LL

Colossal LLaMA 2 13b Base

LlamaGGUF by mradermacherUpdated 2024-09-03
Download on Hugging Face
Parameters13.4B
Context4K
Layers40
Architecturellama
Smallest quant3.1 GB
Quick answer

Colossal LLaMA 2 13b Base needs about 15.3 GB of memory at Q4_K_M (a 8.1 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 24 quants, from 3.1 GB (IQ1_S) to 11.0 GB (Q6_K).

Colossal LLaMA 2 13b Base – 13B parameter LLaMA‑2 model with 4096 token context

Colossal LLaMA‑2‑13B‑Base is a 13 billion‑parameter LLaMA‑2 model. It supports both Chinese and English and offers a 4096‑token context window. The model is released under the LLaMA‑2 license and Apache 2.0 with no commercial restrictions.

Good at

  • Large‑scale Chinese and English NLP tasks
  • Long‑context generation up to 4096 tokens
  • Domain‑specific fine‑tuning

Limits

  • Requires at least 10 GB VRAM for 8 k context
  • No vision or audio capabilities
  • No special hardware beyond GPU with sufficient VRAM

Best for

  • Chatbots needing long context
  • Research on multilingual LLMs
  • Fine‑tuning for domain‑specific applications

Who it’s for. Researchers, developers, and hobbyists who need a large, multilingual LLaMA‑2 model with a generous context window and no commercial licensing constraints. The model is suitable for building specialized fine‑tuned applications.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S3.1 GB 6.9 GB
Fits Get file →
IQ1_M3.3 GB 7.2 GB
Fits Get file →
IQ2_XXS3.7 GB 7.6 GB
Tight Get file →
IQ2_XS4.1 GB 7.9 GB
Tight Get file →
IQ2_S4.4 GB 8.3 GB
Too big Get file →
IQ2_M4.7 GB 8.6 GB
Too big Get file →
Q2_K5.1 GB 8.9 GB
Too big Get file →
IQ3_XXS5.2 GB 9.0 GB
Too big Get file →
IQ3_XS5.6 GB 9.5 GB
Too big Get file →
IQ3_S5.9 GB 9.8 GB
Too big Get file →
Q3_K_S5.9 GB 9.8 GB
Too big Get file →
IQ3_M6.2 GB 10.1 GB
Too big Get file →
Q3_K_M6.6 GB 10.4 GB
Too big Get file →
Q3_K_L7.2 GB 11.0 GB
Too big Get file →
IQ4_XS7.2 GB 11.1 GB
Too big Get file →
Q4_07.6 GB 11.5 GB
Too big Get file →
Q4_07.6 GB 11.5 GB
Too big Get file →
Q4_07.6 GB 11.5 GB
Too big Get file →
Q4_07.7 GB 11.5 GB
Too big Get file →
Q4_K_S7.7 GB 11.5 GB
Too big Get file →
Q4_K_M8.1 GB 12.0 GB
Too big Get file →
Q5_K_S9.3 GB 13.1 GB
Too big Get file →
Q5_K_M9.5 GB 13.4 GB
Too big Get file →
Q6_K11.0 GB 14.8 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (3.09 GB) or IQ1_M (3.33 GB). For 12 GB VRAM, IQ2_S (4.41 GB) or Q2_K (5.07 GB) are good choices. For 16 GB VRAM, Q4_K_M (8.13 GB) or Q5_K_S (9.26 GB) provide higher quality while staying within memory limits. Estimates are approximate.

What GPU do you need for Colossal LLaMA 2 13b Base?

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: no quant fits at an 8K context.
  • 12 GB VRAM: the best quant that fits is IQ2_M (4.7 GB file, about 11.9 GB in use).
  • 16 GB VRAM: the best quant that fits is Q4_K_M (8.1 GB file, about 15.3 GB in use).
  • 24 GB VRAM: the best quant that fits is Q6_K (11.0 GB file, about 18.2 GB in use).

Which Colossal LLaMA 2 13b Base quant should you choose?

Q4_K_M is the usual starting point for Colossal LLaMA 2 13b Base: it balances file size and quality well. Here it is a 8.1 GB file.

If memory is tight, the smallest file is IQ1_S at 3.1 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

Architecturellama
Parameters13.4B
Layers40
Trained context4,096 tokens
KV cache (fp16)800 KB per token

Source

Quantized bymradermacher
FamilyLlama
Licensellama2
Downloads (HF)3,920
Last updated2024-09-03

Frequently asked questions about Colossal LLaMA 2 13b Base

How much VRAM does Colossal LLaMA 2 13b Base need?

About 15.3 GB at Q4_K_M with an 8K context window. That is the 8.1 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 Colossal LLaMA 2 13b Base on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 10.3 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Colossal LLaMA 2 13b Base GGUF?

It depends on the quant: IQ1_S is 3.1 GB, IQ1_M is 3.3 GB, IQ2_XXS is 3.7 GB, IQ2_XS is 4.1 GB, IQ2_S is 4.4 GB, IQ2_M is 4.7 GB and 18 more.

What context length does Colossal LLaMA 2 13b Base support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 6.7 GB at 8K and 3.4 GB at 4K.

Where can I download Colossal LLaMA 2 13b Base GGUF?

The files are on Hugging Face in the mradermacher/Colossal-LLaMA-2-13b-base-i1-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Colossal LLaMA 2 13b Base use?

Hugging Face lists the license as llama2. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run Colossal LLaMA 2 13b Base on 8 GB VRAM?

Yes, with a low‑bit quant such as IQ1_S or IQ1_M, which use about 3 GB of storage and require roughly 10 GB of VRAM at 8 k context.

Does this model support vision or audio input?

No, the model only handles text. It has no vision or audio capabilities.

What license applies to this model?

The model is licensed under LLaMA‑2 and Apache 2.0, with no additional commercial use restrictions.

Other GGUFs of Colossal-LLaMA-2-13b-base

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Llama models

All Llama →

Other medium models

All medium models →