metadata
license: mit
model-index:
- name: Asimov-7B-v2
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: 54.27
name: normalized accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=prithivida/Asimov-7B-v2
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: 78.72
name: normalized accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=prithivida/Asimov-7B-v2
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: 52.59
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=prithivida/Asimov-7B-v2
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: 45.44
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=prithivida/Asimov-7B-v2
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: 71.82
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=prithivida/Asimov-7B-v2
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: 10.92
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=prithivida/Asimov-7B-v2
name: Open LLM Leaderboard
Model Card for Asimov-7B-v2
Asimov will be a series of language models that are trained to act as a useful writing assistant, named after one of the profilic writers of our time Isaac Asimov. Asimov-7B-v2 is the 2nd model in the series, and is a fine-tuned version of mistralai/Mistral-7B-v0.1 with a variety of publicly available analytical, adversarial and quantitative reasoning datasets. This model has not been aligned with any human preference datasets so consider this as a model with no guard rails.
Model Details
- Model type: A 7B parameter GPT-like model fine-tuned on publicly available synthetic datasets.
- Language(s) (NLP): Primarily English
- License: MIT
- Finetuned from model: mistralai/Mistral-7B-v0.1
Model Description
- Developed by: [More Information Needed]
- Funded by [optional]: [More Information Needed]
- Shared by [optional]: [More Information Needed]
- Model type: [More Information Needed]
- Language(s) (NLP): [More Information Needed]
- License: [More Information Needed]
- Finetuned from model [optional]: [More Information Needed]
Bias, Risks, and Limitations
Asimov-7B-v2 has not been aligned to human preferences, so the model can produce problematic outputs (especially when prompted to do so). It is also unknown what the size and composition of the corpus was used to train the base model (mistralai/Mistral-7B-v0.1), however it is likely to have included a mix of Web data and technical sources like books and code. See the Falcon 180B model card for an example of this.
How to Get Started with the Model
Use the code below to get started with the model.
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer,GenerationConfig
from peft import PeftModel, PeftConfig
model_name = "prithivida/Asimov-7B-v2"
peft_config = PeftConfig.from_pretrained(model_name)
base_model = AutoModelForCausalLM.from_pretrained(
peft_config.base_model_name_or_path,
return_dict=True,
device_map="auto",
torch_dtype=torch.float16,
low_cpu_mem_usage=True,
)
model = PeftModel.from_pretrained(
base_model,
model_name,
torch_dtype=torch.float16,
device_map="auto",
)
tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=True)
model.config.pad_token_id = tokenizer.unk_token_id
def run_inference(messages):
chat = []
for i, message in enumerate(messages):
if i % 2 ==0:
chat.append({"role": "Human", "content": f"{message}"})
else:
chat.append({"role": "Assistant", "content": f"{message}"})
prompt = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt")
input_ids = inputs["input_ids"].cuda()
generation_output = model.generate(
input_ids=input_ids,
generation_config=GenerationConfig(pad_token_id=tokenizer.pad_token_id,
do_sample=True,
temperature=1.0,
top_k=50,
top_p=0.95),
return_dict_in_generate=True,
output_scores=True,
max_new_tokens=128
)
for seq in generation_output.sequences:
output = tokenizer.decode(seq)
print(output.split("### Assistant: ")[1].strip())
run_inference(["What's the longest side of the right angled triangle called and how is it related to the Pythagoras theorem?"])
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 52.29 |
| AI2 Reasoning Challenge (25-Shot) | 54.27 |
| HellaSwag (10-Shot) | 78.72 |
| MMLU (5-Shot) | 52.59 |
| TruthfulQA (0-shot) | 45.44 |
| Winogrande (5-shot) | 71.82 |
| GSM8k (5-shot) | 10.92 |