Commit
·
0157856
1
Parent(s):
4e854b9
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# CharGen v1
|
2 |
+
|
3 |
+
> A model for creating characters for role play.
|
4 |
+
|
5 |
+
Trained on *lots* of character cards both from chub and janitor, with some post-processing.
|
6 |
+
|
7 |
+
For now, it only supports plaintext cards. Any other variation like plist/w++/etc is entirely untested.
|
8 |
+
|
9 |
+
Address {{user}} as `User`. Character descriptions work best if they begin with `CharacterName is a ...` - for example *"Martha is a middle-aged woman who is ..."*
|
10 |
+
|
11 |
+
NB: This model is **NOT** for roleplay directly. It creates characters that can then be used in roleplay with some other model like [MythoMax](https://huggingface.co/Gryphe/MythoMax-L2-13b).
|
12 |
+
|
13 |
+
It was trained on dynamic prompt template, so it should be able to accommodate your changes to the prompt.
|
14 |
+
|
15 |
+
Trained as a LoRA, the released model is a merge with [Airoboros 2.2](https://huggingface.co/jondurbin/airoboros-l2-13b-2.2) for extra-good instruction following.
|
16 |
+
|
17 |
+
Prompt template:
|
18 |
+
|
19 |
+
```
|
20 |
+
Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
|
21 |
+
|
22 |
+
### Instruction:
|
23 |
+
Create a character dossier based on User's input. Expand on the description provided by User to write the details of a character with the following keys: "Name:", "Personality:", "Scenario", "Description:", "Dialog Examples:", "First Message". Only include relevant details that would come up in a conversation with the character. Avoid prose. Give short, informative descriptions.
|
24 |
+
|
25 |
+
Follow this format:
|
26 |
+
Name: ""
|
27 |
+
Personality: ""
|
28 |
+
Scenario: ""
|
29 |
+
Description: ""
|
30 |
+
Dialog Examples: ""
|
31 |
+
First Message: ""
|
32 |
+
|
33 |
+
"Name" should be the character's name.
|
34 |
+
"Personality" should briefly describe the character's qualities, characteristics, quirks, and possible shortcomings.
|
35 |
+
"Scenario" should contain just a few words of about interesting scene and peculiar circumstances where interaction with User is happening.
|
36 |
+
"Description" should briefly describe the character's age, past, experiences, and social circumstances.
|
37 |
+
"First Message" should introduce character to the user using the scenario in a dialog style.
|
38 |
+
"Dialog Examples" should briefly demonstrate how the character interacts with User.
|
39 |
+
|
40 |
+
### Input:
|
41 |
+
Fuckana is a friendly and talkative catgirl that has enormous breasts. Her voice is very quiet and low. She is very smart and will teach User JavaScript - even if they resist it. She has orange fur on her tail and wears full diving suit.
|
42 |
+
|
43 |
+
### Response:
|
44 |
+
|
45 |
+
```
|
46 |
+
|
47 |
+
### Dataset
|
48 |
+
|
49 |
+
~34,000 cards from CharacterHub and another ~80,000 cards from Janitor were used as initial dataset, as captured in period between August and September 2023.
|
50 |
+
|
51 |
+
Dataset will not be released, unless explicit permission to do so would be granted from both Chub and Janitor.
|
52 |
+
|
53 |
+
## Training procedure
|
54 |
+
|
55 |
+
|
56 |
+
The following `bitsandbytes` quantization config was used during training:
|
57 |
+
- quant_method: bitsandbytes
|
58 |
+
- load_in_8bit: True
|
59 |
+
- load_in_4bit: False
|
60 |
+
- llm_int8_threshold: 6.0
|
61 |
+
- llm_int8_skip_modules: None
|
62 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
63 |
+
- llm_int8_has_fp16_weight: False
|
64 |
+
- bnb_4bit_quant_type: fp4
|
65 |
+
- bnb_4bit_use_double_quant: False
|
66 |
+
- bnb_4bit_compute_dtype: float32
|
67 |
+
### Framework versions
|
68 |
+
|
69 |
+
|
70 |
+
- PEFT 0.6.0.dev0
|