🦾 Mistralic-7B-1 🦾
Special thanks to Together Compute for sponsoring Skunkworks with compute!
**INFERENCE**
```
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
torch.set_default_device('cuda')
system_prompt = "Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.\n\n"
system_no_input_prompt = "Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n"
def generate_prompt(instruction, input=None):
if input:
prompt = f"### System:\n{system_prompt}\n\n"
else:
prompt = f"### System:\n{system_no_input_prompt}\n\n"
prompt += f"### Instruction:\n{instruction}\n\n"
if input:
prompt += f"### Input:\n{input}\n\n"
return prompt + """### Response:\n"""
device = "cuda"
model = AutoModelForCausalLM.from_pretrained("SkunkworksAI/Mistralic-7B-1")
tokenizer = AutoTokenizer.from_pretrained("SkunkworksAI/Mistralic-7B-1")
while True:
instruction = input("Enter Instruction: ")
instruction = generate_prompt(instruction)
inputs = tokenizer(instruction, return_tensors="pt", return_attention_mask=False)
outputs = model.generate(**inputs, max_length=1000, do_sample=True, temperature=0.01, use_cache=True, eos_token_id=tokenizer.eos_token_id)
text = tokenizer.batch_decode(outputs)[0]
print(text)
```
**EVALUATION**
Average: 0.72157
For comparison:
mistralai/Mistral-7B-v0.1 scores 0.7116
mistralai/Mistral-7B-Instruct-v0.1 scores 0.6794
# [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_SkunkworksAI__Mistralic-7B-1)
| Metric | Value |
|-----------------------|---------------------------|
| Avg. | 55.44 |
| ARC (25-shot) | 60.84 |
| HellaSwag (10-shot) | 82.29 |
| MMLU (5-shot) | 60.8 |
| TruthfulQA (0-shot) | 52.38 |
| Winogrande (5-shot) | 77.03 |
| GSM8K (5-shot) | 11.07 |
| DROP (3-shot) | 43.71 |