RichardErkhov
commited on
Commit
•
1399a9a
1
Parent(s):
7911997
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,302 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Quantization made by Richard Erkhov.
|
2 |
+
|
3 |
+
[Github](https://github.com/RichardErkhov)
|
4 |
+
|
5 |
+
[Discord](https://discord.gg/pvy7H8DZMG)
|
6 |
+
|
7 |
+
[Request more models](https://github.com/RichardErkhov/quant_request)
|
8 |
+
|
9 |
+
|
10 |
+
Llama-160M-Chat-v1 - GGUF
|
11 |
+
- Model creator: https://huggingface.co/Felladrin/
|
12 |
+
- Original model: https://huggingface.co/Felladrin/Llama-160M-Chat-v1/
|
13 |
+
|
14 |
+
|
15 |
+
| Name | Quant method | Size |
|
16 |
+
| ---- | ---- | ---- |
|
17 |
+
| [Llama-160M-Chat-v1.Q2_K.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q2_K.gguf) | Q2_K | 0.07GB |
|
18 |
+
| [Llama-160M-Chat-v1.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.IQ3_XS.gguf) | IQ3_XS | 0.07GB |
|
19 |
+
| [Llama-160M-Chat-v1.IQ3_S.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.IQ3_S.gguf) | IQ3_S | 0.07GB |
|
20 |
+
| [Llama-160M-Chat-v1.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q3_K_S.gguf) | Q3_K_S | 0.07GB |
|
21 |
+
| [Llama-160M-Chat-v1.IQ3_M.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.IQ3_M.gguf) | IQ3_M | 0.08GB |
|
22 |
+
| [Llama-160M-Chat-v1.Q3_K.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q3_K.gguf) | Q3_K | 0.08GB |
|
23 |
+
| [Llama-160M-Chat-v1.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q3_K_M.gguf) | Q3_K_M | 0.08GB |
|
24 |
+
| [Llama-160M-Chat-v1.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q3_K_L.gguf) | Q3_K_L | 0.08GB |
|
25 |
+
| [Llama-160M-Chat-v1.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.IQ4_XS.gguf) | IQ4_XS | 0.09GB |
|
26 |
+
| [Llama-160M-Chat-v1.Q4_0.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q4_0.gguf) | Q4_0 | 0.09GB |
|
27 |
+
| [Llama-160M-Chat-v1.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.IQ4_NL.gguf) | IQ4_NL | 0.09GB |
|
28 |
+
| [Llama-160M-Chat-v1.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q4_K_S.gguf) | Q4_K_S | 0.09GB |
|
29 |
+
| [Llama-160M-Chat-v1.Q4_K.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q4_K.gguf) | Q4_K | 0.1GB |
|
30 |
+
| [Llama-160M-Chat-v1.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q4_K_M.gguf) | Q4_K_M | 0.1GB |
|
31 |
+
| [Llama-160M-Chat-v1.Q4_1.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q4_1.gguf) | Q4_1 | 0.1GB |
|
32 |
+
| [Llama-160M-Chat-v1.Q5_0.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q5_0.gguf) | Q5_0 | 0.11GB |
|
33 |
+
| [Llama-160M-Chat-v1.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q5_K_S.gguf) | Q5_K_S | 0.11GB |
|
34 |
+
| [Llama-160M-Chat-v1.Q5_K.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q5_K.gguf) | Q5_K | 0.11GB |
|
35 |
+
| [Llama-160M-Chat-v1.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q5_K_M.gguf) | Q5_K_M | 0.11GB |
|
36 |
+
| [Llama-160M-Chat-v1.Q5_1.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q5_1.gguf) | Q5_1 | 0.12GB |
|
37 |
+
| [Llama-160M-Chat-v1.Q6_K.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q6_K.gguf) | Q6_K | 0.12GB |
|
38 |
+
| [Llama-160M-Chat-v1.Q8_0.gguf](https://huggingface.co/RichardErkhov/Felladrin_-_Llama-160M-Chat-v1-gguf/blob/main/Llama-160M-Chat-v1.Q8_0.gguf) | Q8_0 | 0.16GB |
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
|
43 |
+
Original model description:
|
44 |
+
---
|
45 |
+
language:
|
46 |
+
- en
|
47 |
+
license: apache-2.0
|
48 |
+
tags:
|
49 |
+
- text-generation
|
50 |
+
base_model: JackFram/llama-160m
|
51 |
+
datasets:
|
52 |
+
- ehartford/wizard_vicuna_70k_unfiltered
|
53 |
+
- totally-not-an-llm/EverythingLM-data-V3
|
54 |
+
- Open-Orca/SlimOrca-Dedup
|
55 |
+
- databricks/databricks-dolly-15k
|
56 |
+
- THUDM/webglm-qa
|
57 |
+
widget:
|
58 |
+
- messages:
|
59 |
+
- role: system
|
60 |
+
content: You are a helpful assistant, who answers with empathy.
