|
--- |
|
license: apache-2.0 |
|
language: |
|
- en |
|
- de |
|
- es |
|
- fr |
|
- it |
|
- pt |
|
- pl |
|
- nl |
|
- tr |
|
- sv |
|
- cs |
|
- el |
|
- hu |
|
- ro |
|
- fi |
|
- uk |
|
- sl |
|
- sk |
|
- da |
|
- lt |
|
- lv |
|
- et |
|
- bg |
|
- no |
|
- ca |
|
- hr |
|
- ga |
|
- mt |
|
- gl |
|
- zh |
|
- ru |
|
- ko |
|
- ja |
|
- ar |
|
- hi |
|
--- |
|
# Model Card for EuroLLM-1.7B-Instruct |
|
|
|
|
|
This is the model card for the first instruction tuned model of the EuroLLM series: EuroLLM-1.7B-Instruct. You can also check the pre-trained version: [EuroLLM-1.7B](https://huggingface.co/utter-project/EuroLLM-1.7B). |
|
|
|
- **Developed by:** Unbabel, Instituto Superior Técnico, University of Edinburgh, CentraleSupélec University of Paris-Saclay. |
|
- **Funded by:** European Union. |
|
- **Model type:** A 1.7B parameter instruction tuned multilingual transfomer LLM. |
|
- **Language(s) (NLP):** Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Irish, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish, Swedish, Arabic, Catalan, Chinese, Galician, Hindi, Japanese, Korean, Norwegian, Russian, Turkish, and Ukrainian. |
|
- **License:** Apache License 2.0. |
|
|
|
## Model Details |
|
|
|
The EuroLLM project has the goal of creating a suite of LLMs capable of understanding and generating text in all European Union languages as well as some additional relevant languages. |
|
EuroLLM-1.7B is a 1.7B parameter model trained on 4 trillion tokens divided across the considered languages and several data sources: Web data, parallel data (en-xx and xx-en), and high-quality datasets. |
|
EuroLLM-1.7B-Instruct was further instruction tuned on EuroBlocks, a dataset predominantly focusing on machine translation. |
|
|
|
|
|
### Model Description |
|
|
|
EuroLLM uses a standard, dense Transformer architecture: |
|
- We use grouped query attention (GQA) with 8 key-value heads, since it has been shown to increase speed at inference time while maintaining downstream performance. |
|
- We perform pre-layer normalization, since it improves the training stability, and use the RMSNorm, which is faster. |
|
- We use the SwiGLU activation function, since it has been shown to lead to good results on downstream tasks. |
|
- We use rotary positional embeddings (RoPE) in every layer, since these have been shown to lead to good performances while allowing the extension of the context length. |
|
|
|
For pre-training, we use 256 Nvidia H100 GPUs of the Marenostrum 5 supercomputer, training the model with a constant batch size of 3,072 sequences, which corresponds to approximately 12 million tokens, using the Adam optimizer, and BF16 precision. |
|
Here is a summary of the model hyper-parameters: |
|
| | | |
|
|--------------------------------------|----------------------| |
|
| Sequence Length | 4,096 | |
|
| Number of Layers | 24 | |
|
| Embedding Size | 2,048 | |
|
| FFN Hidden Size | 5,632 | |
|
| Number of Heads | 16 | |
|
| Number of KV Heads (GQA) | 8 | |
