PH

saiga phi3 medium sft m1 d2 kto m5 d7

PhiGGUF by olkoviUpdated 2024-07-28
Download on Hugging Face
Parameters14B
Context4K
Layers40
Architecturellama
Smallest quant6.3 GB
Quick answer

saiga phi3 medium sft m1 d2 kto m5 d7 needs about 10.6 GB of memory at Q4_K_M (a 8.4 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 9 quants, from 6.3 GB (IQ3_M) to 27.9 GB.

saiga phi3 medium sft m1 d2 kto m5 d7 is a 14B-parameter Phi model in GGUF format, quantized by olkovi. Its trained context length is 4K tokens. The repository has 9 quant files, from 6.3 GB (IQ3_M) to 27.9 GB.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ3_M6.3 GB 7.6 GB
Tight Get file →
Q3_K_M6.8 GB 8.1 GB
Too big Get file →
IQ4_XS7.6 GB 8.9 GB
Too big Get file →
IQ4_NL8.0 GB 9.3 GB
Too big Get file →
Q4_K_M8.4 GB 9.7 GB
Too big Get file →
Q5_K_M9.9 GB 11.2 GB
Too big Get file →
Q6_K11.5 GB 12.8 GB
Too big Get file →
Q8_014.8 GB 16.2 GB
Too big Get file →
—27.9 GB 29.3 GB
Too big Get file →

What GPU do you need for saiga phi3 medium sft m1 d2 kto m5 d7?

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 Q4_K_M (8.4 GB file, about 10.6 GB in use).
  • 16 GB VRAM: the best quant that fits is Q6_K (11.5 GB file, about 13.6 GB in use).
  • 24 GB VRAM: the best quant that fits is Q8_0 (14.8 GB file, about 17.0 GB in use).
  • 32 GB VRAM: the best quant that fits is the main file (27.9 GB file, about 30.1 GB in use).

Which saiga phi3 medium sft m1 d2 kto m5 d7 quant should you choose?

Q4_K_M is the usual starting point for saiga phi3 medium sft m1 d2 kto m5 d7: it balances file size and quality well. Here it is a 8.4 GB file.

The Q8_0 file is 14.8 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is IQ3_M at 6.3 GB, with the quality trade-offs that come with fewer bits. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturellama
Parameters14B
Layers40
Trained context4,096 tokens
KV cache (fp16)200 KB per token

Source

Quantized byolkovi
FamilyPhi
Downloads (HF)65
Last updated2024-07-28

Frequently asked questions about saiga phi3 medium sft m1 d2 kto m5 d7

How much VRAM does saiga phi3 medium sft m1 d2 kto m5 d7 need?

About 10.6 GB at Q4_K_M with an 8K context window. That is the 8.4 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 saiga phi3 medium sft m1 d2 kto m5 d7 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 8.5 GB, so you would need a larger GPU or CPU offloading.

What is the file size of saiga phi3 medium sft m1 d2 kto m5 d7 GGUF?

It depends on the quant: IQ3_M is 6.3 GB, Q3_K_M is 6.8 GB, IQ4_XS is 7.6 GB, IQ4_NL is 8.0 GB, Q4_K_M is 8.4 GB, Q5_K_M is 9.9 GB and 3 more.

What context length does saiga phi3 medium sft m1 d2 kto m5 d7 support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 1.7 GB at 8K and 0.8 GB at 4K.

Where can I download saiga phi3 medium sft m1 d2 kto m5 d7 GGUF?

The files are on Hugging Face in the olkovi/saiga_phi3_medium_sft_m1_d2_kto_m5_d7 repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

Similar models

Browse all →

More Phi models

All Phi →

Other medium models

All medium models →