Exported Embedding Model
This model package was exported from the Stage 2 training run:
- Run directory:
20260716_225939_stage2_pplx_stage1_2048 - Export date:
2026-08-28
It is intended to be downloaded directly from Hugging Face and tested immediately.
Model Info
- Model:
HienDuong/pplx-embed-stage1-2048-deepedu - Model link:
https://huggingface.co/HienDuong/pplx-embed-stage1-2048-deepedu - Base HF model:
perplexity-ai/pplx-embed-v1-0.6b - Base link:
https://huggingface.co/perplexity-ai/pplx-embed-v1-0.6b
Prompt Behavior
- Query prefix:
"" - Document prefix:
""
Loading
from sentence_transformers import SentenceTransformer
model = SentenceTransformer(
"HienDuong/pplx-embed-stage1-2048-deepedu",
token="HF_READ_TOKEN_IF_PRIVATE",
trust_remote_code=True,
)
Best Checkpoint Selection
load_best_model_at_end:Truemetric_for_best_model:eval_DeepEdu_dot_ndcg@10best_metric:0.5831874106543483best_model_checkpoint:output/20260716_225939_stage2_pplx_stage1_2048/checkpoints/checkpoint-100
Artifact Layout
- root: loadable Sentence Transformers model
artifacts/stage2_manifest.json: training/export manifestartifacts/best_summary.json: best-checkpoint summaryartifacts/post_eval_summary.json: post-train benchmark summaryartifacts/post_eval/: raw post-eval outputs
Notes
trust_remote_code=Trueis required.- The base model uses custom code.
- This finetuned model keeps empty query/document prefixes, consistent with the base PPLX usage pattern.
- Downstream teams should still test with their actual retrieval scoring setup before production use.