Llama3-ChatQA-1.5-8B-IMat-GGUF

Llama.cpp imatrix quantization of nvidia/Llama3-ChatQA-1.5-8B

Original Model: nvidia/Llama3-ChatQA-1.5-8B
Original dtype: FP16 (float16)
Quantized by: llama.cpp b3003
IMatrix dataset: here

• Llama3-ChatQA-1.5-8B-IMat-GGUF
  • Files
    • IMatrix
    • Common Quants
    • All Quants
  • Downloading using huggingface-cli
  • Inference
    • Simple chat template
    • 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
Llama3-ChatQA-1.5-8B.Q8_0.gguf	Q8_0	8.54GB	✅ Available	⚪ No	📦 No
Llama3-ChatQA-1.5-8B.Q6_K.gguf	Q6_K	6.60GB	✅ Available	⚪ No	📦 No
Llama3-ChatQA-1.5-8B.Q4_K.gguf	Q4_K	4.92GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.Q3_K.gguf	Q3_K	4.02GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.Q2_K.gguf	Q2_K	3.18GB	✅ Available	🟢 Yes	📦 No

All Quants

Filename	Quant type	File Size	Status	Uses IMatrix	Is Split
Llama3-ChatQA-1.5-8B.FP16.gguf	F16	16.07GB	✅ Available	⚪ No	📦 No
Llama3-ChatQA-1.5-8B.Q5_K.gguf	Q5_K	5.73GB	✅ Available	⚪ No	📦 No
Llama3-ChatQA-1.5-8B.Q5_K_S.gguf	Q5_K_S	5.60GB	✅ Available	⚪ No	📦 No
Llama3-ChatQA-1.5-8B.Q4_K_S.gguf	Q4_K_S	4.69GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.Q3_K_L.gguf	Q3_K_L	4.32GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.Q3_K_S.gguf	Q3_K_S	3.66GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.Q2_K_S.gguf	Q2_K_S	2.99GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ4_NL.gguf	IQ4_NL	4.68GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ4_XS.gguf	IQ4_XS	4.45GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ3_M.gguf	IQ3_M	3.78GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ3_S.gguf	IQ3_S	3.68GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ3_XS.gguf	IQ3_XS	3.52GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ3_XXS.gguf	IQ3_XXS	3.27GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ2_M.gguf	IQ2_M	2.95GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ2_S.gguf	IQ2_S	2.76GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ2_XS.gguf	IQ2_XS	2.61GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ2_XXS.gguf	IQ2_XXS	2.40GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ1_M.gguf	IQ1_M	2.16GB	✅ Available	🟢 Yes	📦 No
Llama3-ChatQA-1.5-8B.IQ1_S.gguf	IQ1_S	2.02GB	✅ Available	🟢 Yes	📦 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/Llama3-ChatQA-1.5-8B-IMat-GGUF --include "Llama3-ChatQA-1.5-8B.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/Llama3-ChatQA-1.5-8B-IMat-GGUF --include "Llama3-ChatQA-1.5-8B.Q8_0/*" --local-dir Llama3-ChatQA-1.5-8B.Q8_0
# see FAQ for merging GGUF's

---

Inference

Simple chat template

<|begin_of_text|>System: This is a chat between a user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions based on the context. The assistant should also indicate when the answer cannot be found in the context.

User: Can you provide ways to eat combinations of bananas and dragonfruits?

Assistant: Sure! Here are some ways to eat bananas and dragonfruits together:
 1. Banana and dragonfruit smoothie: Blend bananas and dragonfruits together with some milk and honey.
 2. Banana and dragonfruit salad: Mix sliced bananas and dragonfruits together with some lemon juice and honey.

User: What about solving an 2x + 3 = 7 equation?

Assistant:

Llama.cpp

llama.cpp/main -m Llama3-ChatQA-1.5-8B.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: Llama3-ChatQA-1.5-8B.Q8_0)
3. Run gguf-split --merge Llama3-ChatQA-1.5-8B.Q8_0/Llama3-ChatQA-1.5-8B.Q8_0-00001-of-XXXXX.gguf Llama3-ChatQA-1.5-8B.Q8_0.gguf
   • Make sure to point gguf-split to the first chunk of the split.

---

Got a suggestion? Ping me @legraphista!