Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
datasets:
|
3 |
+
- agentlans/crash-course
|
4 |
+
base_model:
|
5 |
+
- FuseAI/FuseChat-Gemma-2-9B-Instruct
|
6 |
+
- jsgreenawalt/gemma-2-9B-it-advanced-v2.1
|
7 |
+
---
|
8 |
+
# Gemma2-9B-AdvancedFuse
|
9 |
+
|
10 |
+
Gemma2-9B-AdvancedFuse is an experimental, open-source large language model (LLM) with 9 billion parameters.
|
11 |
+
It aims to combine the strengths of [FuseAI/FuseChat-Gemma-2-9B-Instruct](https://huggingface.co/fuseai/fusechat-gemma-2-9b-instruct) and
|
12 |
+
[jsgreenawalt/gemma-2-9B-it-advanced-v2.1](https://huggingface.co/jsgreenawalt/gemma-2-9b-it-advanced-v2.1) through additive linear merging,
|
13 |
+
further fine-tuned on a 12K row dataset from [agentlans/crash-course](https://huggingface.co/datasets/agentlans/crash-course)
|
14 |
+
for enhanced chat and instruct performance, including math and multilingual prompts.
|
15 |
+
|
16 |
+
## Capabilities
|
17 |
+
- **Text Generation:** Generates coherent emails, summaries, and notes. This model card was primarily generated by the model itself.
|
18 |
+
- **Instruction Following:** Demonstrates strong ability to understand and execute instructions in conversational settings.
|
19 |
+
- **Roleplaying:** Can engage in third-person narrative roleplay but may exhibit common GPT expressions or clichés.
|
20 |
+
|
21 |
+
### Limitations
|
22 |
+
As with most large language models:
|
23 |
+
- **Factual Errors:** May generate incorrect or outdated information due to data biases.
|
24 |
+
- **Mathematical Operations:** Struggles with mathematical calculations requiring symbolic reasoning despite its finetuning data.
|
25 |
+
- **Handling Unsafe Input:** May generate unsafe, biased, or malicious content if provided inappropriate input. Careful prompt engineering is recommended.
|
26 |
+
|
27 |
+
### Model Usage Guidelines
|
28 |
+
1. Use clear and specific instructions for optimal performance.
|
29 |
+
2. Verify generated outputs for factual accuracy when critical information is involved.
|
30 |
+
3. Avoid providing inputs that could lead to harmful or unethical responses.
|
31 |
+
4. Consider using human review, especially in high-stakes applications.
|