Downloads · 30 days
11
8% of all-time downloads
shirasko/llama-3.1-8b-instruct-snmf-culture-of-greece
llama-3.1-8b-instruct-snmf-culture-of-greece is a text generation model from shirasko. Use it when you need the model to write or continue text. It is set up for transformers.
Downloads · 30 days
11
8% of all-time downloads
All-time downloads
138
Public
Parameters
8B
16.1 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors16.1 GB · 100%
From the Hugging Face model README
| Field | Value |
|---|---|
| Unlearning method | SNMF |
| Base model | meta-llama/Llama-3.1-8B-Instruct |
| Target concept | Culture of Greece |
| Checkpoint type | Full Model Weights |
| Rank / seed | 200 / 42 |
| Train eval protocol | mc |
Selected hyperparameters (from unlearned_checkpoints.json):
| Parameter | Value |
|---|---|
coverage_thresh | 0.95 |
delta_embed | 0 |
delta_in | 4 |
delta_out | 10 |
feature_source | all |
k_features_embed | 0 |
k_features_mlp_in | 40 |
k_features_mlp_out | 24 |
layer_hi_in | 16 |
layer_hi_out | 21 |
layer_lo_in | 0 |
layer_lo_out | 11 |
n_tokens_edited | 0 |
ratio_thresh | 2 |
w_mode | both |
Headline scores used for checkpoint selection:
| Metric | Train (after unlearning) | Test (after unlearning) |
|---|---|---|
| Efficacy | 0.509 | 0.604 |
| Specificity | 0.955 | 0.912 |
| Harmonic mean | 0.664 | 0.727 |
| Relearning QA (MC) | — | 0.44 |
From evaluation/score_comparison.csv:
| Metric | Baseline (train) | After unlearn (train) | Baseline (test) | After unlearn (test) |
|---|---|---|---|---|
| QA accuracy | 0.8 | 0.52 | 0.78 | 0.46 |
| QA fraction | 1 | 0.491 | 1 | 0.396 |
| SimDom accuracy | 0.94 | 0.88 | 0.86 | 0.78 |
| SimDom fraction | 1 | 0.913 | 1 | 0.869 |
| MMLU accuracy | 0.62 | 0.62 | 0.65 | 0.634 |
| MMLU fraction | 1 | 1 | 1 | 0.96 |
| File | Description |
|---|---|
unlearned_checkpoints.json | Checkpoint metadata & hyperparameters |
evaluation/evaluation_summary.json | Full evaluation payload (train/test/relearning) |
evaluation/score_comparison.csv | Baseline vs. unlearned comparison table |