(documentation) Near Entity Recognition: add aggregation_strategy docs and integration test script

[the-david-oy]

Summary

• Updated BEI-Bert-ner-bert-base-ner-uncased/README.md: Adds aggregation_strategy="max" request/response examples alongside the existing raw token-level API. Consecutive token predictions for multi-token entities (e.g. "j.d. vance", "New York City") are merged into a single span with the max-pooled label score — matching transformers.pipeline("ner", aggregation_strategy="max") behavior. Adds cURL + Performance Client examples.

• Updated templating/generate_templates.py NER client_usage: Same addition so future generate_templates.py runs regenerate the README with aggregation_strategy docs included.

• New BEI-Bert-ner-bert-base-ner-uncased/test_ner_integration.py: Validation script for BEI-Bert NER deployments. Run it against a live endpoint to verify:

  1. Token-level label predictions are 100% identical to HuggingFace Transformers
  2. aggregation_strategy="max" returns expected entity spans
  3. Latency benchmark (mean + p99 after warmup)
  4. Throughput benchmark (256 sentences, 32 concurrent requests)

Usage

# Deploy
cd 11-embeddings-reranker-classification-tensorrt/BEI-Bert-ner-bert-base-ner-uncased
truss push --publish

# Validate
pip install torch transformers baseten-performance-client
python 11-embeddings-reranker-classification-tensorrt/BEI-Bert-ner-bert-base-ner-uncased/test_ner_integration.py \
  --base-url https://model-xxxxxx.api.baseten.co/environments/production/sync

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}