File size: 1,558 Bytes
3376fe0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
72e02f6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3376fe0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
---
license: apache-2.0
datasets:
- Allanatrix/Scientific_Research_Tokenized
language:
- en
base_model:
- allenai/OLMo-7B
pipeline_tag: text-generation
library_name: peft
tags:
- Olmo
- lora
- peft
- transformers
- scientific-ml
- fine-tuned
- research-assistant
- hypothesis-generation
- scientific-writing
- scientific-reasoning
---

# Model Card for nexa-OLMo-sci7b

## Model Details
**Model Description:**  
nexa-OLMo-sci7b is a fine-tuned variant of allenai/OLMo-7B, optimized for scientific research generation tasks such as hypothesis generation, abstract writing, and methodology completion. Fine-tuning was performed using PEFT with LoRA in 4-bit quantized mode via bitsandbytes.

**Developed by:** Allan (Independent Scientific Intelligence Architect)  
**Shared by:** Allan (https://huggingface.co/allan-wandia)  
**Model type:** Decoder-only transformer (causal language model)  
**Language(s):** English (scientific domain-specific vocabulary)  
**License:** Apache 2.0  
**Fine-tuned from:** allenai/OLMo-7B  
**Repository:** https://huggingface.co/allan-wandia/nexa-olmo-sci7b  

## Training Details
**Training Data:**  
- Size: 100 million tokens  
- Source: Curated scientific literature (Bio, Physics, QST, Astro)  

**Hyperparameters:**  
- Sequence length: 1024  
- Batch size: 1  
- Gradient Accumulation Steps: 64  
- Effective Batch Size: 64  
- Learning rate: 2e-05  
- Epochs: 2  
- LoRA: Enabled (PEFT)  
- Quantization: 4-bit  

**Results:**  
Robust performance in scientific prose tasks, with novelty varying by prompt diversity.