Mykes commited on
Commit
f9fda8c
·
verified ·
1 Parent(s): 857059d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +42 -2
README.md CHANGED
@@ -1,14 +1,19 @@
1
  ---
2
  base_model: unsloth/gemma-7b-bnb-4bit
3
  language:
4
- - en
5
  license: apache-2.0
 
6
  tags:
7
  - text-generation-inference
8
  - transformers
9
  - unsloth
10
  - gemma
11
- - gguf
 
 
 
 
12
  ---
13
 
14
  # Uploaded model
@@ -20,3 +25,38 @@ tags:
20
  This gemma model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
21
 
22
  [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  base_model: unsloth/gemma-7b-bnb-4bit
3
  language:
4
+ - ru
5
  license: apache-2.0
6
+ pipeline_tag: text-generation
7
  tags:
8
  - text-generation-inference
9
  - transformers
10
  - unsloth
11
  - gemma
12
+ - trl
13
+ - sft
14
+ - chemistry
15
+ - biology
16
+ - medical
17
  ---
18
 
19
  # Uploaded model
 
25
  This gemma model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
26
 
27
  [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
28
+
29
+ ## Supervised Fine-tuning of gemma-7b
30
+ I used unsloth lib, unsloth/gemma-7b-bnb-4bit, rus_med_dialogues dataset
31
+ 10 epochs, 2560 training steps on 1 x RTX 4090
32
+ Training time was about 35 minutes
33
+ The Training Loss in the end was 0.0939
34
+
35
+ ## Parameters:
36
+
37
+ `Unsloth: Fast Gemma patching release 2024.7`
38
+ `GPU: NVIDIA GeForce RTX 4090. Max memory: 23.643 GB. Platform = Linux.`
39
+ `Pytorch: 2.2.0+cu121. CUDA = 8.9. CUDA Toolkit = 12.1.`
40
+ `Bfloat16 = TRUE. FA [Xformers = 0.0.24. FA2 = True]`
41
+
42
+
43
+ ### Example of answer before finetuning:
44
+ **input:**
45
+ *'<bos>Below is the context which is your conversation history and the last user question. Write a response according the context and question.*
46
+ *### Context: user: Недавно мне делали операцию, а рана до сих пор не заживает? assistant: Я понимаю, что это может быть неприятно. Сколько времени прошло с момента операции.*
47
+ *### Question: Месяц. Это сильно плохо?`*
48
+
49
+ **output:**
50
+ `### Response: 1. I'm sorry to hear that you're still experiencing pain from your recent surgery. 2. It's understandable that this can be a difficult situation to be in. 3. I'd like to help you find a solution to this problem. 4. Have you considered trying a different'
51
+
52
+ ### After 10 epochs:
53
+ **input**
54
+ *<bos>Below is the context which is your conversation history and the last user question. Write a response according the context and question.*
55
+ *### Context: user: Недавно мне делали операцию, а рана до сих пор не заживает? assistant: Я понимаю, что это может быть неприятно. Сколько времени прошло с момента операции.*
56
+ *### Question: Месяц. Это сильно плохо?*
57
+
58
+ **output:**
59
+ `### Response: 1 месяц - это не так уж и много. В среднем раны заживают за 2-3 месяца. Но если Вы чувствуете сильную боль или не литров, обязательно обратитесь к своему хирургу для проверки.`
60
+
61
+ # Disclaimer
62
+ The data provided is solely for research purposes. It may include inaccuracies, potentially harmful content, or biases, and we disclaim any responsibility for these issues. Users should assess the data's accuracy and appropriateness before utilizing it. The opinions and views represented in the data do not necessarily align with those of the author.