Pharia-1-LLM-7B-control-aligned
Aleph Alpha · Provedores EU
Última revisão:
Soberania e alojamento
Este modelo está catalogado para referência. Ainda não está alojado na LLM EU, e nenhum endpoint próprio o serve aqui. Os pedidos são encaminhados diretamente para o fornecedor, sob os termos desse fornecedor.
Resumo
Pharia-1-LLM-7B-control-aligned é o modelo de controlo 7B da Aleph Alpha, publicado no Hugging Face sob a Open Aleph License, que limita o uso a fins educacionais e de investigação. A Aleph Alpha afirma que os modelos foram treinados em dados curados em conformidade com as regras de direitos de autor e proteção de dados da UE e nacionais, e são otimizados para alemão, francês e espanhol. O uso comercial é feito através de um acordo on-premise que inclui os pesos, e não através do download público.
Especificações
- Fornecedor
- Aleph Alpha (DE)
- Janela de contexto
- 8.2k
- Parâmetros
- 7.04B
- Modalidades
- text
- Licença
- other · licence
- Pesos abertos
- yes
- Data de lançamento
- 27/08/2024
- Tarefas
- Alemão, Dados sensíveis, Sumarização, Recuperação (RAG)
Idiomas
Forte: de, fr, es
Adequado: en, it, pt, nl, pl
Forte
- de
- fr
- es
Adequado
- en
- it
- pt
- nl
- pl
Não verificado: o fornecedor não documentou esta afirmação. — a language is listed as strong only where a source claims it; the card links the evidence where one exists.
Quando usar
Pontos fortes
- German, French and Spanish optimisation with EU-focused data curation
- control variant designed for length-controlled, non-refusing responses
- weights downloadable for research and teaching
- vendor offers commercial on-premise licensing of the same checkpoints
Limitações
- Open Aleph License allows non-commercial research and educational use only
- 8192-token context window
- RLHF-style chat behaviour is limited compared with current instruct models
Perguntas frequentes
Under what licence is Pharia-1-LLM-7B-control-aligned released, and how is commercial use handled?
The Open Aleph License, recorded as non-commercial research and educational use only (https://huggingface.co/Aleph-Alpha/Pharia-1-LLM-7B-control-aligned-hf). The weights are downloadable for research and teaching, and commercial use runs through on-premise agreement including weights and runtime, per the vendor documentation (https://docs.aleph-alpha.com/phariaai-dev-guide/latest/pharia-llm/model-access.html). The control variant is designed for length-controlled, non-refusing responses.
How large is Pharia-1-LLM-7B-control-aligned?
The card records 7.04B (safetensors total 7,041,544,704) (source: https://huggingface.co/api/models/Aleph-Alpha/Pharia-1-LLM-7B-control-aligned-hf) with an 8192-token context window ("Sequence length (Context window)", source: https://huggingface.co/Aleph-Alpha/Pharia-1-LLM-7B-control). The same Open Aleph License as the other Pharia-1-LLM checkpoints.
Which languages is Pharia-1-LLM-7B-control-aligned optimised for?
German, French and Spanish are recorded as strong, with en, it, pt, nl, pl supported. Aleph Alpha states the models were trained on curated data in compliance with EU and national copyright and data-protection rules. Source: https://huggingface.co/Aleph-Alpha/Pharia-1-LLM-7B-control-aligned-hf.
Is Pharia-1-LLM-7B-control-aligned served on LLM EU?
It has no endpoint in the catalogue, so it is catalogue only — not hosted on LLM EU yet, with no region and no retention to report. A card whose status is "listed" is a catalogue entry, not a running deployment.
What benchmark is recorded for Pharia-1-LLM-7B-control-aligned, and who ran it?
One: 0.417 (OpenGPT-X evaluation, listed as Pharia-1-LLM-7B-C-A) — an OpenGPT-X evaluation, not an evaluation by LLM EU, which has run none. Source: https://huggingface.co/openGPT-X/Teuken-7B-instruct-v0.6.