Qwen2-1.5B-Instruct-IMat-GGUF/README.md

base_model, inference, language, library_name, license, pipeline_tag, quantized_by, tags

base_model	inference	language	library_name	license	pipeline_tag	quantized_by	tags
Qwen/Qwen2-1.5B-Instruct	false	en	gguf	apache-2.0	text-generation	legraphista	chat quantized GGUF imatrix quantization imat imatrix static 16bit 8bit 6bit 5bit 4bit 3bit 2bit 1bit

Qwen2-1.5B-Instruct-IMat-GGUF

Llama.cpp imatrix quantization of Qwen/Qwen2-1.5B-Instruct

Original Model: Qwen/Qwen2-1.5B-Instruct
Original dtype: BF16 (bfloat16)
Quantized by: llama.cpp b3091
IMatrix dataset: here

• Files
  • IMatrix
  • Common Quants
  • All Quants
• Downloading using huggingface-cli
• Inference
  • Simple chat template
  • Chat template with system prompt
  • 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
Qwen2-1.5B-Instruct.Q8_0.gguf	Q8_0	1.65GB	✅ Available	⚪ Static	📦 No
Qwen2-1.5B-Instruct.Q6_K.gguf	Q6_K	1.27GB	✅ Available	⚪ Static	📦 No
Qwen2-1.5B-Instruct.Q4_K.gguf	Q4_K	986.05MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.Q3_K.gguf	Q3_K	824.18MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.Q2_K.gguf	Q2_K	676.30MB	✅ Available	🟢 IMatrix	📦 No

All Quants

Filename	Quant type	File Size	Status	Uses IMatrix	Is Split
Qwen2-1.5B-Instruct.BF16.gguf	BF16	3.09GB	✅ Available	⚪ Static	📦 No
Qwen2-1.5B-Instruct.FP16.gguf	F16	3.09GB	✅ Available	⚪ Static	📦 No
Qwen2-1.5B-Instruct.Q8_0.gguf	Q8_0	1.65GB	✅ Available	⚪ Static	📦 No
Qwen2-1.5B-Instruct.Q6_K.gguf	Q6_K	1.27GB	✅ Available	⚪ Static	📦 No
Qwen2-1.5B-Instruct.Q5_K.gguf	Q5_K	1.13GB	✅ Available	⚪ Static	📦 No
Qwen2-1.5B-Instruct.Q5_K_S.gguf	Q5_K_S	1.10GB	✅ Available	⚪ Static	📦 No
Qwen2-1.5B-Instruct.Q4_K.gguf	Q4_K	986.05MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.Q4_K_S.gguf	Q4_K_S	940.31MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ4_NL.gguf	IQ4_NL	936.33MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ4_XS.gguf	IQ4_XS	895.73MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.Q3_K.gguf	Q3_K	824.18MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.Q3_K_L.gguf	Q3_K_L	880.16MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.Q3_K_S.gguf	Q3_K_S	760.94MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ3_M.gguf	IQ3_M	776.66MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ3_S.gguf	IQ3_S	762.40MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ3_XS.gguf	IQ3_XS	731.70MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ3_XXS.gguf	IQ3_XXS	668.79MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.Q2_K.gguf	Q2_K	676.30MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.Q2_K_S.gguf	Q2_K_S	640.13MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ2_M.gguf	IQ2_M	601.05MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ2_S.gguf	IQ2_S	563.81MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ2_XS.gguf	IQ2_XS	550.32MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ2_XXS.gguf	IQ2_XXS	511.01MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ1_M.gguf	IQ1_M	464.46MB	✅ Available	🟢 IMatrix	📦 No
Qwen2-1.5B-Instruct.IQ1_S.gguf	IQ1_S	436.52MB	✅ 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/Qwen2-1.5B-Instruct-IMat-GGUF --include "Qwen2-1.5B-Instruct.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/Qwen2-1.5B-Instruct-IMat-GGUF --include "Qwen2-1.5B-Instruct.Q8_0/*" --local-dir ./
# see FAQ for merging GGUF's

---

Inference

Simple chat template

<|im_start|>system
You are a helpful assistant.<|im_end|>
<|im_start|>user
{user_prompt}<|im_end|>
<|im_start|>assistant
{assistant_response}<|im_end|>
<|im_start|>user
{next_user_prompt}<|im_end|>

Chat template with system prompt

<|im_start|>system
{system_prompt}<|im_end|>
<|im_start|>user
{user_prompt}<|im_end|>
<|im_start|>assistant
{assistant_response}<|im_end|>
<|im_start|>user
{next_user_prompt}<|im_end|>

Llama.cpp

llama.cpp/main -m Qwen2-1.5B-Instruct.Q8_0.gguf --color -i -p "prompt here (according to the chat template)"

---

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: Qwen2-1.5B-Instruct.Q8_0)
3. Run gguf-split --merge Qwen2-1.5B-Instruct.Q8_0/Qwen2-1.5B-Instruct.Q8_0-00001-of-XXXXX.gguf Qwen2-1.5B-Instruct.Q8_0.gguf
   • Make sure to point gguf-split to the first chunk of the split.

---

Got a suggestion? Ping me @legraphista!