|
| Activation Function | SwiGLU | |
|
| Position Encodings | RoPE (\Theta=10,000) | |
|
| Layer Norm | RMSNorm | |
|
| Tied Embeddings | No | |
|
| Embedding Parameters | 0.262B | |
|
| LM Head Parameters | 0.262B | |
|
| Non-embedding Parameters | 1.133B | |
|
| Total Parameters | 1.657B | |
|
|
|
## Run the model |
|
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
|
model_id = "utter-project/EuroLLM-1.7B-Instruct" |
|
tokenizer = AutoTokenizer.from_pretrained(model_id) |
|
model = AutoModelForCausalLM.from_pretrained(model_id) |
|
|
|
text = "English: My name is EuroLLM. Portuguese:" |
|
|
|
inputs = tokenizer(text, return_tensors="pt") |
|
outputs = model.generate(**inputs, max_new_tokens=20) |
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True)) |
|
|
|
|
|
## Results |
|
|
|
### Machine Translation |
|
|
|
We evaluate EuroLLM-1.7B-Instruct on several machine translation benchmarks: FLORES-200, WMT-23, and WMT-24 comparing it with [Gemma-2B](https://huggingface.co/google/gemma-2b) and [Gemma-7B](https://huggingface.co/google/gemma-7b) (also instruction tuned on EuroBlocks-v0.1). |
|
The results show that EuroLLM-1.7B is substantially better than Gemma-2B in Machine Translation and competitive with Gemma-7B. |
|
|
|
#### Flores-200 |
|
| Model | AVG | AVG en-xx | AVG xx-en | en-ar | en-bg | en-ca | en-cs | en-da | en-de | en-el | en-es-latam | en-et | en-fi | en-fr | en-ga | en-gl | en-hi | en-hr | en-hu | en-it | en-ja | en-ko | en-lt | en-lv | en-mt | en-nl | en-no | en-pl | en-pt-br | en-ro | en-ru | en-sk | en-sl | en-sv | en-tr | en-uk | en-zh-cn | ar-en | bg-en | ca-en | cs-en | da-en | de-en | el-en | es-latam-en | et-en | fi-en | fr-en | ga-en | gl-en | hi-en | hr-en | hu-en | it-en | ja-en | ko-en | lt-en | lv-en | mt-en | nl-en | no-en | pl-en | pt-br-en | ro-en | ru-en | sk-en | sl-en | sv-en | tr-en | uk-en | zh-cn-en | |
|
|--------------------------------|------|-----------|-----------|-------|-------|-------|-------|-------|-------|-------|--------------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|----------|-------|-------|-------|-------|-------|-------|-------|----------|-------|-------|-------|-------|-------|-------|-------|--------------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|----------|-------|-------|-------|-------|-------|-------|-------|----------| |
|
| EuroLLM-1.7B-Instruct | 86.10| 85.53 | 86.67 | 83.87 | 88.36 | 84.42 | 88.34 | 88.77 | 86.63 | 86.71 | 85.99 | 86.98 | 87.13 | 87.21 | 72.25 | 85.97 | 74.78 | 82.96 | 85.51 | 87.77 | 89.26 | 86.27 | 86.31 | 86.22 | 67.38 | 86.95 | 88.68 | 87.38 | 89.13 | 88.39 | 87.47 | 87.51 | 85.32 | 89.20 | 86.24 | 86.33 | 86.17 | 85.80 | 87.20 | 87.53 | 87.53 | 89.26 | 88.71 | 86.49 | 86.55 | 87.60 | 88.17 | 88.90 | 79.89 | 87.59 | 87.53 | 86.10 | 86.34 | 87.54 | 86.25 | 86.08 | 85.03 | 85.60 | 78.16 | 86.80 | 89.96 | 85.24 | 88.85 | 88.42 | 85.86 | 87.17 | 86.36 | 89.48 | 86.76 | 86.06 | 85.88 | |
|
