Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
tags:
|
| 6 |
+
- evaluation
|
| 7 |
+
- mmlu
|
| 8 |
+
- exaone
|
| 9 |
+
- amd-mi325
|
| 10 |
+
pretty_name: EXAONE 1.2B Quantized MMLU Results
|
| 11 |
+
size_categories:
|
| 12 |
+
- n<1K
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# EXAONE-4.0-1.2B-GPTQ MMLU Evaluation
|
| 16 |
+
|
| 17 |
+
This repository contains refined evaluation results for the **EXAONE-4.0-1.2B-GPTQ-W8A16** model.
|
| 18 |
+
|
| 19 |
+
## 🚀 Evaluation Setup
|
| 20 |
+
- **Benchmark:** MMLU (Massive Multitask Language Understanding)
|
| 21 |
+
- **Prompting:** 5-shot
|
| 22 |
+
- **Hardware:** AMD Instinct MI325 OAM
|
| 23 |
+
- **Quantization:** GPTQ (W8A16 / fp8_w8a8)
|
| 24 |
+
|
| 25 |
+
## 📊 Quick Results
|
| 26 |
+
The full breakdown of 57 subjects can be found in `mmlu_refined.csv`.
|
| 27 |
+
|
| 28 |
+
| Metric | Value |
|
| 29 |
+
| :--- | :--- |
|
| 30 |
+
| **Tool** | lm-evaluation-harness |
|
| 31 |
+
| **Accelerator** | AMD MI325 |
|
| 32 |
+
| **Status** | Completed |
|
| 33 |
+
|
| 34 |
+
## 📂 File Structure
|
| 35 |
+
- `mmlu_refined.csv`: Key performance metrics per subject (Accuracy %).
|
| 36 |
+
- `raw_data/`: Original JSON output from the evaluation harness.
|