cs4248-nlp/paper-s1-control-bs32-tinybert-general-4l-312d-taco-hf-20260410-234932

Code-search embedding model trained with the CS4248 two-phase KD pipeline.

Model details

Usage

from transformers import AutoModel, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("cs4248-nlp/paper-s1-control-bs32-tinybert-general-4l-312d-taco-hf-20260410-234932")
model = AutoModel.from_pretrained("cs4248-nlp/paper-s1-control-bs32-tinybert-general-4l-312d-taco-hf-20260410-234932")

Mean-pool the last hidden state to get a fixed-size embedding:

import torch

def mean_pool(model_output, attention_mask):
    token_embeddings = model_output.last_hidden_state
    mask = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return (token_embeddings * mask).sum(1) / mask.sum(1).clamp(min=1e-9)

inputs = tokenizer("your query here", return_tensors="pt", truncation=True, max_length=160)
with torch.no_grad():
    outputs = model(**inputs)
embedding = mean_pool(outputs, inputs['attention_mask'])