← Run Locally · Onke amamodeli
ngu-meta-llama
Llama 3.3 70B Instruct
meta-llama/Llama-3.3-70B-Instruct
Onke amanani ngamaqikelelo. Ukusetyenziswa kwememori kunye nesantya kuxhomekeke kwiidrayivar zenu, kwiisetingi nakwezinye iinkqubo ezisebenzayo.
- Iiparamitha
- 71B
- Ilayisenisi
- llama3.3
- Ipapashiwe
- Nov 26, 2024
Ilayisenisi yeyomenzi wemodeli. Khangelani imiqathango yayo ngaphambi kokuba nisebenzise imodeli.
I-Hugging Face iyanicela ukuba namkele imiqathango yomenzi ngaphambi kokuba nikhuphele le modeli. Akukho kiti ethengiswayo yayo.
Eyokuqala kwi-Hugging Face ↗Imemori efunekayo ngokwe-quantization
Kwimixholo ye-8,192 tokens, kubandakanywa i-0.8 GB yendawo yokusebenzela.
| Quantization | Ifayile yemodeli | Isiphelo |
|---|---|---|
| Q4_K_M | 42.8 GB | 47.9 GB(ngokubanzi) |
| Q5_K_M | 50.2 GB | 56.0 GB(ngokubanzi) |
| Q6_K | 57.9 GB | 64.4 GB(ngokubanzi) |
| Q8_0 | 75.0 GB | 83.3 GB(ngokubanzi) |
| F16 | 141.1 GB | 156.0 GB(ngokubanzi) |
Izixhobo eziyisebenzisa kwikhadi lemizobo
Kwi-Q4_K_M nakwimixholo ye-8,192 tokens.
Akukho nanye kwizixhobo esizidwelisayo eyisebenzisa ngokupheleleyo kwikhadi lemizobo kule setingi.
Yijongeni kwikhompyuter yenuImodeli enye leyo ngeAPI
Umthengisi owona uxabiso luphantsi: NanoGPT, $0.99 ngenyanga ekusetyenzisweni okuqhelekileyo (izicelo ezingama-200 ngosuku).
Jonga amaxabiso kwi-Model Price MapIinkcukacha zamamodeli zivela kuluhlu oluphangaleleyo lwe-Hugging Face (amagama, ubungakanani, iilayisenisi). Iphepha ngalinye lemodeli linxibelelanisa nelokuqala.