Hugging Face's logo

ModernoTech
/
Nova-Coder

GGUF
Spanish
English
deepseek_v2
moderno-tech
nova
nova-coder
quantized
code-generation
local-ai
custom_code
conversational
Model card Files
xet
 Community
1
ModernoTech commited on
Commit
5dc541c
·
verified ·
1 Parent(s): b541c77

Nova/GGUF model card identifying Nova Coder variant, quantization table, legal attribution, and no unverified metrics

Browse files
Files changed (1) hide show
  1. README.md +67 -67
README.md CHANGED
@@ -1,88 +1,88 @@
1
  ---
2
- base_model: deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct
3
- inference: false
 
 
 
 
 
 
 
 
 
 
4
  license: other
5
- license_name: deepseek-license
6
- license_link: LICENSE
7
- model_creator: DeepSeek-ai
8
- model_name: DeepSeek-Coder-V2-Lite-Instruct
9
- model_type: deepseek
10
- quantized_by: Second State Inc.
11
  ---
12
 
13
- <!-- header start -->
14
- <!-- 200823 -->
15
- <div style="width: auto; margin-left: auto; margin-right: auto">
16
- <img src="https://github.com/LlamaEdge/LlamaEdge/raw/dev/assets/logo.svg" style="width: 100%; min-width: 400px; display: block; margin: auto;">
17
- </div>
18
- <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
19
- <!-- header end -->
20
-
21
- # DeepSeek-Coder-V2-Lite-Instruct-GGUF
22
 
23
- ## Original Model
24
-
25
- [deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct](https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct)
26
-
27
- ## Run with LlamaEdge
28
 
29
- - LlamaEdge version: [v0.12.3](https://github.com/LlamaEdge/LlamaEdge/releases/tag/0.12.3) and above
30
 
31
- - Prompt template
32
 
33
- - Prompt type: `deepseek-chat-2`
 
34
 
35
- - Prompt string
36
 
37
- ```text
38
- <|begin_of_sentence|>{system_message}
 
 
39
 
40
- User: {user_message_1}
41
 
42
- Assistant: {assistant_message_1}<|end_of_sentence|>User: {user_message_2}
43
 
44
- Assistant:
45
- ```
 
 
 
46
 
47
- - Context size: `128000`
48
 
49
- - Run as LlamaEdge service
 
 
 
 
 
 
 
 
 
 
 
 
 
 
50
 
51
- ```bash
52
- wasmedge --dir .:. \
53
- --nn-preload default:GGML:AUTO:DeepSeek-Coder-V2-Lite-Instruct-Q5_K_M.gguf \
54
- llama-api-server.wasm \
55
- --prompt-template deepseek-chat-2 \
56
- --ctx-size 128000 \
57
- --model-name DeepSeek-Coder-V2-Lite-Instruct
58
- ```
59
 
60
- - Run as LlamaEdge command app
 
 
 
 
 
 
 
 
61
 
62
- ```bash
63
- wasmedge --dir .:. \
64
- --nn-preload default:GGML:AUTO:DeepSeek-Coder-V2-Lite-Instruct-Q5_K_M.gguf \
65
- llama-chat.wasm \
66
- --prompt-template deepseek-chat-2 \
67
- --ctx-size 128000
68
- ```
69
 
70
- ## Quantized GGUF Models
 
 
 