| Gemma-2B-EuroBlocks | 81.56| 78.93 | 84.18 | 75.25 | 82.46 | 83.17 | 82.17 | 84.40 | 83.20 | 79.63 | 84.15 | 72.63 | 81.00 | 85.12 | 38.79 | 82.00 | 67.00 | 81.18 | 78.24 | 84.80 | 87.08 | 82.04 | 73.02 | 68.41 | 56.67 | 83.30 | 86.69 | 83.07 | 86.82 | 84.00 | 84.55 | 77.93 | 76.19 | 80.77 | 79.76 | 84.19 | 84.10 | 83.67 | 85.73 | 86.89 | 86.38 | 88.39 | 88.11 | 84.68 | 86.11 | 83.45 | 86.45 | 88.22 | 50.88 | 86.44 | 85.87 | 85.33 | 85.16 | 86.75 | 85.62 | 85.00 | 81.55 | 81.45 | 67.90 | 85.95 | 89.05 | 84.18 | 88.27 | 87.38 | 85.13 | 85.22 | 83.86 | 87.83 | 84.96 | 85.15 | 85.10 | |
|
| Gemma-7B-EuroBlocks | 86.16| 85.49 | 86.82 | 83.39 | 88.32 | 85.82 | 88.88 | 89.01 | 86.96 | 86.62 | 86.31 | 84.42 | 88.11 | 87.46 | 61.85 | 86.10 | 77.91 | 87.01 | 85.81 | 87.57 | 89.88 | 87.24 | 84.47 | 83.15 | 67.13 | 86.50 | 90.44 | 87.57 | 89.22 | 89.13 | 88.58 | 86.73 | 84.68 | 88.16 | 86.87 | 88.40 | 87.11 | 86.65 | 87.25 | 88.17 | 87.47 | 89.59 | 88.44 | 86.76 | 86.66 | 87.55 | 88.88 | 88.86 | 73.46 | 87.63 | 88.43 | 87.12 | 87.31 | 87.49 | 87.20 | 87.15 | 85.16 | 85.96 | 78.39 | 86.73 | 90.52 | 85.38 | 89.17 | 88.75 | 86.35 | 86.82 | 86.21 | 89.39 | 88.20 | 86.45 | 86.28 | |
|
|
|
|
|
#### WMT-23 |
|
| Model | AVG | AVG en-xx | AVG xx-en | AVG xx-xx | en-de | en-cs | en-uk | en-ru | en-zh-cn | de-en | uk-en | ru-en | zh-cn-en | cs-uk | |
|
|--------------------------------|------|-----------|-----------|-----------|-------|-------|-------|-------|----------|-------|-------|-------|----------|-------| |
|
| EuroLLM-1.7B-Instruct | 82.56| 82.30 | 82.07 | 85.81 | 80.99 | 84.42 | 80.74 | 81.94 | 83.42 | 83.74 | 85.06 | 81.00 | 78.49 | 85.81 | |
|
| Gemma-2B-EuroBlocks | 79.86| 78.35 | 81.32 | 81.56 | 76.54 | 76.35 | 77.62 | 78.88 | 82.36 | 82.85 | 83.83 | 80.17 | 78.42 | 81.56 | |
|
| Gemma-7B-EuroBlocks | 83.90| 83.70 | 83.21 | 87.61 | 82.15 | 84.68 | 83.05 | 83.85 | 84.79 | 84.40 | 85.86 | 82.55 | 80.01 | 87.61 | |
|
|
|
|
|
#### WMT-24 |
|
| Model | AVG | AVG en-xx | AVG xx-xx | en-es-latam | en-cs | en-ru | en-uk | en-ja | en-zh-cn | en-hi | cs-uk | ja-zh-cn | |
|
|---------|------|------|-------|-------|-------|-------|--------|--------|-------|-------|-------|-----| |
|
| EuroLLM-1.7B-Instruct| 78.45|78.65|77.67|79.05|80.93|80.33|78.05|78.72|81.87|80.15|70.10|82.65|72.69| |
|
|Gemma-2B-EuroBlocks | 74.71|74.25|76.57|75.21|78.84|70.40|74.44|75.55|78.32|78.70|62.51|79.97|73.17| |
|
|Gemma-7B-EuroBlocks | 80.88|80.45|82.60|80.43|81.91|80.14|80.32|82.17|84.08|81.86|72.71|85.55|79.65| |
|
|
|
### General Benchmarks |
|
We also compare EuroLLM-1.7B with [TinyLlama-1.1-3T](https://huggingface.co/TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T) and [Gemma-2B](https://huggingface.co/google/gemma-2b) on 3 general benchmarks: Arc Challenge, Hellaswag, and MMLU. |
|
For the non-english languages we use the [Okapi](https://aclanthology.org/2023.emnlp-demo.28.pdf) datasets. |
|
