File size: 7,378 Bytes
fa8ac70
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
09c3cae
fa8ac70
 
 
 
 
 
1b8f33e
fa8ac70
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
784d494
fa8ac70
 
 
 
 
 
09c3cae
 
 
fa8ac70
 
784d494
1125a68
 
 
fa8ac70
 
784d494
fa8ac70
 
 
 
 
 
 
 
 
1b8f33e
fa8ac70
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
784d494
fa8ac70
 
 
 
 
49c3d06
fa8ac70
 
 
 
 
49c3d06
fa8ac70
 
 
 
 
 
 
 
 
 
 
49c3d06
fa8ac70
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
63a9f49
fa8ac70
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
---
language:
  - en
thumbnail: null
tags:
  - text generation
  - conversational
pipeline_tag: text-generation
inference: false
---
<h1 style="text-align: center">Pygmalion 13b</h1>
<h2 style="text-align: center">A conversational LLaMA fine-tune.</h2>

-----
Install 0cc4m's ->latest<- update from his GPTQ+KoboldAI fork, it has proper support for 8bit models in this repo's format out of the box on both windows & Linux:

- https://github.com/0cc4m/KoboldAI/tree/latestgptq

With this fix applied (It will output gibberish without this patch in place!):
- https://github.com/0cc4m/GPTQ-for-LLaMa/pull/14

GPTQ via Ooba UI may not need this patch.

## Eval / Benchmark scores

Current evals out of the Pygmalion-13b/7b model: <br>
<html>
<head>
	<style>
		table {
			border:1px solid #b3adad;
			border-collapse:collapse;
			padding:5px;
		}
		table th {
			border:1px solid #b3adad;
			padding:5px;
			background: #f0f0f0;
			color: #313030;
		}
		table td {
			border:1px solid #b3adad;
			text-align:center;
			padding:5px;
			background: #ffffff;
			color: #313030;
		}
	</style>
</head>
<body>
	<table>
		<thead>
			<tr>
				<th>Model:</th>
				<th>Wikitext2</th>
				<th>Ptb-New</th>
				<th>C4-New</th>
			</tr>
		</thead>
		<tbody>
			<tr>
				<td>Pygmalion 13b - 16bit</td>
				<td>5.710726737976074</td>
				<td>23.633684158325195</td>
				<td>7.6324849128723145</td>
			</tr>
            <tr>
				<td>Pygmalion 13b - 8bit <br> [act-order]</td>
				<td>5.711935997009277</td>
				<td>23.654993057250977</td>
				<td>7.632820129394531</td>
			</tr>
            <tr>
				<td>Pygmalion 7b - 16bit</td>
				<td>5.654823303222656</td>
				<td>40.83400344848633</td>
				<td>7.429622173309326</td>
			</tr>
            <tr>
				<td>Pygmalion 7b - 8bit <br> [act-order]</td>
				<td>5.656460285186768</td>
				<td>40.79701232910156</td>
				<td>7.432109832763672</td>
			</tr>
            <tr>
				<td>Pygmalion 7b - 4bit <br> [act-order]</td>
				<td>6.2477378845215</td>
				<td>46.5129699707031</td>
				<td>7.8470954895020</td>
			</tr>
		</tbody>
	</table>
</body>
</html>

