Comparáid
Blip2 Opt 2.7B vs Blip2 Flan T5 Xl
Cuir dhá nó ceithre mhúnla i gcomparáid ar na réimsí a chinneann imscaradh: comhthéacs, praghas, réigiún, ceadúnas, teangacha, óstáil, agus cad atá fíoraithe.
| Blip2 Opt 2.7B | Blip2 Flan T5 Xl | |
|---|---|---|
| Soláthraí | Salesforce | Salesforce |
| Paraiméadair | 3.7B | 3.9B |
| Fuinneog comhthéacs | Gan fhíorú: níl an t-éileamh seo doiciméadaithe ag an soláthraí. | Gan fhíorú: níl an t-éileamh seo doiciméadaithe ag an soláthraí. |
| Ceadúnas | mit | mit |
| Módlachtaí | text, vision | text, vision |
| Dáta scaoilte | 06 Feabh 2023 | 06 Feabh 2023 |
| Teangacha | Láidir: en Sásúil: — | Láidir: en Sásúil: — |
| Meáchain oscailte | tá | tá |
| Tascanna | Fís, Costas íseal | Fís |
Nuair a bhíonn dhá mhúnla difriúil, taispeánann an tábla seo an difríocht agus stopann sé. Ní fhoilsíonn LLM EU suaitheantas buaiteora gan scór doiciméadaithe don tasc sin, agus fanann figiúr atá ar iarraidh ar iarraidh seachas a bheith ina nialas.
Cathain a úsáidtear é
Blip2 Opt 2.7B
Láidreachtaí
- open weights are published on Hugging Face and can be downloaded without a vendor agreement
- mit licence, so commercial self-hosting is permitted without a separate agreement
- at 3.7B parameters it fits on a single accelerator
- the model page lists en
Teorainneacha
- catalogue only — LLM EU does not run this model, and nothing on this card implies an LLM EU endpoint
- no evaluation has been run by LLM EU
- the published configuration records no context window, so the card carries none
- at 3.7B parameters, the weights fit on a single modest accelerator
- vision checkpoints are heavier to serve than their parameter count alone suggests, because the image encoder and pre-processing run alongside the language model
Tascanna
Blip2 Flan T5 Xl
Láidreachtaí
- open weights are published on Hugging Face and can be downloaded without a vendor agreement
- mit licence, so commercial self-hosting is permitted without a separate agreement
- at 3.9B parameters it fits on a single accelerator
- the model page lists en
Teorainneacha
- catalogue only — LLM EU does not run this model, and nothing on this card implies an LLM EU endpoint
- no evaluation has been run by LLM EU
- the published configuration records no context window, so the card carries none
- at 3.9B parameters, the weights fit on a single modest accelerator
- vision checkpoints are heavier to serve than their parameter count alone suggests, because the image encoder and pre-processing run alongside the language model
Tascanna
Tascanna: Fís