 
71
 
72
- | Name | Quant method | Bits | Size | Use case |
73
- | ---- | ---- | ---- | ---- | ----- |
74
- | [DeepSeek-Coder-V2-Lite-Instruct-Q2_K.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q2_K.gguf) | Q2_K | 2 | 6.43 GB| smallest, significant quality loss - not recommended for most purposes |
75
- | [DeepSeek-Coder-V2-Lite-Instruct-Q3_K_L.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q3_K_L.gguf) | Q3_K_L | 3 | 8.46 GB| small, substantial quality loss |
76
- | [DeepSeek-Coder-V2-Lite-Instruct-Q3_K_M.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q3_K_M.gguf) | Q3_K_M | 3 | 8.13 GB| very small, high quality loss |
77
- | [DeepSeek-Coder-V2-Lite-Instruct-Q3_K_S.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q3_K_S.gguf) | Q3_K_S | 3 | 7.49 GB| very small, high quality loss |
78
- | [DeepSeek-Coder-V2-Lite-Instruct-Q4_0.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q4_0.gguf) | Q4_0 | 4 | 8.91 GB| legacy; small, very high quality loss - prefer using Q3_K_M |
79
- | [DeepSeek-Coder-V2-Lite-Instruct-Q4_K_M.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q4_K_M.gguf) | Q4_K_M | 4 | 10.4 GB| medium, balanced quality - recommended |
80
- | [DeepSeek-Coder-V2-Lite-Instruct-Q4_K_S.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q4_K_S.gguf) | Q4_K_S | 4 | 9.53 GB| small, greater quality loss |
81
- | [DeepSeek-Coder-V2-Lite-Instruct-Q5_0.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q5_0.gguf) | Q5_0 | 5 | 10.8 GB| legacy; medium, balanced quality - prefer using Q4_K_M |
82
- | [DeepSeek-Coder-V2-Lite-Instruct-Q5_K_M.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q5_K_M.gguf) | Q5_K_M | 5 | 11.9 GB| large, very low quality loss - recommended |
83
- | [DeepSeek-Coder-V2-Lite-Instruct-Q5_K_S.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q5_K_S.gguf) | Q5_K_S | 5 | 11.1 GB| large, low quality loss - recommended |
84
- | [DeepSeek-Coder-V2-Lite-Instruct-Q6_K.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q6_K.gguf) | Q6_K | 6 | 14.1 GB| very large, extremely low quality loss |
85
- | [DeepSeek-Coder-V2-Lite-Instruct-Q8_0.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-Q8_0.gguf) | Q8_0 | 8 | 16.7 GB| very large, extremely low quality loss - not recommended |
86
- | [DeepSeek-Coder-V2-Lite-Instruct-f16.gguf](https://huggingface.co/second-state/DeepSeek-Coder-V2-Lite-Instruct-GGUF/blob/main/DeepSeek-Coder-V2-Lite-Instruct-f16.gguf) | f16 | 16 | 31.4 GB| |
87
 
88
- *Quatized with llama.cpp b3212*
 
1
  ---
2
+ language:
3
+ - en
4
+ pretty_name: Nova Coder 2.0 — Moderno Tech
5
+ model_name: Nova Coder 2.0
6
+ model_creator: Moderno Tech
7
+ tags:
8
+ - moderno-tech
9
+ - nova
10
+ - nova-coder
11
+ - code-generation
12
+ - gguf
13
+ - moe
14
  license: other
15
+ inference: false
 
 
 
 
 
16
  ---
17
 
18
+ ![Nova Coder banner](images/nova-coder-banner.jpg)
 
 
 
 
 
 
 
 
19
 
20
+ <div align="center">
21
+ <img alt="Moderno Tech" src="images/moderno-tech-logo.png" width="84"/>
22
+ </div>
 
 
23
 
24
+ # Nova Coder 2.0 GGUF
25
 
26
+ **Moderno Tech — Nova**
27
 
28
+ Nova Coder 2.0 GGUF es la distribución de cuantización de Nova Coder 2.0 publicada por Moderno Tech.
29
+ La distribución original Safetensors BF16 se mantiene en `ModernoTech/Nova-Coder`; esta variante ofrece archivos GGUF para LM Studio, Ollama y llama.cpp.
30
 
31
+ ## Identificación
32
 
33
+ - Marca: Moderno Tech
34
+ - Familia: Nova
35
+ - Producto: Nova Coder 2.0
36
+ - Sitio oficial: https://moderno.com.ar
37
 
38
+ ## Variante recomendada
39
 
40
+ Para uso general con LM Studio, Ollama o llama.cpp:
41
 
42
+ - **`Nova-Coder-2.0-Q4_K_M.gguf`**
43
+ - Tamaño exacto: **10,364,416,480 bytes**
44
+ - SHA-256: `38bc76f3326b49b4d81d1027d092bf7ce5b4ed2de4136d1d2e7e6347c3ec8376`
45
+ - Cuantización: Q4_K_M
46
+ - Carácter: formato compacto con alta calidad para coding general.
47
 
