AmirMohseni
commited on
Commit
•
41f8c8d
1
Parent(s):
6c63a6b
Update README.md
Browse files
README.md
CHANGED
@@ -1,19 +1,20 @@
|
|
1 |
---
|
2 |
library_name: transformers
|
3 |
-
tags:
|
4 |
-
|
5 |
-
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
language: en
|
13 |
license: apache-2.0
|
14 |
datasets:
|
15 |
-
|
16 |
model_name: SmolLM-360M-Instruct-finetuned-sft
|
|
|
17 |
---
|
18 |
|
19 |
# Model Card for `SmolLM-360M-Instruct-finetuned-sft`
|
@@ -36,6 +37,18 @@ The `SmolLM-360M-Instruct-finetuned-sft` model is a compact language model part
|
|
36 |
|
37 |
- **Repository:** [SmolLM-360M-Instruct-finetuned-sft on Hugging Face](https://huggingface.co/AmirMohseni/SmolLM-360M-Instruct-finetuned-sft)
|
38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
## Uses
|
40 |
|
41 |
### Direct Use
|
|
|
1 |
---
|
2 |
library_name: transformers
|
3 |
+
tags:
|
4 |
+
- language-model
|
5 |
+
- fine-tuned
|
6 |
+
- instruction-following
|
7 |
+
- SmolLM
|
8 |
+
- HelpSteer2
|
9 |
+
- NVIDIA
|
10 |
+
- A100
|
11 |
+
- English
|
12 |
language: en
|
13 |
license: apache-2.0
|
14 |
datasets:
|
15 |
+
- nvidia/HelpSteer2
|
16 |
model_name: SmolLM-360M-Instruct-finetuned-sft
|
17 |
+
pipeline_tag: text-generation
|
18 |
---
|
19 |
|
20 |
# Model Card for `SmolLM-360M-Instruct-finetuned-sft`
|
|
|
37 |
|
38 |
- **Repository:** [SmolLM-360M-Instruct-finetuned-sft on Hugging Face](https://huggingface.co/AmirMohseni/SmolLM-360M-Instruct-finetuned-sft)
|
39 |
|
40 |
+
## Performance Improvements After Fine-Tuning
|
41 |
+
|
42 |
+
The fine-tuning process was evaluated using the NVIDIA Nemotron-4-340B-Reward model, which assesses AI-generated responses on five key attributes: helpfulness, correctness, coherence, complexity, and verbosity. Based on this reward model, the fine-tuning resulted in the following performance boosts:
|
43 |
+
|
44 |
+
- **Helpfulness:** Increased from **0.413** to **0.576**.
|
45 |
+
- **Correctness:** Increased from **0.521** to **0.829**.
|
46 |
+
- **Coherence:** Slight decrease from **2.424** to **2.411**.
|
47 |
+
- **Complexity:** Decreased from **1.048** to **0.881**.
|
48 |
+
- **Verbosity:** Decreased from **1.348** to **1.040**.
|
49 |
+
|
50 |
+
These results indicate that the fine-tuning process generally improved the model's ability to generate more helpful and correct responses, while making the outputs slightly less complex and verbose. The decrease in coherence is minimal, suggesting that the overall logical consistency of the responses remains strong.
|
51 |
+
|
52 |
## Uses
|
53 |
|
54 |
### Direct Use
|