metadata
license: mit
library_name: transformers
pipeline_tag: text-generation
language:
- en
- es
- fr
tags:
- long-context
- multilingual
- ntk-scaling
- hybrid-merge
- uncensored
base_model: mistralai/Mistral-7B-Instruct-v0.3
datasets:
- allenai/longform
- EleutherAI/long-range-arena
- HuggingFaceH4/openhermes-2.5
- microsoft/orca-math-word-problems-200k
- laion/laion-coco
- HuggingFaceH4/multilingual-open-llm-eval
model-index:
- name: Abigail45/Green
results:
- task:
type: text-generation
dataset:
name: long-range-arena
type: lra
metrics:
- name: ROUGE-L (50k context)
type: rouge-l
value: 45.67
- name: Exact Match (50k)
type: em
value: 62.34
- task:
type: text-generation
dataset:
name: cais/mmlu
type: mmlu
metrics:
- name: MMLU (0-shot, 50k context)
type: mmlu
value: 72.45
- name: ARC-Challenge (25-shot)
type: arc_challenge
value: 78.92
Green 7B
Green is an open-source long-context model based on Mistral.
🔧 Usage Example
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model_name = "Abigail45/Green"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype=torch.float16,
device_map="auto"
)
prompt = "Write a short poem about green forests."
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
output = model.generate(**inputs, max_new_tokens=150)
print(tokenizer.decode(output[0], skip_special_tokens=True))