_MULTI_API_SERVER=1
export KVCACHE_RDMA_NICS="mlx5_2,mlx5_3,mlx5_4,mlx5_5,mlx5_6,mlx5_7,mlx5_8,mlx5_9"
python -m fastdeploy.entrypoints.openai.multi_api_server --ports ${ports} --num-servers ${server_num} --metrics-ports ${metrics_ports} --args --model ${model_path} --cache-queue-port ${cache_ports} --engine-worker-queue-port ${ep_engine_ports} --config ${config_yaml} >server.log 2>&1 &

Accuracy Tests

Service Satrt

[INFO] Application startup complete.
[INFO] 127.0.0.1:25712 - "GET /v1/models HTTP/1.1" 200
[INFO] 127.0.0.1:25712 - "POST /v1/chat/completions HTTP/1.1" 200

Test

connecting: http://localhost:8188/v1 ...
model: /workspace/FirstBuildFD/bug2_3/bd
📝 Question: '1+1=?'
🤖 Response: The question "1 + 1 = ?" is a basic arithmetic problem. 
In standard arithmetic, when we add 1 and 1 together, the result is 2. 
So, 1 + 1 = 2.

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.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 75.00000% with 3 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@2ffcb3d). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
...y/model_executor/model_loader/default_loader_v1.py	75.00%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #6382   +/-   ##
==========================================
  Coverage           ?   68.36%           
==========================================
  Files              ?      391           
  Lines              ?    52250           
  Branches           ?     8148           
==========================================
  Hits               ?    35723           
  Misses             ?    13918           
  Partials           ?     2609           

Flag	Coverage Δ
GPU	68.36% <75.00%> (?)

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.