LL

Llama2 70B OASST SFT v10

LlamaGGUF by TheBlokeUpdated 2023-09-27
Download on Hugging Face
Parameters69B
Context4K
Layers80
Architecturellama
Smallest quant29.3 GB
Quick answer

Llama2 70B OASST SFT v10 needs about 44.6 GB of memory at Q4_K_M (a 41.4 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 10 quants, from 29.3 GB (Q2_K) to 48.8 GB (Q5_K_M).

Llama2 70B OASST SFT v10 – fine‑tuned for instruction following

Llama2 70B OASST SFT v10 is a 70‑billion parameter causal decoder‑only transformer fine‑tuned on Open‑Assistant data. It supports English and limited multilingual capabilities. The model is available in GGUF format for llama.cpp and related tools.

Good at

  • Instruction following
  • Chat‑style interactions
  • Code assistance

Limits

  • Large VRAM requirement for full precision
  • No vision support
  • License: LLAMA 2 COMMUNITY

Best for

  • Chatbot back‑ends
  • Customer support agents
  • Developer assistance

Who it’s for. Engineers and researchers needing a large instruction‑tuned model for local deployment. The model is suited for users with sufficient GPU memory and who require English and limited multilingual responses.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K29.3 GB 31.1 GB
Too big Get file →
Q3_K_S29.9 GB 31.8 GB
Too big Get file →
Q3_K_M33.2 GB 35.0 GB
Too big Get file →
Q3_K_L36.1 GB 38.0 GB
Too big Get file →
Q4_038.9 GB 40.7 GB
Too big Get file →
Q4_K_S39.1 GB 40.9 GB
Too big Get file →
Q4_K_M41.4 GB 43.3 GB
Too big Get file →
Q5_047.5 GB 49.3 GB
Too big Get file →
Q5_K_S47.5 GB 49.3 GB
Too big Get file →
Q5_K_M48.8 GB 50.6 GB
Too big Get file →

For 8 GB VRAM, use Q4_K (≈32.5 GB memory at 8k context) or Q3_K_S (≈33.1 GB). For 12 GB VRAM, Q4_0 (≈42.1 GB) or Q4_K_M (≈44.6 GB) are suitable. Higher quants like Q5_K_M provide better quality but need more memory.

What GPU do you need for Llama2 70B OASST SFT v10?

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: no quant fits at an 8K context.
  • 16 GB VRAM: no quant fits at an 8K context.
  • 24 GB VRAM: no quant fits at an 8K context.
  • 32 GB VRAM: no quant fits at an 8K context.
  • 48 GB VRAM: the best quant that fits is Q4_K_M (41.4 GB file, about 44.6 GB in use).

Which Llama2 70B OASST SFT v10 quant should you choose?

Q4_K_M is the usual starting point for Llama2 70B OASST SFT v10: it balances file size and quality well. Here it is a 41.4 GB file.

If memory is tight, the smallest file is Q2_K at 29.3 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
Parameters69B
Layers80
Trained context4,096 tokens
KV cache (fp16)320 KB per token

Source

Quantized byTheBloke
FamilyLlama
Licensellama2
Downloads (HF)5,910
Last updated2023-09-27

Frequently asked questions about Llama2 70B OASST SFT v10

How much VRAM does Llama2 70B OASST SFT v10 need?

About 44.6 GB at Q4_K_M with an 8K context window. That is the 41.4 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 Llama2 70B OASST SFT v10 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 32.5 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Llama2 70B OASST SFT v10 GGUF?

It depends on the quant: Q2_K is 29.3 GB, Q3_K_S is 29.9 GB, Q3_K_M is 33.2 GB, Q3_K_L is 36.1 GB, Q4_0 is 38.9 GB, Q4_K_S is 39.1 GB and 4 more.

What context length does Llama2 70B OASST SFT v10 support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 2.7 GB at 8K and 1.3 GB at 4K.

Where can I download Llama2 70B OASST SFT v10 GGUF?

The files are on Hugging Face in the TheBloke/Llama2-70B-OASST-SFT-v10-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 Llama2 70B OASST SFT v10 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 Llama2 70B OASST SFT v10 on 8 GB VRAM?

Only with a very low‑precision quant like Q2_K or Q3_K_S, which still require over 30 GB of VRAM at 8k context. 8 GB is insufficient.

Does this model support image inputs?

No. The model is a text‑only causal transformer.

What license applies to this model?

It is released under the LLAMA 2 COMMUNITY LICENSE AGREEMENT.

Similar models

Browse all →

More from TheBloke

All by TheBloke →

More Llama models

All Llama →

Other large models

All large models →