Skip to content

Latest commit

 

History

History
188 lines (172 loc) · 5.61 KB

README.md

File metadata and controls

188 lines (172 loc) · 5.61 KB
Raw
license tags model-index
apache-2.0
merge
mergekit
name results
NeuralPipe-7B-slerp
task dataset metrics source
type name
text-generation
Text Generation
name type config split args
AI2 Reasoning Challenge (25-Shot)
ai2_arc
ARC-Challenge
test
num_few_shot
25
type value name
acc_norm
67.75
normalized accuracy
task dataset metrics source
type name
text-generation
Text Generation
name type split args
HellaSwag (10-Shot)
hellaswag
validation
num_few_shot
10
type value name
acc_norm
86.15
normalized accuracy
task dataset metrics source
type name
text-generation
Text Generation
name type config split args
MMLU (5-Shot)
cais/mmlu
all
test
num_few_shot
5
type value name
acc
63.94
accuracy
task dataset metrics source
type name
text-generation
Text Generation
name type config split args
TruthfulQA (0-shot)
truthful_qa
multiple_choice
validation
num_few_shot
0
type value
mc2
59.8
task dataset metrics source
type name
text-generation
Text Generation
name type config split args
Winogrande (5-shot)
winogrande
winogrande_xl
validation
num_few_shot
5
type value name
acc
79.64
accuracy
task dataset metrics source
type name
text-generation
Text Generation
name type config split args
GSM8k (5-shot)
gsm8k
main
test
num_few_shot
5
type value name
acc
69.75
accuracy

NeuralPipe-7B

This model is a merge of the following models made with mergekit:

⚡ Quantized models

Thanks to TheBloke for the quantized models:

🧩 Configuration

slices:
  - sources:
      - model: OpenPipe/mistral-ft-optimized-1218
        layer_range: [0, 32]
      - model: mlabonne/NeuralHermes-2.5-Mistral-7B
        layer_range: [0, 32]
merge_method: slerp
base_model: OpenPipe/mistral-ft-optimized-1218
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "mlabonne/NeuralPipe-7B-slerp"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])

Output:

A large language model is an AI system that uses deep learning techniques to process and understand vast amounts of natural language data. It is designed to generate human-like text, perform complex language tasks, and understand the context, nuance, and meaning of textual data. These models are trained on large datasets, often including billions of words, to learn the patterns and relationships in language. As a result, they can generate coherent and contextually relevant text, answer questions, and perform a variety of other language-related tasks. Some well-known large language models include OpenAI's GPT-3, Google's BERT, and Facebook's RoBERTa.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 71.17
AI2 Reasoning Challenge (25-Shot) 67.75
HellaSwag (10-Shot) 86.15
MMLU (5-Shot) 63.94
TruthfulQA (0-shot) 59.80
Winogrande (5-shot) 79.64
GSM8k (5-shot) 69.75