|
61 |
+
- role: user
|
62 |
+
content: Got a question for you!
|
63 |
+
- role: assistant
|
64 |
+
content: "Sure! What's it?"
|
65 |
+
- role: user
|
66 |
+
content: Why do you love cats so much!? 🐈
|
67 |
+
- messages:
|
68 |
+
- role: system
|
69 |
+
content: "You are a helpful assistant who answers user's questions with empathy."
|
70 |
+
- role: user
|
71 |
+
content: Who is Mona Lisa?
|
72 |
+
- messages:
|
73 |
+
- role: system
|
74 |
+
content: You are a helpful assistant who provides concise responses.
|
75 |
+
- role: user
|
76 |
+
content: Heya!
|
77 |
+
- role: assistant
|
78 |
+
content: Hi! How may I help you today?
|
79 |
+
- role: user
|
80 |
+
content: I need to build a simple website. Where should I start learning about web development?
|
81 |
+
- messages:
|
82 |
+
- role: user
|
83 |
+
content: Invited some friends to come home today. Give me some ideas for games to play with them!
|
84 |
+
- messages:
|
85 |
+
- role: system
|
86 |
+
content: "You are a helpful assistant who answers user's questions with details and curiosity."
|
87 |
+
- role: user
|
88 |
+
content: What are some potential applications for quantum computing?
|
89 |
+
- messages:
|
90 |
+
- role: system
|
91 |
+
content: You are a helpful assistant who gives creative responses.
|
92 |
+
- role: user
|
93 |
+
content: Write the specs of a game about mages in a fantasy world.
|
94 |
+
- messages:
|
95 |
+
- role: system
|
96 |
+
content: "You are a helpful assistant who answers user's questions with details."
|
97 |
+
- role: user
|
98 |
+
content: Tell me about the pros and cons of social media.
|
99 |
+
- messages:
|
100 |
+
- role: system
|
101 |
+
content: "You are a helpful assistant who answers user's questions with confidence."
|
102 |
+
- role: user
|
103 |
+
content: What is a dog?
|
104 |
+
- role: assistant
|
105 |
+
content: 'A dog is a four-legged, domesticated animal that is a member of the class Mammalia,
|
106 |
+
which includes all mammals. Dogs are known for their loyalty, playfulness, and
|
107 |
+
ability to be trained for various tasks. They are also used for hunting, herding,
|
108 |
+
and as service animals.'
|
109 |
+
- role: user
|
110 |
+
content: What is the color of an apple?
