Yz

deepseek-coder-v2

An open-source Mixture-of-Experts code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks.

16b 236b
Hızlı Kurulum (Ollama kuruluysa) 
ollama run deepseek-coder-v2

Ollama kurulu değil mi? ollama.com/download — Windows, macOS ve Linux için ücretsiz. İlk çalıştırmada model indirilir, sonrası tamamen çevrimdışıdır.

Varyantlar

Boyut büyüdükçe kalite artar, donanım ihtiyacı yükselir. Başlangıç için küçük varyantı deneyin.

EtiketBoyutBağlamGirdiKomut
latest 8.9GB 160K Text ollama run deepseek-coder-v2:latest
lite 8.9GB 160K Text ollama run deepseek-coder-v2:lite
16b 8.9GB 160K Text ollama run deepseek-coder-v2:16b
236b 133GB 4K Text ollama run deepseek-coder-v2:236b
16b-lite-base-q2_K 6.4GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q2_K
16b-lite-base-q3_K_S 7.5GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q3_K_S
16b-lite-base-q3_K_M 8.1GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q3_K_M
16b-lite-base-q3_K_L 8.5GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q3_K_L
16b-lite-base-q4_0 8.9GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q4_0
16b-lite-base-q4_1 9.9GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q4_1
16b-lite-base-q4_K_S 9.5GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q4_K_S
16b-lite-base-q4_K_M 10GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q4_K_M
16b-lite-base-q5_0 11GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q5_0
16b-lite-base-q5_1 12GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q5_1
16b-lite-base-q5_K_S 11GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q5_K_S
16b-lite-base-q5_K_M 12GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q5_K_M
16b-lite-base-q6_K 14GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q6_K
16b-lite-base-q8_0 17GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-q8_0
16b-lite-base-fp16 31GB 4K Text ollama run deepseek-coder-v2:16b-lite-base-fp16
16b-lite-instruct-q2_K 6.4GB 4K Text ollama run deepseek-coder-v2:16b-lite-instruct-q2_K
16b-lite-instruct-q3_K_S 7.5GB 4K Text ollama run deepseek-coder-v2:16b-lite-instruct-q3_K_S
16b-lite-instruct-q3_K_M 8.1GB 4K Text ollama run deepseek-coder-v2:16b-lite-instruct-q3_K_M
16b-lite-instruct-q3_K_L 8.5GB 4K Text ollama run deepseek-coder-v2:16b-lite-instruct-q3_K_L
16b-lite-instruct-q4_0 8.9GB 160K Text ollama run deepseek-coder-v2:16b-lite-instruct-q4_0
16b-lite-instruct-q4_1 9.9GB 4K Text ollama run deepseek-coder-v2:16b-lite-instruct-q4_1

39 varyant daha var — tümü için ollama.com.

Model Detayları ve Benchmarklar (kaynak: ollama.com)

DeepSeek-Coder-V2 is an open-source Mixture-of-Experts (MoE) code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks. DeepSeek-Coder-V2 is further pre-trained from DeepSeek-Coder-V2-Base with 6 trillion tokens sourced from a high-quality and multi-source corpus.

References

Hugging Face