Translation
Transformers
Safetensors
llama
text-generation
multilingual
lora
astheraltranslate
text-generation-inference
Instructions to use SEE7/AstheralTranslate with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use SEE7/AstheralTranslate with Transformers:
# Use a pipeline as a high-level helper # Warning: Pipeline type "translation" is no longer supported in transformers v5. # You must load the model directly (see below) or downgrade to v4.x with: # pip install "transformers<5.0.0" from transformers import pipeline pipe = pipeline("translation", model="SEE7/AstheralTranslate")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("SEE7/AstheralTranslate") model = AutoModelForCausalLM.from_pretrained("SEE7/AstheralTranslate", device_map="auto") - Notebooks
- Google Colab
- Kaggle
|
Download README.md from SEE7/AstheralTranslate: direct link, hf CLI and curl.
- Browser
- Download file 1.29 kB
-
https://huggingface.co/SEE7/AstheralTranslate/resolve/main/README.md
- Command line
-
hf download hf://SEE7/AstheralTranslate/README.md
-
curl -L -o README.md https://huggingface.co/SEE7/AstheralTranslate/resolve/main/README.md
1.29 kB
metadata
language:
- en
- de
- fr
- es
- it
- pt
license: apache-2.0
library_name: transformers
pipeline_tag: translation
base_model: utter-project/EuroLLM-1.7B
datasets:
- Helsinki-NLP/opus-100
tags:
- translation
- multilingual
- lora
- astheraltranslate
AstheralTranslate
English translation into German, French, Spanish, Italian and Portuguese. Fine-tuned by SEE7 from EuroLLM-1.7B.
Model and training
1,656,850,432 parameters. Merged BF16 weights, rank-32 LoRA over attention and MLP projections, 28,704,768 trainable parameters and 500 optimizer steps.
Use
Prompt format: Translate from English to German. Return only the translation. Then provide the English text and the target-language label. Other supported target names are French, Spanish, Italian and Portuguese.
Reports and files
Base outputs · Fine-tuned outputs · Training manifest
Apache 2.0 base-model license. OPUS-100 source corpus terms and attribution apply to the training data.