|
111 |
+
inference:
|
112 |
+
parameters:
|
113 |
+
max_new_tokens: 250
|
114 |
+
penalty_alpha: 0.5
|
115 |
+
top_k: 4
|
116 |
+
repetition_penalty: 1.01
|
117 |
+
model-index:
|
118 |
+
- name: Llama-160M-Chat-v1
|
119 |
+
results:
|
120 |
+
- task:
|
121 |
+
type: text-generation
|
122 |
+
name: Text Generation
|
123 |
+
dataset:
|
124 |
+
name: AI2 Reasoning Challenge (25-Shot)
|
125 |
+
type: ai2_arc
|
126 |
+
config: ARC-Challenge
|
127 |
+
split: test
|
128 |
+
args:
|
129 |
+
num_few_shot: 25
|
130 |
+
metrics:
|
131 |
+
- type: acc_norm
|
132 |
+
value: 24.74
|
133 |
+
name: normalized accuracy
|
134 |
+
source:
|
135 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Llama-160M-Chat-v1
|
136 |
+
name: Open LLM Leaderboard
|
137 |
+
- task:
|
138 |
+
type: text-generation
|
139 |
+
name: Text Generation
|
140 |
+
dataset:
|
141 |
+
name: HellaSwag (10-Shot)
|
142 |
+
type: hellaswag
|
143 |
+
split: validation
|
144 |
+
args:
|
145 |
+
num_few_shot: 10
|
146 |
+
metrics:
|
147 |
+
- type: acc_norm
|
148 |
+
value: 35.29
|
149 |
+
name: normalized accuracy
|
150 |
+
source:
|
151 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Llama-160M-Chat-v1
|
152 |
+
name: Open LLM Leaderboard
|
153 |
+
- task:
|
154 |
+
type: text-generation
|
155 |
+
name: Text Generation
|
156 |
+
dataset:
|
157 |
+
name: MMLU (5-Shot)
|
158 |
+
type: cais/mmlu
|
159 |
+
config: all
|
160 |
+
split: test
|
161 |
+
args:
|
162 |
+
num_few_shot: 5
|
163 |
+
metrics:
|
164 |
+
- type: acc
|
165 |
+
value: 26.13
|
166 |
+
name: accuracy
|
167 |
+
source:
|
168 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Llama-160M-Chat-v1
|
169 |
+
name: Open LLM Leaderboard
|
170 |
+
- task:
|
171 |
+
type: text-generation
|
172 |
+
name: Text Generation
|
173 |
+
dataset:
|
174 |
+
name: TruthfulQA (0-shot)
|
175 |
+
type: truthful_qa
|
176 |
+
config: multiple_choice
|
177 |
+
split: validation
|
178 |
+
args:
|
179 |
+
num_few_shot: 0
|
180 |
+
metrics:
|
181 |
+
- type: mc2
|
182 |
+
value: 44.16
|
183 |
+
source:
|
184 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Llama-160M-Chat-v1
|
185 |
+
name: Open LLM Leaderboard
|
186 |
+
- task:
|
187 |
+
type: text-generation
|
188 |
+
name: Text Generation
|
189 |
+
dataset:
|
190 |
+
name: Winogrande (5-shot)
|
191 |
+
type: winogrande
|
192 |
+
config: winogrande_xl
|
193 |
+
split: validation
|
194 |
+
args:
|
195 |
+
num_few_shot: 5
|
196 |
+
metrics:
|
197 |
+
- type: acc
|
198 |
+
value: 51.3
|
199 |
+
name: accuracy
|
200 |
+
source:
|
201 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Llama-160M-Chat-v1
|
202 |
+
name: Open LLM Leaderboard
|
203 |
+
- task:
|
204 |
+
type: text-generation
|
205 |
+
name: Text Generation
|
206 |
+
dataset:
|
207 |
+
name: GSM8k (5-shot)
|
208 |
+
type: gsm8k
|
209 |
+
config: main
|
210 |
+
split: test
|
211 |
+
args:
|
212 |
+
num_few_shot: 5
|
213 |
+
metrics:
|
214 |
+
- type: acc
|
215 |
+
value: 0.0
|
216 |
+
name: accuracy
|
217 |
+
source:
|
218 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Llama-160M-Chat-v1
|
219 |
+
name: Open LLM Leaderboard
|
220 |
+
---
|
221 |
+
|
222 |
+
