kevin009 commited on
Commit
0f6ad95
·
verified ·
1 Parent(s): 86d8425

Add model card

Browse files
Files changed (1) hide show
  1. README.md +73 -0
README.md ADDED
@@ -0,0 +1,73 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ ---
5
+ # Model Card for BabyLlama v0.6
6
+
7
+ ## Overview
8
+ **Model Name:** BabyLlama v0.6
9
+ **Repository:** kevin009/babyllama-v0.6
10
+ **Architecture:** LlamaForCausalLM, based on TinyLlama 1.1b
11
+ **Model Type:** llama
12
+ **Version:** 0.5
13
+
14
+ ## Model Description
15
+ BabyLlama v0.6 is an innovative conversational AI model, specialized in generating creative and humorous interactions. While it's built on the Llama2 architecture and specifically draws from the TinyLlama 1.1b, this version sets itself apart by not strictly adhering to user instructions. Instead, it aims to replicate human-like conversation in a manner that's distinctly recognizable from actual human dialogue, focusing on creativity and humor.
16
+
17
+ ## Technical Specifications
18
+
19
+ - **Attention Bias:** False
20
+ - **BOS Token ID:** 1
21
+ - **EOS Token ID:** 2
22
+ - **Hidden Activation Function:** SiLU (silu)
23
+ - **Hidden Size:** 2048
24
+ - **Initializer Range:** 0.02
25
+ - **Intermediate Size:** 5632
26
+ - **Max Position Embeddings:** 2048
27
+ - **Number of Attention Heads:** 32
28
+ - **Number of Hidden Layers:** 22
29
+ - **Number of Key/Value Heads:** 4
30
+ - **Pretraining TP:** 1
31
+ - **RMS Norm Epsilon:** 1e-05
32
+ - **ROPE Scaling:** null
33
+ - **ROPE Theta:** 10000.0
34
+ - **Tie Word Embeddings:** False
35
+ - **Torch DType:** float16
36
+ - **Transformers Version:** 4.35.2
37
+ - **Use Cache:** True
38
+ - **Vocabulary Size:** 32000
39
+
40
+
41
+ ## Use Cases
42
+ This model excels in applications where engaging, entertaining, and uniquely human-distinguishable AI responses are valued. It is particularly suited for chatbots, entertainment platforms, interactive games, and social experiments where the focus is on creativity, humor, and the unexpected.
43
+
44
+ ## How to Use
45
+ BabyLlama v0.6 is particularly adept for applications within the realm of role-playing games (RPGs), interactive storytelling, and simulation-based training where dynamic, engaging, and character-driven dialogues are essential. It can serve as an AI companion or NPC (non-player character) that interacts with users in a variety of scenarios, ranging from fantasy and adventure to everyday social simulations. Its unique conversational style enhances the immersive experience by providing responses that are not only contextually relevant but also infused with creativity and humor, thereby elevating the narrative depth and engagement of the role-playing environment.
46
+
47
+
48
+ # Example usage for playful interaction
49
+
50
+
51
+ ```python
52
+ from transformers import AutoModelForCausalLM, AutoTokenizer
53
+
54
+ model_name = "kevin009/babyllama-v0.6"
55
+ tokenizer = AutoTokenizer.from_pretrained(model_name)
56
+ model = AutoModelForCausalLM.from_pretrained(model_name)
57
+
58
+ scenario_prompt = "User: Can you teach me how to plant a magical garden? \n\n Bot:"
59
+
60
+ ```
61
+
62
+ ## Limitations and Considerations
63
+ Due to its design for generating creative and humorous content, BabyLlama v0.5 might not strictly follow provided instructions, reflecting its unique training approach. Users should be mindful of its propensity for unexpected outputs and incorporate suitable moderation or guidance mechanisms as necessary.
64
+ Limitations and Considerations
65
+
66
+ BabyLlama v0.6's focus on playful and fictional dialogues means it may not be suitable for applications requiring factual accuracy or serious outcomes. Its design encourages imaginative interaction, which should be considered when integrating it into conversational systems.
67
+ BabyLlama v0.6 might not strictly follow provided instructions, reflecting its unique training approach, Users should be mindful of its propensity for unexpected outputs and incorporate suitable moderation or guidance mechanisms as necessary.
68
+
69
+ ## Acknowledgments
70
+ The development of BabyLlama v0.5 reflects a significant effort to push the boundaries of conversational AI, drawing from the foundational TinyLlama 1.1b model and incorporating advanced fine-tuning techniques to achieve its distinctive capabilities.
71
+
72
+ ## Version History
73
+ - **v0.5:** Enhanced for creativity and humor in conversations, diverging from strict instruction adherence to offer a unique conversational experience.