48
+ ## Todas las variantes disponibles
49
 
50
+ | Archivo | Cuantización | Tamaño (bytes) | SHA-256 |
51
+ |---|---|---:|---|
52
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q2_K.gguf` | Q2_K | 6,430,464,480 | `d4eb2f3365ac103d4231352b331577135be67931cecedad474c38e4cecb9182d` |
53
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q3_K_L.gguf` | Q3_K_L | 8,459,398,624 | `9d9b1b46859bb2208acf2174192547e53389575fb2af820a988c91ecb94c2613` |
54
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q3_K_M.gguf` | Q3_K_M | 8,126,606,816 | `3918c980cc193aa63733db160fdd26ef3a668fecd5219d0e1aa62d12c86cc91f` |
55
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q3_K_S.gguf` | Q3_K_S | 7,487,663,584 | `15476c1c838f3a8205d0d7725a1f2b377da0ba1c287149680dbffc8d755904d2` |
56
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q4_0.gguf` | Q4_0 | 8,905,109,984 | `7d932d742f5e09b66cf017bb0e27bf5acf1d0a12d0e65bcc4561528020c4b56c` |
57
+ | `Nova-Coder-2.0-Q4_K_M.gguf` | Q4_K_M - recomendada | 10,364,416,480 | `38bc76f3326b49b4d81d1027d092bf7ce5b4ed2de4136d1d2e7e6347c3ec8376` |
58
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q4_K_S.gguf` | Q4_K_S | 9,533,608,416 | `48a9e203ee6f5ed208c4d8859b7c2c2914bd1af8621d90f366dd7b0bd9112d5c` |
59
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q5_0.gguf` | Q5_0 | 10,841,764,320 | `fbd09a611559c2f97767a41e555dd05d0c6a61425bf14d2d1e2cf0ea817690dd` |
60
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q5_K_M.gguf` | Q5_K_M | 11,851,313,632 | `08b4ff12b045ec0001a690249931d91acbcbe007694352d6ac5bf86e16f6fc6b` |
61
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q5_K_S.gguf` | Q5_K_S | 11,143,057,888 | `03f2eacb23d597e7793a57bcaabcc3ebceec1515b04a0fe49996817ff169922d` |
62
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q6_K.gguf` | Q6_K | 14,066,972,128 | `0cfe5211943fffec745261aba45023b07d1a9004356a4efdfbf855bf47ca6782` |
63
+ | `DeepSeek-Coder-V2-Lite-Instruct-Q8_0.gguf` | Q8_0 | 16,702,517,728 | `373dcfc92e01372709b6164fc836f677a6280e25e9eac5c434c64223207bfc4f` |
64
+ | `DeepSeek-Coder-V2-Lite-Instruct-f16.gguf` | f16 | 31,424,033,248 | `2de57aa898372715c457b7c06362bb9826f3e47ba2a83679f162fcb7b40e763b` |
65
 
66
+ ## Arquitectura
 
 
 
 
 
 
 
67
 
68
+ - Arquitectura: `DeepseekV2ForCausalLM`
69
+ - Tipo: MoE
70
+ - Parámetros totales: ~16B
71
+ - Parámetros activos por token: ~2.4B
72
+ - Capas: 27
73
+ - Hidden size: 2048
74
+ - Cabezas de atención: 16
75
+ - Contexto soportado: 163840
76
+ - Vocabulario: 102400
77
 
78
+ ## Licencia
 
 
 
 
 
 
79
 
80
+ - La identidad comercial es Moderno Tech / Nova Coder 2.0.
81
+ - Los avisos legales y atribuciones de origen se conservan.
82
+ - Condiciones aplicables provistas por el origen: MIT / model license de DeepSeek.
83
+ - Original: `https://github.com/deepseek-ai/DeepSeek-V2/blob/main/LICENSE-CODE`
84
+ - Original: `https://github.com/deepseek-ai/DeepSeek-V2/blob/main/LICENSE-MODEL`
85
 
86
+ ## Animación
 
 
 
 
 
 
 
 
 
 
 
 
 
 
87
 
88
+ ![Animación de inferencia](images/nova-coder-inference-process.gif)