K2-Chat-IMat-GGUF

Llama.cpp imatrix quantization of LLM360/K2-Chat

Original Model: LLM360/K2-Chat
Original dtype: FP16 (float16)
Quantized by: llama.cpp b3051
IMatrix dataset: here

• Files
  • IMatrix
  • Common Quants
  • All Quants
• Downloading using huggingface-cli
• Inference
  • Llama.cpp
• FAQ
  • Why is the IMatrix not applied everywhere?
  • How do I merge a split GGUF?

---

Files

IMatrix

Status: ✅ Available
Link: here

Common Quants

Filename	Quant type	File Size	Status	Uses IMatrix	Is Split
K2-Chat.Q8_0/*	Q8_0	69.37GB	✅ Available	⚪ Static	✂ Yes
K2-Chat.Q6_K/*	Q6_K	53.56GB	✅ Available	⚪ Static	✂ Yes
K2-Chat.Q4_K.gguf	Q4_K	39.35GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.Q3_K.gguf	Q3_K	31.63GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.Q2_K.gguf	Q2_K	24.11GB	✅ Available	🟢 IMatrix	📦 No

All Quants

Filename	Quant type	File Size	Status	Uses IMatrix	Is Split
K2-Chat.FP16/*	F16	130.58GB	✅ Available	⚪ Static	✂ Yes
K2-Chat.Q8_0/*	Q8_0	69.37GB	✅ Available	⚪ Static	✂ Yes
K2-Chat.Q6_K/*	Q6_K	53.56GB	✅ Available	⚪ Static	✂ Yes
K2-Chat.Q5_K/*	Q5_K	46.24GB	✅ Available	⚪ Static	✂ Yes
K2-Chat.Q5_K_S.gguf	Q5_K_S	44.92GB	✅ Available	⚪ Static	📦 No
K2-Chat.Q4_K.gguf	Q4_K	39.35GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.Q4_K_S.gguf	Q4_K_S	37.06GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ4_NL.gguf	IQ4_NL	36.80GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ4_XS.gguf	IQ4_XS	34.76GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.Q3_K.gguf	Q3_K	31.63GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.Q3_K_L.gguf	Q3_K_L	34.65GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.Q3_K_S.gguf	Q3_K_S	28.16GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ3_M.gguf	IQ3_M	29.83GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ3_S.gguf	IQ3_S	28.16GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ3_XS.gguf	IQ3_XS	26.64GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ3_XXS.gguf	IQ3_XXS	24.67GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.Q2_K.gguf	Q2_K	24.11GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.Q2_K_S.gguf	Q2_K_S	21.98GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ2_M.gguf	IQ2_M	22.41GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ2_S.gguf	IQ2_S	20.78GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ2_XS.gguf	IQ2_XS	19.27GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ2_XXS.gguf	IQ2_XXS	17.47GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ1_M.gguf	IQ1_M	15.43GB	✅ Available	🟢 IMatrix	📦 No
K2-Chat.IQ1_S.gguf	IQ1_S	14.21GB	✅ Available	🟢 IMatrix	📦 No

Downloading using huggingface-cli

If you do not have hugginface-cli installed:

pip install -U "huggingface_hub[cli]"

Download the specific file you want:

huggingface-cli download legraphista/K2-Chat-IMat-GGUF --include "K2-Chat.Q8_0.gguf" --local-dir ./

If the model file is big, it has been split into multiple files. In order to download them all to a local folder, run:

huggingface-cli download legraphista/K2-Chat-IMat-GGUF --include "K2-Chat.Q8_0/*" --local-dir ./
# see FAQ for merging GGUF's

---

Inference

Llama.cpp

llama.cpp/main -m K2-Chat.Q8_0.gguf --color -i -p "prompt here"

---

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: K2-Chat.Q8_0)
3. Run gguf-split --merge K2-Chat.Q8_0/K2-Chat.Q8_0-00001-of-XXXXX.gguf K2-Chat.Q8_0.gguf
   • Make sure to point gguf-split to the first chunk of the split.

---

Got a suggestion? Ping me @legraphista!