metadata

license: mit
model-index:
  - name: ALMA-13B-R
    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: 55.55
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=haoranxu/ALMA-13B-R
          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: 79.45
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=haoranxu/ALMA-13B-R
          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: 49.52
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=haoranxu/ALMA-13B-R
          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: 36.09
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=haoranxu/ALMA-13B-R
          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: 75.3
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=haoranxu/ALMA-13B-R
          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: 0
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=haoranxu/ALMA-13B-R
          name: Open LLM Leaderboard

ALMA-R builds upon ALMA models, with further LoRA fine-tuning with our proposed Contrastive Preference Optimization (CPO) as opposed to the Supervised Fine-tuning used in ALMA. CPO fine-tuning requires our triplet preference data for preference learning. ALMA-R now can matches or even exceeds GPT-4 or WMT winners!

@misc{xu2024contrastive,
      title={Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translation}, 
      author={Haoran Xu and Amr Sharaf and Yunmo Chen and Weiting Tan and Lingfeng Shen and Benjamin Van Durme and Kenton Murray and Young Jin Kim},
      year={2024},
      eprint={2401.08417},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

@misc{xu2023paradigm,
      title={A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models}, 
      author={Haoran Xu and Young Jin Kim and Amr Sharaf and Hany Hassan Awadalla},
      year={2023},
      eprint={2309.11674},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

Download ALMA(-R) Models and Dataset 🚀

We release six translation models presented in the paper:

ALMA-7B
ALMA-7B-LoRA
ALMA-7B-R (NEW!): Further LoRA fine-tuning upon ALMA-7B-LoRA with contrastive preference optimization.
ALMA-13B
ALMA-13B-LoRA
ALMA-13B-R (NEW!): Further LoRA fine-tuning upon ALMA-13B-LoRA with contrastive preference optimization (BEST MODEL!).

Model checkpoints are released at huggingface:

Models	Base Model Link	LoRA Link
ALMA-7B	haoranxu/ALMA-7B	-
ALMA-7B-LoRA	haoranxu/ALMA-7B-Pretrain	haoranxu/ALMA-7B-Pretrain-LoRA
ALMA-7B-R (NEW!)	haoranxu/ALMA-7B-R (LoRA merged)	-
ALMA-13B	haoranxu/ALMA-13B	-
ALMA-13B-LoRA	haoranxu/ALMA-13B-Pretrain	haoranxu/ALMA-13B-Pretrain-LoRA
ALMA-13B-R (NEW!)	haoranxu/ALMA-13B-R (LoRA merged)	-

Note that ALMA-7B-Pretrain and ALMA-13B-Pretrain are NOT translation models. They only experience stage 1 monolingual fine-tuning (20B tokens for the 7B model and 12B tokens for the 13B model), and should be utilized in conjunction with their LoRA models.

Datasets used by ALMA and ALMA-R are also released at huggingface now (NEW!)

Datasets	Train / Validation	Test
Human-Written Parallel Data (ALMA)	train and validation	WMT'22
Triplet Preference Data	train	WMT'22 and WMT'23

A quick start to use our best system (ALMA-13B-R) for translation. An example of translating "我爱机器翻译。" into English:

import torch
from transformers import AutoModelForCausalLM
from transformers import AutoTokenizer

# Load base model and LoRA weights
model = AutoModelForCausalLM.from_pretrained("haoranxu/ALMA-13B-R", torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("haoranxu/ALMA-13B-R", padding_side='left')

# Add the source sentence into the prompt template
prompt="Translate this from Chinese to English:\nChinese: 我爱机器翻译。\nEnglish:"
input_ids = tokenizer(prompt, return_tensors="pt", padding=True, max_length=40, truncation=True).input_ids.cuda()

# Translation
with torch.no_grad():
    generated_ids = model.generate(input_ids=input_ids, num_beams=5, max_new_tokens=20, do_sample=True, temperature=0.6, top_p=0.9)
outputs = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
print(outputs)

Please find more details in our GitHub repository

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	49.32
AI2 Reasoning Challenge (25-Shot)	55.55
HellaSwag (10-Shot)	79.45
MMLU (5-Shot)	49.52
TruthfulQA (0-shot)	36.09
Winogrande (5-shot)	75.30
GSM8k (5-shot)	0.00