Spestly commited on
Commit
191fd0e
·
verified ·
1 Parent(s): a7dd6fb

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +104 -1
README.md CHANGED
@@ -9,5 +9,108 @@ tags:
9
  license: mit
10
  language:
11
  - en
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
12
  ---
13
- ![Header](./Atlas-Pro.png)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9
  license: mit
10
  language:
11
  - en
12
+ - zh
13
+ - fr
14
+ - es
15
+ - pt
16
+ - de
17
+ - it
18
+ - ru
19
+ - ja
20
+ - ko
21
+ - vi
22
+ - th
23
+ - ar
24
+ - fa
25
+ - he
26
+ - tr
27
+ - cs
28
+ - pl
29
+ - hi
30
+ - bn
31
+ - ur
32
+ - id
33
+ - ms
34
+ - lo
35
+ - my
36
+ - ceb
37
+ - km
38
+ - tl
39
+ - nl
40
+ datasets:
41
+ - openai/gsm8k
42
+ - HuggingFaceH4/ultrachat_200k
43
+ library_name: transformers
44
  ---
45
+ ![Header](./Atlas-Pro.png)
46
+ # **Atlas Pro**
47
+
48
+ ### **Model Overview**
49
+ **Atlas Pro** (Previously known as '🏆 Atlas-Experiment 0403 🧪' in AtlasUI) is an advanced language model (LLM) built on top of **Atlas Flash**. It's designed to provide exceptional performance for professional tasks like coding, mathematics, and scientific problem-solving. Atlas Pro builds on Atlas Flash by adding more fine-tuning and specialization, making it perfect for researchers and advanced users.
50
+
51
+ ---
52
+
53
+ ### **Key Features**
54
+ - **Improved Problem-Solving:** Handles tricky tasks in programming, math, and sciences better than most models.
55
+ - **Advanced Code Generation:** Produces clean and efficient code, but may still miss edge cases occasionally.
56
+ - **Domain Expertise:** Focused on technical and scientific domains but works well in general contexts too.
57
+ - **Reasoning Improvement:** In this version of Atlas, I have enhanced it's reasoning via synthetic data from models such as Gemini-2.0 Flash Thinking so that it can improve on reasoning.
58
+ ---
59
+
60
+ ### **Intended Use Cases**
61
+ Atlas Pro works best for:
62
+ - **Technical Professionals:** Helping developers, engineers, and scientists solve complex problems.
63
+ - **Educational Assistance:** Offering clear, step-by-step help for students and teachers.
64
+ - **Research Support:** Assisting in theoretical and applied science work.
65
+ - **Enterprise Tools:** Integrating into company workflows for smarter systems.
66
+
67
+ ---
68
+
69
+ ### **NOTICE**
70
+ Atlas Pro is built on **Atlas Flash** and improved to meet high standards. Here’s how it’s made:
71
+
72
+ 1. **Base Model:** Built upon **Atlas Flash**, which is already quite capable.
73
+ 2. **Fine-Tuning Details:**
74
+ - Used datasets specific to programming, math, and scientific challenges and overall reasoning abilities.
75
+ - Refined its performance for professional scenarios.
76
+ 3. **Performance Highlights:**
77
+ - Beats benchmarks with high accuracy, though occasional tweaks might still improve outputs.
78
+ ---
79
+
80
+ ### **Limitations**
81
+ - **Knowledge Cutoff:** It doesn’t know about anything recent unless updated.
82
+ - **Hardware Requirements:** Needs high-end GPUs to run smoothly.
83
+ - **Specialization Bias:** While amazing in its focus areas, general chat capabilities might not be as good as other models.
84
+ - **Token Leakage:** In some very rare cases (~1/167), Atlas Pro will experience some token leakage.
85
+
86
+ ---
87
+
88
+ ### **Licensing**
89
+ Atlas Pro is released under the **MIT**, which prohibits harmful uses. Make sure to follow the rules in the license agreement.
90
+
91
+ ---
92
+
93
+ ### **Acknowledgments**
94
+ Created by **Spestly** as part of the **Astral Model Family**, Atlas Pro builds on the strong foundation of **Atlas Flash**. Special thanks to **Deepseek's R1 Qwen Distilles** for helping make it happen.
95
+
96
+ ---
97
+
98
+ ### **Usage**
99
+ You can use Atlas Pro with this code snippet:
100
+
101
+ ```python
102
+ from transformers import AutoModelForCausalLM, AutoTokenizer
103
+
104
+ # Load the Atlas Pro model
105
+ model_name = "Spestly/Atlas-R1-Pro-1.5B-Preview"
106
+ tokenizer = AutoTokenizer.from_pretrained(model_name)
107
+ model = AutoModelForCausalLM.from_pretrained(model_name)
108
+
109
+ # Generate a response
110
+ prompt = "Write a Python function to calculate the Fibonacci sequence."
111
+ inputs = tokenizer(prompt, return_tensors="pt")
112
+ outputs = model.generate(**inputs, max_length=200)
113
+ response = tokenizer.decode(outputs[0], skip_special_tokens=True)
114
+
115
+ print(response)
116
+ ```