Kquant03 commited on
Commit
e9e9d8a
1 Parent(s): fbb08ac

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +75 -0
README.md ADDED
@@ -0,0 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - merge
4
+ - mergekit
5
+ - lazymergekit
6
+ - mlabonne/OmniBeagle-7B
7
+ - flemmingmiguel/MBX-7B-v3
8
+ - AiMavenAi/AiMaven-Prometheus
9
+ base_model:
10
+ - mlabonne/OmniBeagle-7B
11
+ - flemmingmiguel/MBX-7B-v3
12
+ - AiMavenAi/AiMaven-Prometheus
13
+ license: apache-2.0
14
+ ---
15
+
16
+
17
+ # NeuralTrix-7B-v1
18
+
19
+ NeuralTrix-7B-v1 is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
20
+ * [mlabonne/OmniBeagle-7B](https://huggingface.co/mlabonne/OmniBeagle-7B)
21
+ * [flemmingmiguel/MBX-7B-v3](https://huggingface.co/flemmingmiguel/MBX-7B-v3)
22
+ * [AiMavenAi/AiMaven-Prometheus](https://huggingface.co/AiMavenAi/AiMaven-Prometheus)
23
+
24
+ It was then trained with DPO using:
25
+ * https://huggingface.co/datasets/jondurbin/truthy-dpo-v0.1
26
+
27
+ ## 🧩 Configuration
28
+
29
+ ```yaml
30
+ models:
31
+ - model: mistralai/Mistral-7B-v0.1
32
+ # no parameters necessary for base model
33
+ - model: mlabonne/OmniBeagle-7B
34
+ parameters:
35
+ density: 0.65
36
+ weight: 0.4
37
+ - model: flemmingmiguel/MBX-7B-v3
38
+ parameters:
39
+ density: 0.6
40
+ weight: 0.35
41
+ - model: AiMavenAi/AiMaven-Prometheus
42
+ parameters:
43
+ density: 0.6
44
+ weight: 0.35
45
+ merge_method: dare_ties
46
+ base_model: mistralai/Mistral-7B-v0.1
47
+ parameters:
48
+ int8_mask: true
49
+ dtype: float16
50
+ ```
51
+
52
+ ## 💻 Usage
53
+
54
+ ```python
55
+ !pip install -qU transformers accelerate
56
+
57
+ from transformers import AutoTokenizer
58
+ import transformers
59
+ import torch
60
+
61
+ model = "CultriX/NeuralTrix-7B-v1"
62
+ messages = [{"role": "user", "content": "What is a large language model?"}]
63
+
64
+ tokenizer = AutoTokenizer.from_pretrained(model)
65
+ prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
66
+ pipeline = transformers.pipeline(
67
+ "text-generation",
68
+ model=model,
69
+ torch_dtype=torch.float16,
70
+ device_map="auto",
71
+ )
72
+
73
+ outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
74
+ print(outputs[0]["generated_text"])
75
+ ```