RichardErkhov
commited on
Commit
•
0ad47bb
1
Parent(s):
c7ac8ff
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,245 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
smol_llama-220M-openhermes - GGUF
|
11 |
+
- Model creator: https://huggingface.co/BEE-spoke-data/
|
12 |
+
- Original model: https://huggingface.co/BEE-spoke-data/smol_llama-220M-openhermes/
|
13 |
+
|
14 |
+
|
15 |
+
| Name | Quant method | Size |
|
16 |
+
| ---- | ---- | ---- |
|
17 |
+
| [smol_llama-220M-openhermes.Q2_K.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q2_K.gguf) | Q2_K | 0.09GB |
|
18 |
+
| [smol_llama-220M-openhermes.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.IQ3_XS.gguf) | IQ3_XS | 0.1GB |
|
19 |
+
| [smol_llama-220M-openhermes.IQ3_S.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.IQ3_S.gguf) | IQ3_S | 0.1GB |
|
20 |
+
| [smol_llama-220M-openhermes.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q3_K_S.gguf) | Q3_K_S | 0.1GB |
|
21 |
+
| [smol_llama-220M-openhermes.IQ3_M.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.IQ3_M.gguf) | IQ3_M | 0.1GB |
|
22 |
+
| [smol_llama-220M-openhermes.Q3_K.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q3_K.gguf) | Q3_K | 0.11GB |
|
23 |
+
| [smol_llama-220M-openhermes.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q3_K_M.gguf) | Q3_K_M | 0.11GB |
|
24 |
+
| [smol_llama-220M-openhermes.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q3_K_L.gguf) | Q3_K_L | 0.11GB |
|
25 |
+
| [smol_llama-220M-openhermes.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.IQ4_XS.gguf) | IQ4_XS | 0.12GB |
|
26 |
+
| [smol_llama-220M-openhermes.Q4_0.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q4_0.gguf) | Q4_0 | 0.12GB |
|
27 |
+
| [smol_llama-220M-openhermes.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.IQ4_NL.gguf) | IQ4_NL | 0.12GB |
|
28 |
+
| [smol_llama-220M-openhermes.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q4_K_S.gguf) | Q4_K_S | 0.12GB |
|
29 |
+
| [smol_llama-220M-openhermes.Q4_K.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q4_K.gguf) | Q4_K | 0.13GB |
|
30 |
+
| [smol_llama-220M-openhermes.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q4_K_M.gguf) | Q4_K_M | 0.13GB |
|
31 |
+
| [smol_llama-220M-openhermes.Q4_1.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q4_1.gguf) | Q4_1 | 0.13GB |
|
32 |
+
| [smol_llama-220M-openhermes.Q5_0.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q5_0.gguf) | Q5_0 | 0.14GB |
|
33 |
+
| [smol_llama-220M-openhermes.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q5_K_S.gguf) | Q5_K_S | 0.14GB |
|
34 |
+
| [smol_llama-220M-openhermes.Q5_K.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q5_K.gguf) | Q5_K | 0.15GB |
|
35 |
+
| [smol_llama-220M-openhermes.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q5_K_M.gguf) | Q5_K_M | 0.15GB |
|
36 |
+
| [smol_llama-220M-openhermes.Q5_1.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q5_1.gguf) | Q5_1 | 0.16GB |
|
37 |
+
| [smol_llama-220M-openhermes.Q6_K.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q6_K.gguf) | Q6_K | 0.17GB |
|
38 |
+
| [smol_llama-220M-openhermes.Q8_0.gguf](https://huggingface.co/RichardErkhov/BEE-spoke-data_-_smol_llama-220M-openhermes-gguf/blob/main/smol_llama-220M-openhermes.Q8_0.gguf) | Q8_0 | 0.22GB |
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
|
43 |
+
Original model description:
|
44 |
+
---
|
45 |
+
license: apache-2.0
|
46 |
+
datasets:
|
47 |
+
- teknium/openhermes
|
48 |
+
base_model: BEE-spoke-data/smol_llama-220M-GQA
|
49 |
+
inference:
|
50 |
+
parameters:
|
51 |
+
do_sample: true
|
52 |
+
renormalize_logits: true
|
53 |
+
temperature: 0.25
|
54 |
+
top_p: 0.95
|
55 |
+
top_k: 50
|
56 |
+
min_new_tokens: 2
