uploaded readme
Browse files
README.md
ADDED
|
@@ -0,0 +1,109 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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 |
+
TinyLlama-ContextQuestionPair-Classifier-Reranker - bnb 4bits
|
| 11 |
+
- Model creator: https://huggingface.co/cnmoro/
|
| 12 |
+
- Original model: https://huggingface.co/cnmoro/TinyLlama-ContextQuestionPair-Classifier-Reranker/
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
Original model description:
|
| 18 |
+
---
|
| 19 |
+
license: cc-by-nc-2.0
|
| 20 |
+
language:
|
| 21 |
+
- en
|
| 22 |
+
- pt
|
| 23 |
+
tags:
|
| 24 |
+
- classification
|
| 25 |
+
- llama
|
| 26 |
+
- tinyllama
|
| 27 |
+
- rag
|
| 28 |
+
- rerank
|
| 29 |
+
---
|
| 30 |
+
```python
|
| 31 |
+
template = """<s><|system|>
|
| 32 |
+
You are a chatbot who always responds in JSON format indicating if the context contains relevant information to answer the question</s>
|
| 33 |
+
<|user|>
|
| 34 |
+
Context:
|
| 35 |
+
{Text}
|
| 36 |
+
|
| 37 |
+
Question:
|
| 38 |
+
{Prompt}</s>
|
| 39 |
+
<|assistant|>
|
| 40 |
+
"""
|
| 41 |
+
|
| 42 |
+
# Output should be:
|
| 43 |
+
|
| 44 |
+
{"relevant": true}
|
| 45 |
+
|
| 46 |
+
# or
|
| 47 |
+
|
| 48 |
+
{"relevant": false}
|
| 49 |
+
```
|
| 50 |
+
|
| 51 |
+
Example:
|
| 52 |
+
```text
|
| 53 |
+
<s><|system|>
|
| 54 |
+
You are a chatbot who always responds in JSON format indicating if the context contains relevant information to answer the question</s>
|
| 55 |
+
<|user|>
|
| 56 |
+
Context:
|
| 57 |
+
old. NFT were observed in almost all patients over 60 years of age, but the incidence was low.
|
| 58 |
+
Many ubiquitin-positive small-sized granules were observed in the second and third layer of the parahippocampal gyrus of aged patients,
|
| 59 |
+
and the incidence rose with increasing age. On the other hand, few of these granules were in patients with Alzheimer\'s type dementia.
|
| 60 |
+
Granulovacuolar degeneration was examined. Many centrally-located granules were positive for ubiquitin. Based on electron microscopic
|
| 61 |
+
observation of these granules at several stages, the granules were thought to be a type of autophagosome. During the first stage of
|
| 62 |
+
granulovacuolar degeneration, electron-dense materials appeared in the cytoplasm, following which they were surrounded by smooth cytoplasm,
|
| 63 |
+
following which they were surrounded by smooth endoplasmic reticulum. Analytical electron microscopy disclosed that the granules contained
|
| 64 |
+
some aluminium. Several senile changes in the central nervous system in cadavers were examined. The pattern of extension of Alzheimer\'s
|
| 65 |
+
neurofibrillary tangles (NFT) and senile plaques (SP) in the olfactory bulbs of 100 specimens was examined during routine autopsy by
|
| 66 |
+
immunohistochemical staining. NFT were first observed in the anterior olfactory nucleus after the age of 60, and incidence rose with
|
| 67 |
+
increasing age. Senile plaques were found in the nucleus when there were many SP in the cerebral cortex. Of 25 non-demented amyotrophic
|
| 68 |
+
lateral sclerosis patients, SP were found in the cerebral cortices of 10, and 9 of 10 were over 60 years old. NFT were observed in almost
|
| 69 |
+
all patients over
|
| 70 |
+
|
| 71 |
+
Question:
|
| 72 |
+
What is granulovacuolar degeneration and what was its observation on electron microscopy?</s>
|
| 73 |
+
<|assistant|>
|
| 74 |
+
{"relevant": true}</s>
|
| 75 |
+
```
|
| 76 |
+
|
| 77 |
+
vLLM recommended request parameters:
|
| 78 |
+
|
| 79 |
+
```python
|
| 80 |
+
prompt = "<s><|system|>\nYou are a chatbot who always responds in JSON format indicating if the context contains relevant information to answer the question</s>\n<|user|>\nContext:\nConhecida como missão de imagem de raios-x e espectroscopia (da sigla em inglês XRISM), a estratégia é utilizar o telescópio para ampliar os estudos da humanidade a níveis celestiais com uma fração dos pixels da tela de um Gameboy original, lançado em 1989. Isso é possível por meio de uma ferramenta chamada “Resolve”. Apesar de utilizar a medição em pixels, a tecnologia é bastante diferente de uma câmera. Com um conjunto de microcalorímetros de seis pixels quadrados que mede 0,5 cm², ela detecta a temperatura de cada raio-x que o atinge. Como funciona o Resolve do telescópio XRISM? Cientista do projeto XRISM da NASA, Brian Williams explicou em um comunicado o funcionamento do telescópio. “Chamamos o Resolve de espectrômetro de microcalorímetros porque cada um de seus 36 pixels está medindo pequenas quantidades de calor entregues por cada raio-x recebido, nos permitindo ver as impressões digitais químicas dos elementos que compõem as fontes com detalhes sem precedentes”.\n\nQuestion:\nQual é a sigla em alemão mencionada?</s>\n<|assistant|>\n{\"relevant\":"
|
| 81 |
+
|
| 82 |
+
headers = {
|
| 83 |
+
"Accept": "text/event-stream",
|
| 84 |
+
"Authorization": "Bearer EMPTY"
|
| 85 |
+
}
|
| 86 |
+
|
| 87 |
+
body = {
|
| 88 |
+
"model": model,
|
| 89 |
+
"prompt": [prompt],
|
| 90 |
+
"best_of": 5,
|
| 91 |
+
"max_tokens": 1,
|
| 92 |
+
"temperature": 0,
|
| 93 |
+
"top_p": 1,
|
| 94 |
+
"use_beam_search": True,
|
| 95 |
+
"top_k": -1,
|
| 96 |
+
"min_p": 0,
|
| 97 |
+
"repetition_penalty": 1,
|
| 98 |
+
"length_penalty": 1,
|
| 99 |
+
"min_tokens": 1,
|
| 100 |
+
"logprobs": 1
|
| 101 |
+
}
|
| 102 |
+
|
| 103 |
+
result = requests.post(base_uri, headers=headers, json=body)
|
| 104 |
+
result = result.json()
|
| 105 |
+
|
| 106 |
+
boolean_response = bool(eval(json_result['choices'][0]['text'].strip().title()))
|
| 107 |
+
print(boolean_response)
|
| 108 |
+
```
|
| 109 |
+
|