Hugging Face's logo

legraphista
/
xLAM-8x7b-r-IMat-GGUF

Text Generation
GGUF
PyTorch
English
function-calling
LLM Agent
tool-use
mistral
quantized
GGUF
quantization
imat
imatrix
static
16bit
8bit
6bit
5bit
4bit
3bit
2bit
1bit
conversational
Model card Files Community
xLAM-8x7b-r-IMat-GGUF / README.md
legraphista's picture
legraphista
Upload README.md with huggingface_hub
f30fbb5 verified
preview code
|
raw
history blame
7.59 kB
metadata
base_model: Salesforce/xLAM-8x7b-r
datasets:
  - Salesforce/xlam-function-calling-60k
extra_gated_button_content: Agree and access repository
extra_gated_heading: Acknowledge to follow corresponding license to access the repository
inference: false
language:
  - en
library_name: gguf
license: cc-by-nc-4.0
pipeline_tag: text-generation
quantized_by: legraphista
tags:
  - function-calling
  - LLM Agent
  - tool-use
  - mistral
  - pytorch
  - quantized
  - GGUF
  - quantization
  - imat
  - imatrix
  - static
  - 16bit
  - 8bit
  - 6bit
  - 5bit
  - 4bit
  - 3bit
  - 2bit
  - 1bit

xLAM-8x7b-r-IMat-GGUF

Llama.cpp imatrix quantization of Salesforce/xLAM-8x7b-r

Original Model: Salesforce/xLAM-8x7b-r
Original dtype: BF16 (bfloat16)
Quantized by: llama.cpp b3647
IMatrix dataset: here

Files

IMatrix

Status: ✅ Available
Link: here

Common Quants

Filename Quant type File Size Status Uses IMatrix Is Split
xLAM-8x7b-r.Q8_0/* Q8_0 49.63GB ✅ Available ⚪ Static ✂ Yes
xLAM-8x7b-r.Q6_K.gguf Q6_K 38.38GB ✅ Available ⚪ Static 📦 No
xLAM-8x7b-r.Q4_K.gguf Q4_K 28.45GB ✅ Available 🟢 IMatrix 📦 No
xLAM-8x7b-r.Q3_K.gguf Q3_K 22.55GB ✅ Available 🟢 IMatrix 📦 No
xLAM-8x7b-r.Q2_K.gguf Q2_K 17.31GB ✅ Available 🟢 IMatrix 📦 No

All Quants

Filename Quant type File Size Status Uses IMatrix Is Split
xLAM-8x7b-r.BF16/* BF16 93.41GB ✅ Available ⚪ Static ✂ Yes
xLAM-8x7b-r.FP16/* F16 93.41GB ✅ Available ⚪ Static ✂ Yes
xLAM-8x7b-r.Q8_0/* Q8_0 49.63GB ✅ Available ⚪ Static ✂ Yes
xLAM-8x7b-r.Q6_K.gguf Q6_K 38.38GB ✅ Available ⚪ Static 📦 No
xLAM-8x7b-r.Q5_K.gguf Q5_K 33.23GB ✅ Available ⚪ Static 📦 No
xLAM-8x7b-r.Q5_K_S.gguf Q5_K_S 32.23GB ✅ Available ⚪ Static 📦 No
xLAM-8x7b-r.Q4_K.gguf Q4_K 28.45GB ✅ Available 🟢 IMatrix 📦 No
xLAM-8x7b-r.Q4_K_S.gguf Q4_K_S 26.75GB ✅ Available 🟢 IMatrix 📦 No
xLAM-8x7b-r.IQ4_NL IQ4_NL - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ4_XS IQ4_XS - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.Q3_K.gguf Q3_K 22.55GB ✅ Available 🟢 IMatrix 📦 No
xLAM-8x7b-r.Q3_K_L Q3_K_L - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.Q3_K_S Q3_K_S - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ3_M IQ3_M - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ3_S IQ3_S - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ3_XS IQ3_XS - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ3_XXS IQ3_XXS - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.Q2_K.gguf Q2_K 17.31GB ✅ Available 🟢 IMatrix 📦 No
xLAM-8x7b-r.Q2_K_S Q2_K_S - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ2_M IQ2_M - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ2_S IQ2_S - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ2_XS IQ2_XS - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ2_XXS IQ2_XXS - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ1_M IQ1_M - ⏳ Processing 🟢 IMatrix -
xLAM-8x7b-r.IQ1_S IQ1_S - ⏳ Processing 🟢 IMatrix -

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/xLAM-8x7b-r-IMat-GGUF --include "xLAM-8x7b-r.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/xLAM-8x7b-r-IMat-GGUF --include "xLAM-8x7b-r.Q8_0/*" --local-dir ./
# see FAQ for merging GGUF's

Inference

Simple chat template 

<s> [INST] {user_prompt} [/INST] {assistant_response}</s> [INST] {next_user_prompt} [/INST]

Chat template with system prompt 

<s> [INST] {system_prompt}

{user_prompt} [/INST] {assistant_response}</s> [INST] {next_user_prompt} [/INST]

Llama.cpp 

llama.cpp/main -m xLAM-8x7b-r.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
  2. Locate your GGUF chunks folder (ex: xLAM-8x7b-r.Q8_0)
  3. Run gguf-split --merge xLAM-8x7b-r.Q8_0/xLAM-8x7b-r.Q8_0-00001-of-XXXXX.gguf xLAM-8x7b-r.Q8_0.gguf
    • Make sure to point gguf-split to the first chunk of the split.

Got a suggestion? Ping me @legraphista!