|
57 |
+
max_new_tokens: 96
|
58 |
+
repetition_penalty: 1.03
|
59 |
+
no_repeat_ngram_size: 5
|
60 |
+
epsilon_cutoff: 0.0008
|
61 |
+
widget:
|
62 |
+
- text: "Below is an instruction that describes a task, paired with an input that\
|
63 |
+
\ provides further context. Write a response that appropriately completes the\
|
64 |
+
\ request. \n \n### Instruction: \n \nWrite an ode to Chipotle burritos.\
|
65 |
+
\ \n \n### Response: \n"
|
66 |
+
example_title: burritos
|
67 |
+
model-index:
|
68 |
+
- name: smol_llama-220M-openhermes
|
69 |
+
results:
|
70 |
+
- task:
|
71 |
+
type: text-generation
|
72 |
+
name: Text Generation
|
73 |
+
dataset:
|
74 |
+
name: AI2 Reasoning Challenge (25-Shot)
|
75 |
+
type: ai2_arc
|
76 |
+
config: ARC-Challenge
|
77 |
+
split: test
|
78 |
+
args:
|
79 |
+
num_few_shot: 25
|
80 |
+
metrics:
|
81 |
+
- type: acc_norm
|
82 |
+
value: 25.17
|
83 |
+
name: normalized accuracy
|
84 |
+
source:
|
85 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BEE-spoke-data/smol_llama-220M-openhermes
|
86 |
+
name: Open LLM Leaderboard
|
87 |
+
- task:
|
88 |
+
type: text-generation
|
89 |
+
name: Text Generation
|
90 |
+
dataset:
|
91 |
+
name: HellaSwag (10-Shot)
|
92 |
+
type: hellaswag
|
93 |
+
split: validation
|
94 |
+
args:
|
95 |
+
num_few_shot: 10
|
96 |
+
metrics:
|
97 |
+
- type: acc_norm
|
98 |
+
value: 28.98
|
99 |
+
name: normalized accuracy
|
100 |
+
source:
|
101 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BEE-spoke-data/smol_llama-220M-openhermes
|
102 |
+
name: Open LLM Leaderboard
|
103 |
+
- task:
|
104 |
+
type: text-generation
|
105 |
+
name: Text Generation
|
106 |
+
dataset:
|
107 |
+
name: MMLU (5-Shot)
|
108 |
+
type: cais/mmlu
|
109 |
+
config: all
|
110 |
+
split: test
|
111 |
+
args:
|
112 |
+
num_few_shot: 5
|
113 |
+
metrics:
|
114 |
+
- type: acc
|
115 |
+
value: 26.17
|
116 |
+
name: accuracy
|
117 |
+
source:
|
118 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BEE-spoke-data/smol_llama-220M-openhermes
|
119 |
+
name: Open LLM Leaderboard
|
120 |
+
- task:
|
121 |
+
type: text-generation
|
122 |
+
name: Text Generation
|
123 |
+
dataset:
|
124 |
+
name: TruthfulQA (0-shot)
|
125 |
+
type: truthful_qa
|
126 |
+
config: multiple_choice
|
127 |
+
split: validation
|
128 |
+
args:
|
129 |
+
num_few_shot: 0
|
130 |
+
metrics:
|
131 |
+
- type: mc2
|
132 |
+
value: 43.08
|
133 |
+
source:
|
134 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BEE-spoke-data/smol_llama-220M-openhermes
|
135 |
+
name: Open LLM Leaderboard
|
136 |
+
- task:
|
137 |
+
type: text-generation
|
138 |
+
name: Text Generation
|
139 |
+
dataset:
|
140 |
+
name: Winogrande (5-shot)
|
141 |
+
type: winogrande
|
142 |
+
config: winogrande_xl
|
143 |
+
split: validation
|
144 |
+
args:
|
145 |
+
num_few_shot: 5
|
146 |
+
metrics:
|
147 |
+
- type: acc
|
148 |
+
value: 52.01
|
149 |
+
name: accuracy
|
150 |
+
source:
|
151 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BEE-spoke-data/smol_llama-220M-openhermes
|
152 |
+
name: Open LLM Leaderboard
|
153 |
+
- task:
|
154 |
+
type: text-generation
|
155 |
+
name: Text Generation
|
156 |
+
dataset:
|
157 |
+
name: GSM8k (5-shot)
|
158 |
+
type: gsm8k
|
159 |
+
config: main
|
160 |
+
split: test
|
161 |
+
args:
|
162 |
+
num_few_shot: 5
|
163 |
+
metrics:
|
164 |
+
- type: acc
|
165 |
+
value: 0.61
|
166 |
+
name: accuracy
|
167 |
+
source:
|
168 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BEE-spoke-data/smol_llama-220M-openhermes
|
169 |
+
name: Open LLM Leaderboard
|
170 |
+
---
|
171 |
+
|
172 |
+
|
173 |
+
# BEE-spoke-data/smol_llama-220M-openhermes
|
174 |
+
|
175 |
+
> Please note that this is an experiment, and the model has limitations because it is smol.
