【PaddleNLP No.20】Create ernie-3.0 deploy with paddle.inference & add training without hf

[hanlintang]

PR types

Function optimization

PR changes

Models

Description

1. 将目录中的slm/model_zoo/ernie-3.0/infer.py加入文档，并进行PIR适配；
2. 为NER训练提供网络无法连接到HF的替代脚本slm/model_zoo/ernie-3.0/run_token_cls_without_hf.py;
3. 修改文档slm/model_zoo/ernie-3.0/README.md;
4. 针对FastDeploy与paddle 3.0.0不适配的情况，重新实现基于paddle.inference的推理脚本，包括文本分类与NER两个任务；

• 文本分类：slm/model_zoo/ernie-3.0/deploy/python/seq_cls_infer.py
• NER：slm/model_zoo/ernie-3.0/deploy/python/token_cls_infer.py

5. 针对推理示例变化更新部署文档slm/model_zoo/ernie-3.0/deploy/python/README.md。

Issue: #9763
@DrownFish19

[paddle-bot]

Thanks for your contribution!

[Copilot]

Pull Request Overview

This PR implements a Paddle Inference–based deploy for the Ernie‑3.0 model and introduces an alternative training script for NER when Hugging Face connectivity is unavailable.

• Adds a new script (run_token_cls_without_hf.py) for NER training without HF.
• Updates inference scripts (infer.py, token_cls_infer.py, seq_cls_infer.py) to leverage paddle.inference with new model file suffixes.
• Revises documentation in README files to reflect deployment changes.

Reviewed Changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
slm/model_zoo/ernie-3.0/run_token_cls_without_hf.py	Adds an alternative training script for NER.
slm/model_zoo/ernie-3.0/infer.py	Updates inference file paths to use new suffixes.
slm/model_zoo/ernie-3.0/deploy/python/token_cls_infer.py	Replaces fastdeploy with paddle.inference and adjusts preprocessing/postprocessing.
slm/model_zoo/ernie-3.0/deploy/python/seq_cls_infer.py	Updates prediction logic and removes fastdeploy dependencies.
slm/model_zoo/ernie-3.0/deploy/python/README.md	Revises usage examples to reflect updated inference arguments.
slm/model_zoo/ernie-3.0/README.md	Adds deploy instructions for infer.py and run_token_cls_without_hf.py.

Comments suppressed due to low confidence (2)

slm/model_zoo/ernie-3.0/infer.py:190

• Ensure that args.model_path includes a trailing path separator or consider using os.path.join to construct the complete model file path, so that the concatenated file name is formed correctly.

model_file=args.model_path + f"model{PADDLE_INFERENCE_MODEL_SUFFIX}"

slm/model_zoo/ernie-3.0/deploy/python/token_cls_infer.py:100

• Verify that converting infer_data to a numpy array returns the expected shape; if the inference API now wraps the results differently compared to previous versions, you may need to adjust the postprocessing accordingly.

result = np.array(infer_data)

[DrownFish19]

LGTM

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 48.92%. Comparing base (ce7b4cc) to head (b96de4b).
Report is 45 commits behind head on develop.

❌ Your project status has failed because the head coverage (48.92%) is below the target coverage (58.00%). You can increase the head coverage or adjust the target coverage.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop   #10475      +/-   ##
===========================================
- Coverage    48.99%   48.92%   -0.07%     
===========================================
  Files          765      767       +2     
  Lines       125974   126238     +264     
===========================================
+ Hits         61720    61766      +46     
- Misses       64254    64472     +218     

☔ 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.