File size: 6,889 Bytes
c5c6450 c88bb26 0feafe0 c88bb26 0feafe0 c5c6450 c88bb26 9d34981 c88bb26 08b1404 c88bb26 08b1404 c88bb26 08b1404 c88bb26 08b1404 c88bb26 08b1404 c88bb26 08b1404 0feafe0 |
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 |
---
license: apache-2.0
library_name: transformers
tags:
- finetune
- gpt4
- synthetic data
- custom_code
- h2oai
datasets:
- Locutusque/Hercules-v3.0
model-index:
- name: Cypher-Mini-1.8B
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: 39.59
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/Cypher-Mini-1.8B
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: 67.45
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/Cypher-Mini-1.8B
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: 31.14
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/Cypher-Mini-1.8B
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: 40.44
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/Cypher-Mini-1.8B
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: 65.19
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/Cypher-Mini-1.8B
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: 14.48
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/Cypher-Mini-1.8B
name: Open LLM Leaderboard
---
- This is an experimental model, Finetuned [h2oai/h2o-danube-1.8b-chat](https://huggingface.co/h2oai/h2o-danube-1.8b-chat), on Hercules v3 & private dataset.
- The original idea was to use this 1.8B model, divide the dataset based on task specific capabilities, train models and transform them into a mixture of experts.
- Hyperparameters: adamw with eps of 1e-8, cosine decay w/ 20% warmup, lr=2e-5.
## Format:
```
<|system|></s><|prompt|></s><|answer|>
```
## Benchamrks:
WIP
## Example:
```
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer, StoppingCriteria
import torch
class MyStoppingCriteria(StoppingCriteria):
def __init__(self, target_sequence, prompt):
self.target_sequence = target_sequence
self.prompt=prompt
def __call__(self, input_ids, scores, **kwargs):
generated_text = tokenizer.decode(input_ids[0])
generated_text = generated_text.replace(self.prompt,'')
if self.target_sequence in generated_text:
return True
return False
def __len__(self):
return 1
def __iter__(self):
yield self
modelpath="aloobun/Cypher-Mini-1.8B"
model = AutoModelForCausalLM.from_pretrained(
modelpath,
torch_dtype=torch.bfloat16,
device_map="cuda",
trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(
modelpath,
trust_remote_code=True,
use_fast=False,
)
prompt = "<|prompt|>Reflect on a time when you encountered a logical fallacy in an argument. How did you identify it, and what was the consequence?</s><|answer|>"
encoded_input = tokenizer(prompt, return_tensors='pt')
input_ids=encoded_input['input_ids'].cuda()
streamer = TextStreamer(tokenizer=tokenizer, skip_prompt=True)
op = model.generate(
input_ids,
streamer=streamer,
pad_token_id=tokenizer.eos_token_id,
do_sample=True,
temperature=0.7,
top_p=0.8,
max_new_tokens=512,
stopping_criteria=MyStoppingCriteria("</s>", prompt)
)
```
## Output:
>I do not have personal experiences or emotions, but I can provide you with an example of a logical fallacy and its consequences:
>
>One common logical fallacy is the appeal to authority fallacy. This occurs when someone argues that a particular opinion or belief is true because of who holds it (i.e., "because the doctor said so"). However, this approach does not take into account other factors that may influence the validity of the claim. For instance, if a doctor says that eating a certain food will cure cancer, it does not necessarily mean that it will work for everyone. Other factors such as genetics, lifestyle, and environmental factors could also play a role in whether or not a person gets cancer.
>
>The consequence of using the appeal to authority fallacy is that it often leads to hasty conclusions and misinformation. It can be difficult to separate fact from fiction, especially when people rely on authority figures to make decisions. As a result, individuals may end up making poor choices based on incomplete information. This can lead to unintended consequences, such as harming oneself or others.
>
>To avoid falling prey to the appeal to authority fallacy, it is important to seek out multiple sources of information and consider all available evidence before making a decision. This can help individuals make more informed choices and reduce the likelihood of being swayed by unsubstantiated claims.</s>
# [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_aloobun__Cypher-Mini-1.8B)
| Metric |Value|
|---------------------------------|----:|
|Avg. |43.05|
|AI2 Reasoning Challenge (25-Shot)|39.59|
|HellaSwag (10-Shot) |67.45|
|MMLU (5-Shot) |31.14|
|TruthfulQA (0-shot) |40.44|
|Winogrande (5-shot) |65.19|
|GSM8k (5-shot) |14.48|
|