|
176 |
+
|
177 |
+
|
178 |
+
prompt format is alpaca
|
179 |
+
|
180 |
+
|
181 |
+
```
|
182 |
+
Below is an instruction that describes a task, paired with an input that
|
183 |
+
provides further context. Write a response that appropriately completes
|
184 |
+
the request.
|
185 |
+
|
186 |
+
### Instruction:
|
187 |
+
|
188 |
+
How can I increase my meme production/output? Currently, I only create them in ancient babylonian which is time consuming.
|
189 |
+
|
190 |
+
### Inputs:
|
191 |
+
|
192 |
+
### Response:
|
193 |
+
```
|
194 |
+
|
195 |
+
It was trained on inputs so if you have inputs (like some text to ask a question about) then include it under `### Inputs:`
|
196 |
+
|
197 |
+
|
198 |
+
## Example
|
199 |
+
|
200 |
+
Output on the text above ^. The inference API is set to sample with low temp so you should see (_at least slightly_) different generations each time.
|
201 |
+
|
202 |
+
|
203 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/60bccec062080d33f875cd0c/0nFP2jsBkritnryKmI8NV.png)
|
204 |
+
|
205 |
+
Note that the inference API parameters used here are an initial educated guess, and may be updated over time:
|
206 |
+
|
207 |
+
```yml
|
208 |
+
inference:
|
209 |
+
parameters:
|
210 |
+
do_sample: true
|
211 |
+
renormalize_logits: true
|
212 |
+
temperature: 0.25
|
213 |
+
top_p: 0.95
|
214 |
+
top_k: 50
|
215 |
+
min_new_tokens: 2
|
216 |
+
max_new_tokens: 96
|
217 |
+
repetition_penalty: 1.03
|
218 |
+
no_repeat_ngram_size: 5
|
219 |
+
epsilon_cutoff: 0.0008
|
220 |
+
```
|
221 |
+
|
222 |
+
Feel free to experiment with the parameters using the model in Python and let us know if you have improved results with other params!
|
223 |
+
|
224 |
+
## Data
|
225 |
+
|
226 |
+
Note that **this checkpoint** was fine-tuned on `teknium/openhermes`, which is generated/synthetic data by an OpenAI model. This means usage of this checkpoint should follow their terms of use: https://openai.com/policies/terms-of-use
|
227 |
+
|
228 |
+
|
229 |
+
---
|
230 |
+
|
231 |
+
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
232 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_BEE-spoke-data__smol_llama-220M-openhermes)
|
233 |
+
|
234 |
+
| Metric |Value|
|
235 |
+
|---------------------------------|----:|
|
236 |
+
|Avg. |29.34|
|
237 |
+
|AI2 Reasoning Challenge (25-Shot)|25.17|
|
238 |
+
|HellaSwag (10-Shot) |28.98|
|
239 |
+
|MMLU (5-Shot) |26.17|
|
240 |
+
|TruthfulQA (0-shot) |43.08|
|
241 |
+
|Winogrande (5-shot) |52.01|
|
242 |
+
|GSM8k (5-shot) | 0.61|
|
243 |
+
|
244 |
+
|
245 |
+
|