alextripplet commited on
Commit
220c7b1
·
verified ·
1 Parent(s): 8bf14e4

Add Majuli 3.1 model card

Browse files
Files changed (1) hide show
  1. README.md +70 -0
README.md ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ language:
4
+ - en
5
+ - ru
6
+ tags:
7
+ - transformers
8
+ - safetensors
9
+ - gemma3
10
+ - image-text-to-text
11
+ - creative
12
+ - roleplay
13
+ - conversational
14
+ - text-generation-inference
15
+ - endpoints_compatible
16
+ base_model:
17
+ - OddTheGreat/Mars_27B_V.1
18
+ model_name: Majuli 3.1
19
+ pipeline_tag: image-text-to-text
20
+ ---
21
+
22
+ # Majuli 3.1
23
+
24
+ **By Tripplet AI** (Tripplet Artificial General Intelligence Research Institute)
25
+
26
+ Majuli 3.1 is a powerful 27B parameter multimodal language model built on the Gemma 3 architecture, optimized for creative writing, roleplay, and general-purpose instruction following.
27
+
28
+ ## Model Details
29
+
30
+ - **Parameters:** 28.8B
31
+ - **Architecture:** Gemma 3 (Gemma3ForConditionalGeneration)
32
+ - **Context Length:** 131,072 tokens
33
+ - **Hidden Size:** 5376
34
+ - **Layers:** 62
35
+ - **Attention Heads:** 32 (16 KV heads)
36
+ - **Vision Encoder:** SigLIP (896px, 27 layers)
37
+ - **Languages:** English, Russian
38
+ - **Precision:** bfloat16
39
+
40
+ ## Key Features
41
+
42
+ - Long context support up to 128K tokens
43
+ - Multimodal capabilities (image + text)
44
+ - Hybrid attention with sliding window (1024) and full attention layers
45
+ - Optimized for creative and roleplay tasks
46
+
47
+ ## Usage
48
+
49
+ ```python
50
+ from transformers import AutoProcessor, AutoModelForImageTextToText
51
+
52
+ model = AutoModelForImageTextToText.from_pretrained("tripplet-research/majuli-3.1")
53
+ processor = AutoProcessor.from_pretrained("tripplet-research/majuli-3.1")
54
+
55
+ messages = [
56
+ {"role": "user", "content": "Hello, tell me about yourself."}
57
+ ]
58
+
59
+ inputs = processor.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True)
60
+ output = model.generate(**inputs, max_new_tokens=512)
61
+ print(processor.decode(output[0], skip_special_tokens=True))
62
+ ```
63
+
64
+ ## About Tripplet AI
65
+
66
+ Tripplet Artificial General Intelligence Research Institute is dedicated to advancing the frontiers of artificial general intelligence through open research and model development.
67
+
68
+ ## License
69
+
70
+ Apache 2.0