PEFT
code
instruct
llama2
souvik0306 commited on
Commit
e843c58
1 Parent(s): c4a1d81

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +5 -10
README.md CHANGED
@@ -35,22 +35,17 @@ With the utilization of [MonsterAPI](https://monsterapi.ai)'s [LLM finetuner](ht
35
  - **Total Finetuning Cost:** $1.313
36
  - **Model Path:** meta-llama/Llama-2-7b-hf
37
  - **Learning Rate:** 0.0002
38
- - **Data Split:** 99% train 1% validation
39
  - **Gradient Accumulation Steps:** 4
40
  - **lora r:** 32
41
  - **lora alpha:** 64
42
 
43
- ---
44
- Prompt Structure
45
  ```
46
- ### INSTRUCTION:
47
- [instruction]
48
-
49
- ### RESPONSE:
50
- [text]
51
  ```
52
- Train loss :
53
 
54
- ![eval loss](https://cdn-uploads.huggingface.co/production/uploads/63ba46aa0a9866b28cb19a14/D84FFl8hAorzJbtSfiiIT.png)
55
 
56
  license: apache-2.0
 
35
  - **Total Finetuning Cost:** $1.313
36
  - **Model Path:** meta-llama/Llama-2-7b-hf
37
  - **Learning Rate:** 0.0002
38
+ - **Data Split:** 100% train
39
  - **Gradient Accumulation Steps:** 4
40
  - **lora r:** 32
41
  - **lora alpha:** 64
42
 
43
+ #### Prompt Structure
 
44
  ```
45
+ <|system|> <|endoftext|> <|user|> [USER PROMPT]<|endoftext|> <|assistant|> [ASSISTANT ANSWER] <|endoftext|>
 
 
 
 
46
  ```
47
+ #### Train loss :
48
 
49
+ ![eval loss](https://cdn-uploads.huggingface.co/production/uploads/63ba46aa0a9866b28cb19a14/_UwicIoHhj1RrMjt_63vQ.png)
50
 
51
  license: apache-2.0