metadata
pipeline_tag: text-generation
inference: false
license: apache-2.0
datasets:
- bigcode/commitpackft
- TIGER-Lab/MathInstruct
- meta-math/MetaMathQA
- glaiveai/glaive-code-assistant-v3
- glaive-function-calling-v2
- bugdaryan/sql-create-context-instruction
- garage-bAInd/Open-Platypus
- nvidia/HelpSteer
- bigcode/self-oss-instruct-sc2-exec-filter-50k
metrics:
- code_eval
library_name: transformers
tags:
- code
- granite
- TensorBlock
- GGUF
base_model: ibm-granite/granite-8b-code-instruct-128k
model-index:
- name: granite-8B-Code-instruct-128k
results:
- task:
type: text-generation
dataset:
name: HumanEvalSynthesis (Python)
type: bigcode/humanevalpack
metrics:
- type: pass@1
value: 62.2
name: pass@1
verified: false
- type: pass@1
value: 51.4
name: pass@1
verified: false
- type: pass@1
value: 38.9
name: pass@1
verified: false
- type: pass@1
value: 38.3
name: pass@1
verified: false
- task:
type: text-generation
dataset:
name: RepoQA (Python@16K)
type: repoqa
metrics:
- type: pass@1 (thresh=0.5)
value: 73
name: pass@1 (thresh=0.5)
verified: false
- type: pass@1 (thresh=0.5)
value: 37
name: pass@1 (thresh=0.5)
verified: false
- type: pass@1 (thresh=0.5)
value: 73
name: pass@1 (thresh=0.5)
verified: false
- type: pass@1 (thresh=0.5)
value: 62
name: pass@1 (thresh=0.5)
verified: false
- type: pass@1 (thresh=0.5)
value: 63
name: pass@1 (thresh=0.5)
verified: false
Feedback and support: TensorBlock's Twitter/X, Telegram Group and Discord server
ibm-granite/granite-8b-code-instruct-128k - GGUF
This repo contains GGUF format model files for ibm-granite/granite-8b-code-instruct-128k.
The files were quantized using machines provided by TensorBlock, and they are compatible with llama.cpp as of commit b4011.
Prompt template
System:
{system_prompt}
Question:
{prompt}
Answer:
Model file specification
| Filename | Quant type | File Size | Description |
|---|---|---|---|
| granite-8b-code-instruct-128k-Q2_K.gguf | Q2_K | 2.852 GB | smallest, significant quality loss - not recommended for most purposes |
| granite-8b-code-instruct-128k-Q3_K_S.gguf | Q3_K_S | 3.304 GB | very small, high quality loss |
| granite-8b-code-instruct-128k-Q3_K_M.gguf | Q3_K_M | 3.674 GB | very small, high quality loss |
| granite-8b-code-instruct-128k-Q3_K_L.gguf | Q3_K_L | 3.993 GB | small, substantial quality loss |
| granite-8b-code-instruct-128k-Q4_0.gguf | Q4_0 | 4.276 GB | legacy; small, very high quality loss - prefer using Q3_K_M |
| granite-8b-code-instruct-128k-Q4_K_S.gguf | Q4_K_S | 4.305 GB | small, greater quality loss |
| granite-8b-code-instruct-128k-Q4_K_M.gguf | Q4_K_M | 4.548 GB | medium, balanced quality - recommended |
| granite-8b-code-instruct-128k-Q5_0.gguf | Q5_0 | 5.190 GB | legacy; medium, balanced quality - prefer using Q4_K_M |
| granite-8b-code-instruct-128k-Q5_K_S.gguf | Q5_K_S | 5.190 GB | large, low quality loss - recommended |
| granite-8b-code-instruct-128k-Q5_K_M.gguf | Q5_K_M | 5.330 GB | large, very low quality loss - recommended |
| granite-8b-code-instruct-128k-Q6_K.gguf | Q6_K | 6.161 GB | very large, extremely low quality loss |
| granite-8b-code-instruct-128k-Q8_0.gguf | Q8_0 | 7.977 GB | very large, extremely low quality loss - not recommended |
Downloading instruction
Command line
Firstly, install Huggingface Client
pip install -U "huggingface_hub[cli]"
Then, downoad the individual model file the a local directory
huggingface-cli download tensorblock/granite-8b-code-instruct-128k-GGUF --include "granite-8b-code-instruct-128k-Q2_K.gguf" --local-dir MY_LOCAL_DIR
If you wanna download multiple model files with a pattern (e.g., *Q4_K*gguf), you can try:
huggingface-cli download tensorblock/granite-8b-code-instruct-128k-GGUF --local-dir MY_LOCAL_DIR --local-dir-use-symlinks False --include='*Q4_K*gguf'