Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,59 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: llama3
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
base_model:
|
6 |
+
- m3rg-iitd/llamat-3
|
7 |
+
tags:
|
8 |
+
- crystal structure generation
|
9 |
+
- CIF understanding
|
10 |
+
---
|
11 |
+
# Model Card for LLaMat-3-CIF
|
12 |
+
|
13 |
+
**LLaMat-3-CIF** is a specialized large language model designed to generate and extract information from Crystallographic Information Files.
|
14 |
+
|
15 |
+
The model is developed after continued pretraining of **LLaMat-3** on 7M instruction-output pairs obtained using CIFs from Materials Project, Google GNoME, and AMCSD
|
16 |
+
|
17 |
+
---
|
18 |
+
|
19 |
+
## Overview
|
20 |
+
|
21 |
+
- **Model Type:** Large Language Model (LLM)
|
22 |
+
- **Base Model:** LLaMat-3 (continued pretraining of LLaMA-3 on material science data)
|
23 |
+
- **Language:** English
|
24 |
+
- **License:** LLaMA-3 License
|
25 |
+
- **Tags:** Material Science, Domain Adaptation, Crystal Structure Generation
|
26 |
+
|
27 |
+
---
|
28 |
+
|
29 |
+
## Model Details
|
30 |
+
|
31 |
+
### Key Features
|
32 |
+
- **Instruction Following Abilities:** Answers questions based on CIF files.
|
33 |
+
- **Applications:** Crystal structure generation
|
34 |
+
|
35 |
+
### Development and Support
|
36 |
+
- **Developed by:** M3RG, IIT Delhi
|
37 |
+
- **Compute Support:**
|
38 |
+
- **Edinburgh International Data Facility (EIDF):** Provided access to Cerebras CS2 clusters for pretraining.
|
39 |
+
- **IIT Delhi High-Performance Computing Cluster:** Supported fine-tuning and inference stages.
|
40 |
+
|
41 |
+
---
|
42 |
+
|
43 |
+
## Technical Specifications
|
44 |
+
|
45 |
+
### Hardware Infrastructure
|
46 |
+
- **Pretraining:** 2 Cerebras CS-2 Wafer-Scale Engines (WSE-2)
|
47 |
+
- **Finetuning:** 2 Cerebras CS-2 Wafer-Scale Engines (WSE-2)
|
48 |
+
- **Inferencing:** 1 NVIDIA A100 80GB GPU
|
49 |
+
|
50 |
+
### Software Stack
|
51 |
+
- **Frameworks:** PyTorch, Hugging Face Transformers
|
52 |
+
|
53 |
+
---
|
54 |
+
|
55 |
+
## Model Sources
|
56 |
+
- **Repository:** [LLaMat on GitHub](https://github.com/M3RG-IITD/llamat)
|
57 |
+
- **Compute Resources:** [EIDF Cerebras CS Clusters](https://edinburgh-international-data-facility.ed.ac.uk/services/computing/cerebras-cs)
|
58 |
+
|
59 |
+
---
|