README.md

metadata

base_model: microsoft/Phi-3-medium-4k-instruct
inference: false
language:
  - multilingual
library_name: gguf
license: mit
license_link: https://huggingface.co/microsoft/Phi-3-medium-4k-instruct/resolve/main/LICENSE
pipeline_tag: text-generation
quantized_by: legraphista
tags:
  - quantized
  - GGUF
  - imatrix
  - quantization
  - imat
  - imatrix
  - static

Phi-3-medium-4k-instruct-IMat-GGUF

Llama.cpp imatrix quantization of microsoft/Phi-3-medium-4k-instruct

Original Model: microsoft/Phi-3-medium-4k-instruct
Original dtype: BF16 (bfloat16)
Quantized by: llama.cpp b2998
IMatrix dataset: here

Files

IMatrix

Status: ✅ Available
Link: here

Common Quants

Filename	Quant type	File Size	Status	Uses IMatrix	Is Split
Phi-3-medium-4k-instruct.Q8_0.gguf	Q8_0	14.83GB	✅ Available	⚪ No	📦 No
Phi-3-medium-4k-instruct.Q6_K.gguf	Q6_K	11.45GB	✅ Available	⚪ No	📦 No
Phi-3-medium-4k-instruct.Q4_K.gguf	Q4_K	8.57GB	✅ Available	🟢 Yes	📦 No
Phi-3-medium-4k-instruct.Q3_K.gguf	Q3_K	6.92GB	✅ Available	🟢 Yes	📦 No
Phi-3-medium-4k-instruct.Q2_K.gguf	Q2_K	5.14GB	✅ Available	🟢 Yes	📦 No

All Quants

Filename	Quant type	File Size	Status	Uses IMatrix	Is Split
Phi-3-medium-4k-instruct.FP16.gguf	F16	27.92GB	✅ Available	⚪ No	📦 No
Phi-3-medium-4k-instruct.BF16.gguf	BF16	27.92GB	✅ Available	⚪ No	📦 No
Phi-3-medium-4k-instruct.Q5_K.gguf	Q5_K	10.07GB	✅ Available	⚪ No	📦 No
Phi-3-medium-4k-instruct.Q5_K_S.gguf	Q5_K_S	9.62GB	✅ Available	⚪ No	📦 No
Phi-3-medium-4k-instruct.Q4_K_S.gguf	Q4_K_S	7.95GB	✅ Available	🟢 Yes	📦 No
Phi-3-medium-4k-instruct.Q3_K_L.gguf	Q3_K_L	7.49GB	✅ Available	🟢 Yes	📦 No
Phi-3-medium-4k-instruct.Q3_K_S.gguf	Q3_K_S	6.06GB	✅ Available	🟢 Yes	📦 No
Phi-3-medium-4k-instruct.Q2_K_S.gguf	Q2_K_S	4.77GB	✅ Available	🟢 Yes	📦 No
Phi-3-medium-4k-instruct.IQ4_NL.gguf	IQ4_NL	7.90GB	✅ Available	🟢 Yes	📦 No
Phi-3-medium-4k-instruct.IQ4_XS.gguf	IQ4_XS	7.47GB	✅ Available	🟢 Yes	📦 No
Phi-3-medium-4k-instruct.IQ3_M.gguf	IQ3_M	6.47GB	✅ Available	🟢 Yes	📦 No
Phi-3-medium-4k-instruct.IQ3_S.gguf	IQ3_S	6.06GB	✅ Available	🟢 Yes	📦 No
Phi-3-medium-4k-instruct.IQ3_XS	IQ3_XS	-	⏳ Processing	🟢 Yes	-
Phi-3-medium-4k-instruct.IQ3_XXS	IQ3_XXS	-	⏳ Processing	🟢 Yes	-
Phi-3-medium-4k-instruct.IQ2_M	IQ2_M	-	⏳ Processing	🟢 Yes	-
Phi-3-medium-4k-instruct.IQ2_S	IQ2_S	-	⏳ Processing	🟢 Yes	-
Phi-3-medium-4k-instruct.IQ2_XS	IQ2_XS	-	⏳ Processing	🟢 Yes	-
Phi-3-medium-4k-instruct.IQ2_XXS	IQ2_XXS	-	⏳ Processing	🟢 Yes	-
Phi-3-medium-4k-instruct.IQ1_M	IQ1_M	-	⏳ Processing	🟢 Yes	-
Phi-3-medium-4k-instruct.IQ1_S	IQ1_S	-	⏳ Processing	🟢 Yes	-

Downloading using huggingface-cli

First, make sure you have hugginface-cli installed:

pip install -U "huggingface_hub[cli]"

Then, you can target the specific file you want:

huggingface-cli download legraphista/Phi-3-medium-4k-instruct-IMat-GGUF --include "Phi-3-medium-4k-instruct.Q8_0.gguf" --local-dir ./

If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:

huggingface-cli download legraphista/Phi-3-medium-4k-instruct-IMat-GGUF --include "Phi-3-medium-4k-instruct.Q8_0/*" --local-dir Phi-3-medium-4k-instruct.Q8_0
# see FAQ for merging GGUF's

FAQ

Why is the IMatrix not applied everywhere?

According to this investigation, it appears that lower quantizations are the only ones that benefit from the imatrix input (as per hellaswag results).

How do I merge a split GGUF?

1. Make sure you have gguf-split available
   • To get hold of gguf-split, navigate to https://github.com/ggerganov/llama.cpp/releases
   • Download the appropriate zip for your system from the latest release
   • Unzip the archive and you should be able to find gguf-split
2. Locate your GGUF chunks folder (ex: Phi-3-medium-4k-instruct.Q8_0)
3. Run gguf-split --merge Phi-3-medium-4k-instruct.Q8_0/Phi-3-medium-4k-instruct.Q8_0-00001-of-XXXXX.gguf Phi-3-medium-4k-instruct.Q8_0.gguf
   • Make sure to point gguf-split to the first chunk of the split.

---

Got a suggestion? Ping me @legraphista!