File size: 1,229 Bytes
689e66f 468d58c 63078da 689e66f 468d58c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 |
---
license: bigcode-openrail-m
pipeline_tag: text-generation
library_name: gguf
base_model: TechxGenus/starcoder2-15b-instruct
---
<u>**NOTE**</u>: You will need a recent build of llama.cpp to run these quants (i.e. at least commit `494c870`).
GGUF importance matrix (imatrix) quants for https://huggingface.co/TechxGenus/starcoder2-15b-instruct
* The importance matrix was trained for ~50K tokens (105 batches of 512 tokens) using a [general purpose imatrix calibration dataset](https://github.com/ggerganov/llama.cpp/discussions/5263#discussioncomment-8395384).
* The [imatrix is being used on the K-quants](https://github.com/ggerganov/llama.cpp/pull/4930) as well.
> Fine-tuned starcoder2-15b with an additional 0.7 billion high-quality, code-related tokens for 3 epochs. We used DeepSpeed ZeRO 3 and Flash Attention 2 to accelerate the training process. It achieves 77.4 pass@1 on HumanEval-Python. This model operates using the Alpaca instruction format (excluding the system prompt).
| Layers | Context | [Template](https://huggingface.co/TechxGenus/starcoder2-15b-instruct#usage) |
| --- | --- | --- |
| <pre>40</pre> | <pre>16384</pre> | <pre>### Instruction<br>{instruction}<br>### Response<br>{response}</pre> |
|