File size: 8,487 Bytes
c36152d
c360bdf
a39a670
 
c360bdf
a39a670
c360bdf
 
a39a670
 
 
 
 
 
c360bdf
cdd4a85
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c360bdf
 
 
 
738e353
 
a39a670
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c36152d
c360bdf
 
 
 
 
 
 
 
 
 
 
1036205
5b690ee
1036205
fd57078
c360bdf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
738e353
 
c360bdf
a39a670
cdd4a85
a39a670
 
 
 
 
 
 
 
 
 
 
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
---
language:
- en
license: apache-2.0
tags:
- text-generation
base_model: BEE-spoke-data/smol_llama-101M-GQA
datasets:
- Open-Orca/SlimOrca-Dedup
- VMware/open-instruct
- LDJnr/Capybara
- cognitivecomputations/ultrachat-uncensored
- starfishmedical/webGPT_x_dolly
- THUDM/webglm-qa
widget:
  - messages:
      - role: system
        content: You are a helpful assistant who gives creative responses.
      - role: user
        content: Write the background story of a game about wizards and llamas in a sci-fi world.
  - messages:
      - role: system
        content: A friendly chat between a user and an assistant.
      - role: user
        content: Got a question for you!
      - role: assistant
        content: "Sure! What's it?"
      - role: user
        content: I need to build a simple website. Where should I start learning about web development?
  - messages:
      - role: system
        content: "You are a helpful assistant who provides concise answers to the user's questions."
      - role: user
        content: How to become more healthy?
  - messages:
      - role: system
        content: You are a helpful assistant, who always answers with empathy.
      - role: user
        content: List the pros and cons of social media.
  - messages:
      - role: system
        content: You are a helpful assistant, who always answers with empathy.
      - role: user
        content: Hello!
      - role: assistant
        content: Hi! How can I help you today?
      - role: user
        content: 'Take a look at the info below.

          - The tape inside the VHS cassettes is very delicate and can be easily ruined,
          making them unplayable and unrepairable. The reason the tape deteriorates is that
          the magnetic charge needed for them to work is not permanent, and the magnetic
          particles end up losing their charge in a process known as remanence decay. These
          particles could also become demagnetised via being stored too close to a magnetic
          source.
      
          - One of the most significant issues with VHS tapes is that they have moving parts,
          meaning that there are more occasions when something can go wrong, damaging your
          footage or preventing it from playing back. The tape itself is a prominent cause
          of this, and tape slippage can occur. Tapes slippage can be caused when the tape
          loses its tension, or it has become warped. These problems can occur in storage
          due to high temperatures or frequent changes in humidity.
      
          - VHS tapes deteriorate over time from infrequent or overuse. Neglect means mold
          and dirt, while overuse can lead to scratches and technical difficulties. This
          is why old VHS tapes inevitably experience malfunctions after a long period of
          time. Usually anywhere between 10 to 25+ years.
      
          - Some VHS tapes like newer mini DVs and Digital 8 tapes can suffer from digital
          corruption, meaning that the footage becomes lost and cannot be recovered. These
          tapes were the steppingstone from VHS to the digital age when capturing footage
          straight to digital became the norm. Unfortunately,they are susceptible to digital
          corruption, which causes video pixilation and/or loss of audio.'
      - role: assistant
        content: Alright!
      - role: user
        content: 'Now I''m going to write my question, and if the info above is useful, you can use them in your response.
      
          Ready?'
      - role: assistant
        content: Ready for your question!
      - role: user
        content: Why do VHS tapes deteriorate over time?
inference:
  parameters:
    max_new_tokens: 250
    penalty_alpha: 0.5
    top_k: 4
    repetition_penalty: 1.105
model-index:
- name: Smol-Llama-101M-Chat-v1
  results:
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: AI2 Reasoning Challenge (25-Shot)
      type: ai2_arc
      config: ARC-Challenge
      split: test
      args:
        num_few_shot: 25
    metrics:
    - type: acc_norm
      value: 22.87
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Smol-Llama-101M-Chat-v1
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: HellaSwag (10-Shot)
      type: hellaswag
      split: validation
      args:
        num_few_shot: 10
    metrics:
    - type: acc_norm
      value: 28.69
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Smol-Llama-101M-Chat-v1
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MMLU (5-Shot)
      type: cais/mmlu
      config: all
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 24.93
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Smol-Llama-101M-Chat-v1
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: TruthfulQA (0-shot)
      type: truthful_qa
      config: multiple_choice
      split: validation
      args:
        num_few_shot: 0
    metrics:
    - type: mc2
      value: 45.76
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Smol-Llama-101M-Chat-v1
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: Winogrande (5-shot)
      type: winogrande
      config: winogrande_xl
      split: validation
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 50.04
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Smol-Llama-101M-Chat-v1
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: GSM8k (5-shot)
      type: gsm8k
      config: main
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 0.08
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Felladrin/Smol-Llama-101M-Chat-v1
      name: Open LLM Leaderboard
---

# A Llama Chat Model of 101M Parameters

- Base model: [BEE-spoke-data/smol_llama-101M-GQA](https://huggingface.co/BEE-spoke-data/smol_llama-101M-GQA)
- Datasets:
  - [Open-Orca/SlimOrca-Dedup](https://huggingface.co/datasets/Open-Orca/SlimOrca-Dedup)
  - [VMware/open-instruct](https://huggingface.co/datasets/VMware/open-instruct)
  - [LDJnr/Capybara](https://huggingface.co/datasets/LDJnr/Capybara)
  - [cognitivecomputations/ultrachat-uncensored](https://huggingface.co/datasets/cognitivecomputations/ultrachat-uncensored)
  - [starfishmedical/webGPT_x_dolly](https://huggingface.co/datasets/starfishmedical/webGPT_x_dolly)
  - [THUDM/webglm-qa](https://huggingface.co/datasets/THUDM/webglm-qa)
- Availability in other ML formats:
  - GGUF: [Felladrin/gguf-Smol-Llama-101M-Chat-v1](https://huggingface.co/Felladrin/gguf-Smol-Llama-101M-Chat-v1)
  - ONNX: [Felladrin/onnx-Smol-Llama-101M-Chat-v1](https://huggingface.co/Felladrin/onnx-Smol-Llama-101M-Chat-v1)
  - MLC: [Felladrin/mlc-q4f16-Smol-Llama-101M-Chat-v1](https://huggingface.co/Felladrin/mlc-q4f16-Smol-Llama-101M-Chat-v1)

## Recommended Prompt Format

```
<|im_start|>system
{system_message}<|im_end|>
<|im_start|>user
{user_message}<|im_end|>
<|im_start|>assistant
```

## Recommended Inference Parameters

```yml
penalty_alpha: 0.5
top_k: 4
repetition_penalty: 1.105
```

## [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_Felladrin__Smol-Llama-101M-Chat-v1)

|             Metric              |Value|
|---------------------------------|----:|
|Avg.                             |28.73|
|AI2 Reasoning Challenge (25-Shot)|22.87|
|HellaSwag (10-Shot)              |28.69|
|MMLU (5-Shot)                    |24.93|
|TruthfulQA (0-shot)              |45.76|
|Winogrande (5-shot)              |50.04|
|GSM8k (5-shot)                   | 0.08|