Exllama v2 Quantizations of starcoder2-15b-instruct-v0.1
Using turboderp's ExLlamaV2 v0.0.20 for quantization.
The "main" branch only contains the measurement.json, download one of the other branches for the model (see below)
Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.
Original model: https://huggingface.co/bigcode/starcoder2-15b-instruct-v0.1
Prompt format
<|endoftext|>You are an exceptionally intelligent coding assistant that consistently delivers accurate and reliable responses to user instructions.
### Instruction
{prompt}
### Response
<|endoftext|>
Available sizes
Branch | Bits | lm_head bits | VRAM (4k) | VRAM (16k) | VRAM (32k) | Description |
---|---|---|---|---|---|---|
8_0 | 8.0 | 8.0 | 15.8 GB | 16.8 GB | 18.1 GB | Maximum quality that ExLlamaV2 can produce, near unquantized performance. |
6_5 | 6.5 | 8.0 | 13.9 GB | 14.9 GB | 16.2 GB | Near unquantized performance at vastly reduced size, recommended. |
5_0 | 5.0 | 6.0 | 11.0 GB | 12.0 GB | 13.2 GB | Slightly lower quality vs 6.5. |
4_25 | 4.25 | 6.0 | 9.5 GB | 10.5 GB | 11.8 GB | GPTQ equivalent bits per weight. |
3_5 | 3.5 | 6.0 | 8.1 GB | 9.1 GB | 10.4 GB | Lower quality, not recommended. |
Download instructions
With git:
git clone --single-branch --branch 6_5 https://huggingface.co/bartowski/starcoder2-15b-instruct-v0.1-exl2 starcoder2-15b-instruct-v0.1-exl2-6_5
With huggingface hub (credit to TheBloke for instructions):
pip3 install huggingface-hub
To download a specific branch, use the --revision
parameter. For example, to download the 6.5 bpw branch:
Linux:
huggingface-cli download bartowski/starcoder2-15b-instruct-v0.1-exl2 --revision 6_5 --local-dir starcoder2-15b-instruct-v0.1-exl2-6_5 --local-dir-use-symlinks False
Windows (which apparently doesn't like _ in folders sometimes?):
huggingface-cli download bartowski/starcoder2-15b-instruct-v0.1-exl2 --revision 6_5 --local-dir starcoder2-15b-instruct-v0.1-exl2-6.5 --local-dir-use-symlinks False
Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for bartowski/starcoder2-15b-instruct-v0.1-exl2
Base model
bigcode/starcoder2-15bDataset used to train bartowski/starcoder2-15b-instruct-v0.1-exl2
Evaluation results
- pass@1 on LiveCodeBench (code generation)self-reported20.400
- pass@1 on LiveCodeBench (self repair)self-reported20.900
- pass@1 on LiveCodeBench (test output prediction)self-reported29.800
- pass@1 on LiveCodeBench (code execution)self-reported28.100
- pass@1 on HumanEvalself-reported72.600
- pass@1 on HumanEval+self-reported63.400
- pass@1 on MBPPself-reported75.200
- pass@1 on MBPP+self-reported61.200
- pass@1 on DS-1000self-reported40.600