DE

DeepSeek-V2-Lite

DeepSeekGGUF by legraphistaUpdated 2024-05-26
Download on Hugging Face
Parameters15.7B
Context160K
Layers27
Architecturedeepseek2
Smallest quant5.0 GB
Quick answer

DeepSeek-V2-Lite needs about 10.3 GB of memory at IQ3_M (a 7.6 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 25 quants, from 5.0 GB (IQ1_S) to 31.4 GB. Longer context needs more: at 32K tokens the KV cache alone adds about 9.1 GB.

DeepSeek-V2-Lite – 17B parameter Mixture‑of‑Experts model

DeepSeek‑V2‑Lite is a 17B parameter language model that uses Multi‑head Latent Attention and DeepSeekMoE. It is trained on 5.7T tokens and supports a context length of 163840 tokens. The model is available in multiple GGUF quantizations for llama.cpp.

Good at

  • Efficient inference with MLA
  • Sparse MoE architecture
  • Large context support

Limits

  • No vision capability
  • License not specified
  • Requires GPU for inference

Best for

  • Chat applications on a single 40G GPU
  • Fine‑tuning on 8x80G GPUs
  • Research on MoE models

Who it’s for. Researchers and developers who need a large‑scale MoE model that can run on a single high‑VRAM GPU. The model is suitable for those working with llama.cpp and interested in efficient inference.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S5.0 GB 7.8 GB
Tight Get file →
IQ1_M5.2 GB 8.0 GB
Too big Get file →
IQ2_XXS5.6 GB 8.4 GB
Too big Get file →
IQ2_XS6.0 GB 8.7 GB
Too big Get file →
IQ2_S6.0 GB 8.8 GB
Too big Get file →
IQ2_M6.3 GB 9.1 GB
Too big Get file →
Q2_K6.4 GB 9.2 GB
Too big Get file →
Q2_K_S6.5 GB 9.2 GB
Too big Get file →
IQ3_XXS7.0 GB 9.7 GB
Too big Get file →
IQ3_XS7.1 GB 9.9 GB
Too big Get file →
IQ3_S7.5 GB 10.3 GB
Too big Get file →
Q3_K_S7.5 GB 10.3 GB
Too big Get file →
IQ3_M7.6 GB 10.3 GB
Too big Get file →
Q3_K8.1 GB 10.9 GB
Too big Get file →
Q3_K_L8.5 GB 11.2 GB
Too big Get file →
IQ4_XS8.6 GB 11.3 GB
Too big Get file →
IQ4_NL8.9 GB 11.7 GB
Too big Get file →
Q4_K_S9.5 GB 12.3 GB
Too big Get file →
Q4_K10.4 GB 13.1 GB
Too big Get file →
Q5_K_S11.1 GB 13.9 GB
Too big Get file →
Q5_K11.9 GB 14.6 GB
Too big Get file →
Q6_K14.1 GB 16.8 GB
Too big Get file →
Q8_016.7 GB 19.5 GB
Too big Get file →
BF1631.4 GB 34.2 GB
Too big Get file →
—31.4 GB 34.2 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (4.99 GB, 7.8 GB memory) or IQ1_M (5.24 GB, 8 GB memory). For 16 GB VRAM, IQ2_S (6.01 GB, 8.8 GB memory) or Q2_K (6.43 GB, 9.2 GB memory) are good choices. Higher VRAM allows larger quants such as Q4_K (10.36 GB, 13.1 GB memory).

What GPU do you need for DeepSeek-V2-Lite?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: no quant fits at an 8K context.
  • 6 GB VRAM: no quant fits at an 8K context.
  • 8 GB VRAM: the best quant that fits is IQ1_M (5.2 GB file, about 8.0 GB in use).
  • 12 GB VRAM: the best quant that fits is IQ4_NL (8.9 GB file, about 11.7 GB in use).
  • 16 GB VRAM: the best quant that fits is Q5_K (11.9 GB file, about 14.6 GB in use).
  • 24 GB VRAM: the best quant that fits is Q8_0 (16.7 GB file, about 19.5 GB in use).
  • 32 GB VRAM: the best quant that fits is Q8_0 (16.7 GB file, about 19.5 GB in use).
  • 48 GB VRAM: the best quant that fits is BF16 (31.4 GB file, about 34.2 GB in use).

Which DeepSeek-V2-Lite quant should you choose?

IQ3_M (7.6 GB) is the middle option in this repository. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.

If memory is tight, the smallest file is IQ1_S at 5.0 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturedeepseek2
Parameters15.7B (approximate, mixture of experts)
Layers27
Trained context163,840 tokens
KV cache (fp16)270 KB per token

Source

Quantized bylegraphista
FamilyDeepSeek
Downloads (HF)13,748
Last updated2024-05-26

Frequently asked questions about DeepSeek-V2-Lite

How much VRAM does DeepSeek-V2-Lite need?

About 10.3 GB at IQ3_M with an 8K context window. That is the 7.6 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run DeepSeek-V2-Lite on an 8 GB GPU?

Yes, with IQ1_M: the 5.2 GB file needs about 8.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of DeepSeek-V2-Lite GGUF?

It depends on the quant: IQ1_S is 5.0 GB, IQ1_M is 5.2 GB, IQ2_XXS is 5.6 GB, IQ2_XS is 6.0 GB, IQ2_S is 6.0 GB, IQ2_M is 6.3 GB and 19 more.

What context length does DeepSeek-V2-Lite support?

The GGUF header lists a trained context length of 163,840 tokens (160K). The KV cache grows with the context you use: about 2.3 GB at 8K and 9.1 GB at 32K.

Where can I download DeepSeek-V2-Lite GGUF?

The files are on Hugging Face in the legraphista/DeepSeek-V2-Lite-IMat-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

Can I run DeepSeek-V2-Lite on 8 GB VRAM?

Yes, use IQ1_S or IQ1_M quantizations. They fit within 8 GB VRAM and use about 8 GB of memory at 8k context.

Does DeepSeek-V2-Lite support vision tasks?

No, the model has no vision capability.

What is the context length of DeepSeek-V2-Lite?

The model supports a context length of 163840 tokens.

Other GGUFs of DeepSeek-V2-Lite

Similar models

Browse all →

More from legraphista

All by legraphista →

More DeepSeek models

All DeepSeek →

Other medium models

All medium models →