Downloads · 30 days
0
hw2-distilbert-lighting
srivathsanb14/hw2-distilbert-lighting
hw2-distilbert-lighting is a machine learning model from srivathsanb14. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
'''--- language: en license: mit libraryname: autogluon pipelinetag: text-classification basemodel: distilbert-base-uncased tags: - autogluon - distilbert - education - 24-679 datasets: - {DATASETID} ---
Access
Public
Updated Sep 21, 2026
Repo size
266 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.ckpt265 MB · 100%
At a glance
- Access
- Public
- Created
- Sep 21, 2026
- Updated
- Sep 21, 2026
- SHA
- 61b33b1a
Model card
From the Hugging Face model README
'''--- language: en license: mit library_name: autogluon pipeline_tag: text-classification base_model: distilbert-base-uncased tags:
- autogluon
- distilbert
- education
- 24-679 datasets:
- {DATASET_ID}
HW2: DistilBERT lighting classifier (AutoMM)
Same AutoMM DistilBERT recipe as discussed in class, applied to a classmate lighting dataset. (https://huggingface.co/datasets/jackstev/hw1-text-lighting-natural-artificial)
Purpose
Classify a short room description as natural or artificial lighting.
Task
| id | label |
|---|---|
| 0 | natural |
| 1 | artificial |
Data and splits
Source: {DATASET_ID} (jackstev). Published splits used as-is: train {len(df_train)} / val {len(df_val)} / test {len(df_test)}. Holdouts are unaugmented; no parent of a holdout text appears in train.
Training setup
| Setting | Value |
|---|---|
| Encoder | distilbert-base-uncased, CLS pooling |
| Adaptation | {FINETUNE_MODE} + trainable classification head |
| Trainable parameters | {trainable_parameters:,} / {total_parameters:,} ({trainable_percent:.3f}%) |
| LR | 2e-5, layerwise decay |
| Epochs | 5; best val accuracy checkpoint |
| Batch | 8 |
| Max length | 128 subword tokens |
| Seed | {SEED} |
| Extra text aug | off (text_trivial_aug_maxscale=0) |
Metrics
| Split | Accuracy | Weighted F1 |
|---|---|---|
| test | {accuracy:.4f} | {f1:.4f} |
See the notebook for the confusion matrix and written error analysis.
Limitations and ethics
Do not use for occupancy or energy billing.
License
MIT for this repo. Dataset CC BY 4.0 (jackstev). DistilBERT Apache-2.0.
Hardware / compute
Google Colab GPU, 5 epochs, seed {SEED}. Python {platform.python_version()}.
Load
from pathlib import Path
import pandas as pd
from huggingface_hub import snapshot_download
from autogluon.multimodal import MultiModalPredictor
snapshot = Path(snapshot_download("{HF_MODEL_ID}", allow_patterns=["automm/*"]))
model = MultiModalPredictor.load(str(snapshot / "automm"))
model.set_num_gpus(0)
model.predict(pd.DataFrame({{"text": ["Sunlight slides across the oak floor."]}}))
AI usage disclosure
I chose the classmate dataset and wrote the miss hypotheses. Cursor aligned this notebook with the class DistilBERT AutoMM notebook. '''
- Created
- Sep 21, 2026
- Updated
- Sep 21, 2026