Commit
•
5af5d78
1
Parent(s):
e4b5496
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: llama2
|
| 3 |
+
datasets:
|
| 4 |
+
- JeanKaddour/minipile
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
Meta's Llama 3 70B pruned to 42B parameters using the methodology described in [The Unreasonable Ineffectiveness of the Deeper Layers](https://arxiv.org/abs/2403.17887). Post-pruning trained using QLoRA for ~100M tokens from [JeanKaddour/minipile](https://huggingface.co/datasets/JeanKaddour/minipile).
|
| 10 |
+
|
| 11 |
+
Layers to prune selected using [PruneMe](https://github.com/arcee-ai/PruneMe).
|
| 12 |
+
|
| 13 |
+
Still evaluating, don't get too excited! Might be incredibly dumb. Check out these zero-shot MMLU numbers though:
|
| 14 |
+
|
| 15 |
+
|
| 16 |
+
| Groups |Version|Filter|n-shot|Metric|Value | |Stderr|
|
| 17 |
+
|------------------|-------|------|-----:|------|-----:|---|-----:|
|
| 18 |
+
|mmlu |N/A |none | 0|acc |0.7319|± |0.0034|
|
| 19 |
+
| - humanities |N/A |none | 0|acc |0.6582|± |0.0063|
|
| 20 |
+
| - other |N/A |none | 0|acc |0.7927|± |0.0069|
|
| 21 |
+
| - social_sciences|N/A |none | 0|acc |0.8466|± |0.0064|
|
| 22 |
+
| - stem |N/A |none | 0|acc |0.6702|± |0.0079|
|