Current evals out of the Metharme-13b/7b model: <br>
<html>
<head>
	<style>
		table {
			border:1px solid #b3adad;
			border-collapse:collapse;
			padding:5px;
		}
		table th {
			border:1px solid #b3adad;
			padding:5px;
			background: #f0f0f0;
			color: #313030;
		}
		table td {
			border:1px solid #b3adad;
			text-align:center;
			padding:5px;
			background: #ffffff;
			color: #313030;
		}
	</style>
</head>
<body>
	<table>
		<thead>
			<tr>
				<th>Model:</th>
				<th>Wikitext2</th>
				<th>Ptb-New</th>
				<th>C4-New</th>
			</tr>
		</thead>
		<tbody>
			<tr>
				<td>Metharme 13b - 16bit</td>
				<td>5.253076553344727</td>
				<td>27.53407859802246</td>
				<td>7.038073539733887</td>
			</tr>
          <tr>
				<td>Metharme 13b - 8bit <br> [act-order]</td>
				<td>5.253607273101807</td>
				<td>27.52388572692871</td>
				<td>7.038473129272461</td>
			</tr>
          <tr>
				<td>Metharme 13b - 8bit <br>[true-sequential & 128g]</td>
				<td>5.2532830238342285</td>
				<td>27.54250144958496</td>
				<td>7.038838863372803</td>
			</tr>
          <tr>
				<td>Metharme 13b - 4bit <br>[true-sequential & 128g]</td>
				<td>5.420501708984375</td>
				<td>28.37093734741211</td>
				<td>7.1930413246154785</td>
			</tr>
          <tr>
				<td>Metharme 7b - 16bit</td>
				<td>5.7208476066589355</td>
				<td>41.61103439331055</td>
				<td>7.512405872344971</td>
			</tr>
          <tr>
				<td>Metharme 7b - 4bit <br>[act-order]</td>
				<td>6.2369050979614</td>
				<td>47.5177230834960</td>
				<td>7.9044938087463</td>
			</tr>
		</tbody>
	</table>
</body>
</html>

<hr>

-----
## Model Details:

Converted from the XORs weights from PygmalionAI's release https://huggingface.co/PygmalionAI/pygmalion-13b

Pygmalion 13b is a dialogue model based on Meta's LLaMA-13b.

This is version 1. It has been fine-tuned using a subset of the data from Pygmalion-6B-v8-pt4, 
for those of you familiar with the project.

The current Pygmalion-13b has been trained as a LoRA, then merged down to the base model for distribuition.

It has also been quantized down to 8Bit using the GPTQ library available here: https://github.com/0cc4m/GPTQ-for-LLaMa

```
python llama.py .\TehVenom_Metharme-13b-Merged c4 --wbits 8 --act-order --save_safetensors Metharme-13b-GPTQ-8bit.act-order.safetensors
```

## Model Details:

Pygmalion 13b is a dialogue model based on Meta's LLaMA-13b.

This is version 1. It has been fine-tuned using a subset of the data from Pygmalion-6B-v8-pt4, 
for those of you familiar with the project.

The current Pygmalion-13b has been trained as a LoRA, then merged down to the base model for distribuition. 

## Applying the XORs

This models has the XOR files pre-applied out of the box.
Converted from the XORs weights from PygmalionAI's release https://huggingface.co/PygmalionAI/pygmalion-13b

## Prompting

The model was trained on the usual Pygmalion persona + chat format, so any of the usual UIs should already handle everything correctly. If you're using the model directly, this is the expected formatting:
```
[CHARACTER]'s Persona: [A few sentences about the character you want the model to play]
<START>
[DIALOGUE HISTORY]
You: [User's input message here]
[CHARACTER]:
```

Where `[CHARACTER]` is, as you can probably guess, the name of the character you want the model to portray, `<START>` should be used verbatim as a delimiter token to separate persona and scenario data from the dialogue, and `[DIALOGUE HISTORY]` is a sliding window of chat history so the model can have conversational context to draw from. Here's a concrete example:

```
Assistant's Persona: Assistant is a highly intelligent language model trained to comply with user requests.
<START>
Assistant: Hello! How may I help you today?
You: What is Zork?
Assistant:
```

Which will generate something like:

```
 Zork is an interactive fiction computer game created in the 1970s by Infocom, Inc., which was later acquired by Activision Blizzard. It is widely considered one of the most influential games ever made and has been credited with popularizing text-based adventure games. The original version of Zork was written in the programming language MACRO-10, but it was ported to many other platforms over the years."
```

The model will automatically emit an end-of-text token (`</s>`) when it judges that the response is complete.


## Other notes

- When prompted correctly, the model will always start by generating a BOS token. This behavior is an accidental side-effect which we plan to address in future model versions and should not be relied upon.
- The model was trained as a LoRA with a somewhat unorthodox configuration which causes errors when used with the current version of `peft`, hence we release it as a full model instead.


## Limitations and biases

The intended use-case for this model is fictional conversation for entertainment purposes. Any other sort of usage is out of scope.

As such, it was **not** fine-tuned to be safe and harmless: the base model _and_ this fine-tune have been trained on data known to contain profanity and texts that are lewd or otherwise offensive. It may produce socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive. Outputs might often be factually wrong or misleading.