Downloads · 30 days
6
19% of all-time downloads
akshayballal/grug-ettin-68m-meetingbank
grug-ettin-68m-meetingbank is a token classification model from akshayballal. Use it when you need labels on individual words, such as names. It is set up for transformers. The card lists the license as cc-by-nc-sa-4.0.
Binary preserve/discard token classifier for prompt compression, trained with grug train on microsoft/MeetingBank-LLMCompressed.
Downloads · 30 days
6
19% of all-time downloads
All-time downloads
31
Public
Parameters
68.4M
274 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors274 MB · 99%
From the Hugging Face model README
Binary preserve/discard token classifier for prompt compression, trained with
grug train on microsoft/MeetingBank-LLMCompressed.
import grug
from grug.backends.modern import ModernBackend
comp = grug.Compressor(ModernBackend(model_name='akshayballal/grug-ettin-68m-meetingbank-v2'))
result = comp.compress(text, rate=0.4)
{
"model_name": "jhu-clsp/ettin-encoder-68m",
"epochs": 6,
"learning_rate": 1e-05,
"batch_size": 32,
"max_length": 512,
"weight_decay": 0.01,
"warmup_ratio": 0.06,
"device": "cuda",
"seed": 0,
"push_to": "akshayballal/grug-ettin-68m-meetingbank-v2",
"select_on": "val_f1",
"patience": 2,
"cs_weight": 0.0
}
{
"epoch": 6,
"train_loss": 0.32849070216068116,
"val_loss": 0.3391664388044825,
"val_accuracy": 0.848027204080612,
"val_precision": 0.8069981607731251,
"val_recall": 0.7997938874754437,
"val_f1": 0.8033798734488425,
"val_perplexity": 1.4037769687306128
}
The training corpus is CC-BY-NC-SA-4.0, so this model inherits a non-commercial constraint.