femiari commited on
Commit
5b44560
1 Parent(s): e1bdcd6

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +44 -0
README.md CHANGED
@@ -20,7 +20,51 @@ QwenMoEAriel is a Mixture of Experts (MoE) made with the following models using
20
  * [Replete-AI/Replete-Coder-Qwen2-1.5b](https://huggingface.co/Replete-AI/Replete-Coder-Qwen2-1.5b)
21
 
22
  ## 🧩 Configuration
 
23
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
 
25
  ## 💻 Usage
26
 
 
20
  * [Replete-AI/Replete-Coder-Qwen2-1.5b](https://huggingface.co/Replete-AI/Replete-Coder-Qwen2-1.5b)
21
 
22
  ## 🧩 Configuration
23
+ base_model : Qwen/Qwen2-1.5B
24
 
25
+ architecture: qwen
26
+
27
+ experts:
28
+
29
+ - source_model: Qwen/Qwen2-1.5B
30
+
31
+ positive_prompts:
32
+
33
+ - "chat"
34
+
35
+ - "assistant"
36
+
37
+ - "tell me"
38
+
39
+ - "explain"
40
+
41
+ - "I want"
42
+
43
+ - source_model: Replete-AI/Replete-Coder-Qwen2-1.5b
44
+
45
+ positive_prompts:
46
+
47
+ - "code"
48
+
49
+ - "python"
50
+
51
+ - "javascript"
52
+
53
+ - "programming"
54
+
55
+ - "algorithm"
56
+
57
+ shared_experts:
58
+
59
+ - source_model: Qwen/Qwen2-1.5B
60
+
61
+ positive_prompts: # required by Qwen MoE for "hidden" gate mode, otherwise not allowed
62
+
63
+ - "chat"
64
+
65
+ # (optional, but recommended:)
66
+
67
+ residual_scale: 0.1 # downweight output from shared expert to prevent overcooking the model
68
 
69
  ## 💻 Usage
70