deepseek-coder#

  • Context Length: 16384

  • Model Name: deepseek-coder

  • Languages: en, zh

  • Abilities: generate

  • Description: Deepseek Coder is composed of a series of code language models, each trained from scratch on 2T tokens, with a composition of 87% code and 13% natural language in both English and Chinese.

Specifications#

Model Spec 1 (pytorch, 1_3 Billion)#

  • Model Format: pytorch

  • Model Size (in billions): 1_3

  • Quantizations: 4-bit, 8-bit, none

  • Engines: vLLM, Transformers (vLLM only available for quantization none)

  • Model ID: deepseek-ai/deepseek-coder-1.3b-base

  • Model Hubs: Hugging Face, ModelScope

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 1_3 --model-format pytorch --quantization ${quantization}

Model Spec 2 (pytorch, 6_7 Billion)#

  • Model Format: pytorch

  • Model Size (in billions): 6_7

  • Quantizations: 4-bit, 8-bit, none

  • Engines: vLLM, Transformers (vLLM only available for quantization none)

  • Model ID: deepseek-ai/deepseek-coder-6.7b-base

  • Model Hubs: Hugging Face, ModelScope

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 6_7 --model-format pytorch --quantization ${quantization}

Model Spec 3 (pytorch, 7 Billion)#

  • Model Format: pytorch

  • Model Size (in billions): 7

  • Quantizations: 4-bit, 8-bit, none

  • Engines: vLLM, Transformers (vLLM only available for quantization none)

  • Model ID: deepseek-ai/deepseek-coder-7b-base-v1.5

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 7 --model-format pytorch --quantization ${quantization}

Model Spec 4 (pytorch, 33 Billion)#

  • Model Format: pytorch

  • Model Size (in billions): 33

  • Quantizations: 4-bit, 8-bit, none

  • Engines: vLLM, Transformers (vLLM only available for quantization none)

  • Model ID: deepseek-ai/deepseek-coder-33b-base

  • Model Hubs: Hugging Face, ModelScope

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 33 --model-format pytorch --quantization ${quantization}

Model Spec 5 (ggufv2, 1_3 Billion)#

  • Model Format: ggufv2

  • Model Size (in billions): 1_3

  • Quantizations: Q2_K, Q3_K_L, Q3_K_M, Q3_K_S, Q4_0, Q4_K_M, Q4_K_S, Q5_0, Q5_K_M, Q5_K_S, Q6_K, Q8_0

  • Engines: llama.cpp

  • Model ID: TheBloke/deepseek-coder-1.3b-base-GGUF

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 1_3 --model-format ggufv2 --quantization ${quantization}

Model Spec 6 (ggufv2, 6_7 Billion)#

  • Model Format: ggufv2

  • Model Size (in billions): 6_7

  • Quantizations: Q2_K, Q3_K_L, Q3_K_M, Q3_K_S, Q4_0, Q4_K_M, Q4_K_S, Q5_0, Q5_K_M, Q5_K_S, Q6_K, Q8_0

  • Engines: llama.cpp

  • Model ID: TheBloke/deepseek-coder-6.7B-base-GGUF

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 6_7 --model-format ggufv2 --quantization ${quantization}

Model Spec 7 (ggufv2, 7 Billion)#

  • Model Format: ggufv2

  • Model Size (in billions): 7

  • Quantizations: Q2_K, Q3_K_L, Q3_K_M, Q3_K_S, Q4_K_M, Q4_K_S, Q5_0, Q5_K_M, Q5_K_S, Q6_K, Q8_0

  • Engines: llama.cpp

  • Model ID: dagbs/deepseek-coder-7b-base-v1.5-GGUF

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 7 --model-format ggufv2 --quantization ${quantization}

Model Spec 8 (ggufv2, 33 Billion)#

  • Model Format: ggufv2

  • Model Size (in billions): 33

  • Quantizations: Q2_K, Q3_K_L, Q3_K_M, Q3_K_S, Q4_0, Q4_K_M, Q4_K_S, Q5_0, Q5_K_M, Q5_K_S, Q6_K, Q8_0

  • Engines: llama.cpp

  • Model ID: TheBloke/deepseek-coder-33B-base-GGUF

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 33 --model-format ggufv2 --quantization ${quantization}

Model Spec 9 (gptq, 1_3 Billion)#

  • Model Format: gptq

  • Model Size (in billions): 1_3

  • Quantizations: Int4

  • Engines: vLLM, Transformers

  • Model ID: TheBloke/deepseek-coder-1.3b-base-GPTQ

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 1_3 --model-format gptq --quantization ${quantization}

Model Spec 10 (gptq, 6_7 Billion)#

  • Model Format: gptq

  • Model Size (in billions): 6_7

  • Quantizations: Int4

  • Engines: vLLM, Transformers

  • Model ID: TheBloke/deepseek-coder-6.7B-base-GPTQ

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 6_7 --model-format gptq --quantization ${quantization}

Model Spec 11 (gptq, 33 Billion)#

  • Model Format: gptq

  • Model Size (in billions): 33

  • Quantizations: Int4

  • Engines: vLLM, Transformers

  • Model ID: TheBloke/deepseek-coder-33B-base-GPTQ

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 33 --model-format gptq --quantization ${quantization}

Model Spec 12 (awq, 1_3 Billion)#

  • Model Format: awq

  • Model Size (in billions): 1_3

  • Quantizations: Int4

  • Engines: vLLM, Transformers

  • Model ID: TheBloke/deepseek-coder-1.3b-base-AWQ

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 1_3 --model-format awq --quantization ${quantization}

Model Spec 13 (awq, 6_7 Billion)#

  • Model Format: awq

  • Model Size (in billions): 6_7

  • Quantizations: Int4

  • Engines: vLLM, Transformers

  • Model ID: TheBloke/deepseek-coder-6.7B-base-AWQ

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 6_7 --model-format awq --quantization ${quantization}

Model Spec 14 (awq, 33 Billion)#

  • Model Format: awq

  • Model Size (in billions): 33

  • Quantizations: Int4

  • Engines: vLLM, Transformers

  • Model ID: TheBloke/deepseek-coder-33B-base-AWQ

  • Model Hubs: Hugging Face

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name deepseek-coder --size-in-billions 33 --model-format awq --quantization ${quantization}