# A Llama Chat Model of 160M Parameters
|
223 |
+
|
224 |
+
- Base model: [JackFram/llama-160m](https://huggingface.co/JackFram/llama-160m)
|
225 |
+
- Datasets:
|
226 |
+
- [ehartford/wizard_vicuna_70k_unfiltered](https://huggingface.co/datasets/ehartford/wizard_vicuna_70k_unfiltered)
|
227 |
+
- [totally-not-an-llm/EverythingLM-data-V3](https://huggingface.co/datasets/totally-not-an-llm/EverythingLM-data-V3)
|
228 |
+
- [Open-Orca/SlimOrca-Dedup](https://huggingface.co/datasets/Open-Orca/SlimOrca-Dedup)
|
229 |
+
- [databricks/databricks-dolly-15k](https://huggingface.co/datasets/databricks/databricks-dolly-15k)
|
230 |
+
- [THUDM/webglm-qa](https://huggingface.co/datasets/THUDM/webglm-qa)
|
231 |
+
- Availability in other ML formats:
|
232 |
+
- GGUF: [Felladrin/gguf-Llama-160M-Chat-v1](https://huggingface.co/Felladrin/gguf-Llama-160M-Chat-v1)
|
233 |
+
- ONNX: [Felladrin/onnx-Llama-160M-Chat-v1](https://huggingface.co/Felladrin/onnx-Llama-160M-Chat-v1)
|
234 |
+
- MLC: [Felladrin/mlc-q4f16-Llama-160M-Chat-v1](https://huggingface.co/Felladrin/mlc-q4f16-Llama-160M-Chat-v1)
|
235 |
+
- MLX: [mlx-community/Llama-160M-Chat-v1-4bit-mlx](https://huggingface.co/mlx-community/Llama-160M-Chat-v1-4bit-mlx)
|
236 |
+
|
237 |
+
## Recommended Prompt Format
|
238 |
+
|
239 |
+
```
|
240 |
+
<|im_start|>system
|
241 |
+
{system_message}<|im_end|>
|
242 |
+
<|im_start|>user
|
243 |
+
{user_message}<|im_end|>
|
244 |
+
<|im_start|>assistant
|
245 |
+
```
|
246 |
+
|
247 |
+
## Recommended Inference Parameters
|
248 |
+
|
249 |
+
```yml
|
250 |
+
penalty_alpha: 0.5
|
251 |
+
top_k: 4
|
252 |
+
repetition_penalty: 1.01
|
253 |
+
```
|
254 |
+
|
255 |
+
## Usage Example
|
256 |
+
|
257 |
+
```python
|
258 |
+
from transformers import pipeline
|
259 |
+
|
260 |
+
generate = pipeline("text-generation", "Felladrin/Llama-160M-Chat-v1")
|
261 |
+
|
262 |
+
messages = [
|
263 |
+
{
|
264 |
+
"role": "system",
|
265 |
+
"content": "You are a helpful assistant who answers user's questions with details and curiosity.",
|
266 |
+
},
|
267 |
+
{
|
268 |
+
"role": "user",
|
269 |
+
"content": "What are some potential applications for quantum computing?",
|
270 |
+
},
|
271 |
+
]
|
272 |
+
|
273 |
+
prompt = generate.tokenizer.apply_chat_template(
|
274 |
+
messages, tokenize=False, add_generation_prompt=True
|
275 |
+
)
|
276 |
+
|
277 |
+
output = generate(
|
278 |
+
prompt,
|
279 |
+
max_new_tokens=1024,
|
280 |
+
penalty_alpha=0.5,
|
281 |
+
top_k=4,
|
282 |
+
repetition_penalty=1.01,
|
283 |
+
)
|
284 |
+
|
285 |
+
print(output[0]["generated_text"])
|
286 |
+
```
|
287 |
+
|
288 |
+
## [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
289 |
+
|
290 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_Felladrin__Llama-160M-Chat-v1)
|
291 |
+
|
292 |
+
| Metric |Value|
|
293 |
+
|---------------------------------|----:|
|
294 |
+
|Avg. |30.27|
|
295 |
+
|AI2 Reasoning Challenge (25-Shot)|24.74|
|
296 |
+
|HellaSwag (10-Shot) |35.29|
|
297 |
+
|MMLU (5-Shot) |26.13|
|
298 |
+
|TruthfulQA (0-shot) |44.16|
|
299 |
+
|Winogrande (5-shot) |51.30|
|
300 |
+
|GSM8k (5-shot) | 0.00|
|
301 |
+
|
302 |
+
|