Results show that EuroLLM-1.7B is superior to TinyLlama-1.1-3T and similar to Gemma-2B on Hellaswag but worse on Arc Challenge and MMLU. This can be due to the lower number of parameters of EuroLLM-1.7B (1.133B non-embedding parameters against 1.981B). |
|
|
|
#### Arc Challenge |
|
| Model | Average | English | German | Spanish | French | Italian | Portuguese | Chinese | Russian | Dutch | Arabic | Swedish | Hindi | Hungarian | Romanian | Ukrainian | Danish | Catalan | |
|
|--------------------|---------|---------|--------|---------|--------|---------|------------|---------|---------|-------|--------|---------|--------|-----------|----------|-----------|--------|---------| |
|
| EuroLLM-1.7B | 0.3130 | 0.4215 | 0.3148 | 0.3376 | 0.3259 | 0.3396 | 0.3410 | 0.3068 | 0.2626 | 0.3037| 0.2652 | 0.3279 | 0.2688 | 0.3039 | 0.3085 | 0.2943 | 0.2956 | 0.3027 | |
|
| TinyLlama-1.1-3T | 0.2621 | 0.3473 | 0.2541 | 0.2726 | 0.2797 | 0.2643 | 0.2829 | 0.2573 | 0.2421 | 0.2404| 0.2335 | 0.2661 | 0.2337 | 0.244 | 0.2536 | 0.2626 | 0.2476 | 0.2736 | |
|
| Gemma-2B | 0.3617 | 0.4846 | 0.3755 | 0.3940 | 0.4080 | 0.3687 | 0.3872 | 0.3726 | 0.3456 | 0.3328| 0.3122 | 0.3519 | 0.2851 | 0.3039 | 0.3590 | 0.3601 | 0.3565 | 0.3516 | |
|
|
|
#### Hellaswag |
|
| Model | Average | English | German | Spanish | French | Italian | Portuguese | Russian | Dutch | Arabic | Swedish | Hindi | Hungarian | Romanian | Ukrainian | Danish | Catalan | |
|
|--------------------|---------|---------|--------|---------|--------|---------|------------|---------|--------|--------|---------|--------|-----------|----------|-----------|--------|---------| |
|
| EuroLLM-1.7B | 0.4653 | 0.6199 | 0.4653 | 0.5187 | 0.5173 | 0.5024 | 0.5116 | 0.4582 | 0.4821 | 0.3939 | 0.4722 | 0.3505 | 0.3970 | 0.4441 | 0.4224 | 0.4556 | 0.4329 | |
|
| TinyLlama-1.1-3T | 0.3710 | 0.6027 | 0.3652 | 0.4136 | 0.4104 | 0.3780 | 0.4008 | 0.3544 | 0.3637 | 0.2981 | 0.3569 | 0.2904 | 0.3147 | 0.3337 | 0.3440 | 0.3464 | 0.3628 | |
|
| Gemma-2B | 0.4666 | 0.7165 | 0.4756 | 0.5414 | 0.5180 | 0.4841 | 0.5081 | 0.4664 | 0.4655 | 0.3868 | 0.4383 | 0.3413 | 0.3710 | 0.4316 | 0.4291 | 0.4471 | 0.4448 | |
|
|
|
#### MMLU |
|
| Model | Average | English | German | Spanish | French | Italian | Portuguese | Chinese | Russian | Dutch | Arabic | Swedish | Hindi | Hungarian | Romanian | Ukrainian | Danish | Catalan | |
|
|--------------------|---------|---------|--------|---------|--------|---------|------------|---------|---------|--------|--------|---------|--------|-----------|----------|-----------|--------|---------| |
|
| EuroLLM-1.7B | 0.2631 | 0.2553 | 0.2626 | 0.2653 | 0.2589 | 0.2628 | 0.2634 | 0.2546 | 0.2626 | 0.2677 | 0.2608 | 0.2656 | 0.2690 | 0.2551 | 0.2677 | 0.2655 | 0.2675 | 0.2689 | |
|
| TinyLlama-1.1-3T | 0.2546 | 0.2604 | 0.2498 | 0.2528 | 0.2535 | 0.2531 | 0.2511 | 0.2629 | 0.2541 | 0.2521 | 0.2591 | 0.2528 | 0.2550 | 0.2566 | 0.2548 | 0.2651 | 0.2419 | 0.2528 | |
|
| Gemma-2B | 0.3356 | 0.4168 | 0.3519 | 0.3475 | 0.3463 | 0.3433 | 0.3383 | 0.3345 | 0.3261 | 0.3429 | 0.3158 | 0.3318 | 0.2842 | 0.3185 | 0.3243 | 0.3152 | 0.3377 | 0.3307 | |
|
|
