A Gradio web UI for Large Language Models. 0-GPTQ. . It is the result of quantising to 4bit using AutoGPTQ. . It applies to software engineers as well. You can either load quantized models from the Hub or your own HF quantized models. This repository showcases how we get an overview of this LM's capabilities. Text Generation • Updated Sep 14 • 65. cpp, etc. ; model_file: The name of the model file in repo or directory. Just don't bother with the powershell envs. GitHub Copilot vs. Contribution. Note: The reproduced result of StarCoder on MBPP. cpp (GGUF), Llama models. 1. Reload to refresh your session. It also significantly outperforms text-davinci-003, a model that's more than 10 times its size. | AutoGPTQTheBloke/Llama-2-13B-chat-GGML. New comments cannot be posted. 81k • 629. Project starcoder’s online platform provides video tutorials and recorded live class sessions which enable K-12 students to learn coding. The table below lists all the compatible models families and the associated binding repository. It also generates comments that explain what it is doing. its called hallucination and thats why you just insert the string where you want it to stop. Supported Models. Resources. What’s the difference between GPT-4 and StarCoder? Compare GPT-4 vs. 4, 5, and 8-bit GGML models for CPU+GPU inference. LocalAI - :robot: The free, Open Source OpenAI alternative. StarCoder in 2023 by cost, reviews, features, integrations, deployment, target market, support options, trial offers, training options, years in business, region, and more using the chart below. 0. Note: Though PaLM is not an open-source model, we still include its results here. This adds full GPU acceleration to llama. Model card Files Files and versions Community 1 Train Deploy Use in Transformers. License: bigcode-openrail-m. 你可以使用 model. Also, we release the technical report. If you don't have enough RAM, try increasing swap. Results StarCoder Bits group-size memory(MiB) wikitext2 ptb c4 stack checkpoint size(MB) FP32: 32-10. . Since GGUF is not yet available for Text Generation Inference yet, we will stick to GPTQ. StarCoder. cpp and ggml, including support GPT4ALL-J which is licensed under Apache 2. 11 tokens/s. Two other test models, TheBloke/CodeLlama-7B-GPTQ and TheBloke/Samantha-1. safetensors file: . We opensource our Qwen series, now including Qwen, the base language models, namely Qwen-7B and Qwen-14B, as well as Qwen-Chat, the chat models, namely Qwen-7B-Chat and Qwen-14B-Chat. StarCoder, a new open-access large language model (LLM) for code generation from ServiceNow and Hugging Face, is now available for Visual Studio Code, positioned as an alternative to GitHub Copilot. Convert the model to ggml FP16 format using python convert. 0: WizardLM-30B 1. TheBloke/starcoder-GPTQ. starcoder-GPTQ-4bit-128g. like 16. License: bigcode-openrail-m. RAM Requirements. 408: 1. You signed out in another tab or window. ; Our WizardMath-70B-V1. Note: The above table conducts a comprehensive comparison of our WizardCoder with other models on the HumanEval and MBPP benchmarks. - Home · oobabooga/text-generation-webui Wiki. The model will automatically load, and is now. Bigcode's Starcoder GGML These files are GGML format model files for Bigcode's Starcoder. My current research focuses on private local GPT solutions using open source LLMs, fine-tuning these models to adapt to specific domains and languages, and creating valuable workflows using. License. For coding assistance have you tried StarCoder? Also I find helping out with small functional modes is only helpful to a certain extent. On the command line, including multiple files at once. cpp. 0-GPTQ" # Or to load it locally, pass the local download pathreplit-code-v1-3b is a 2. TheBloke/guanaco-65B-GPTQ. The following tutorials and live class recording are available in starcoder. And many of these are 13B models that should work well with lower VRAM count GPUs! I recommend trying to load with Exllama (HF if possible). co/datasets/bigco de/the-stack. In the top left, click the refresh icon next to Model. cpp (GGUF), Llama models. Embeddings support. Commit . Reload to refresh your session. New VS Code Tool: StarCoderEx (AI Code Generator) By David Ramel. / gpt4all-lora. You switched accounts on another tab or window. 17323. Please note that these GGMLs are not compatible with llama. 4; Inference String Format The inference string is a concatenated string formed by combining conversation data (human and bot contents) in the training data format. So I doubt this would work, but maybe this does something "magic",. RAM Requirements. The extremely high inference cost, in both time and memory, is a big bottleneck for adopting a powerful transformer for solving. A comprehensive benchmark is available here. Note: The reproduced result of StarCoder on MBPP. How to run starcoder-GPTQ-4bit-128g? Question | Help I am looking at running this starcoder locally -- someone already made a 4bit/128 version (. ”. bigcode-tokenizer Public Jupyter Notebook 13 Apache-2. On a data science benchmark called DS-1000 it clearly beats it as well as all other open-access. The table below lists all the compatible models families and the associated binding repository. If you mean running time - then that is still pending with int-3 quant and quant 4 with 128 bin size. Discussion. starcoder-GPTQ-4bit-128g. High-throughput serving with various decoding algorithms, including parallel sampling, beam search, and more. The model will start downloading. Investigating the generalization behavior of LM probes trained to predict truth labels: (1) from one annotator to another, and (2) from easy questions to hard. It is used as input during the inference process. langchain-visualizer - Visualization and debugging tool for LangChain. At inference time, thanks to ALiBi, MPT-7B-StoryWriter-65k+ can extrapolate even beyond 65k tokens. Note: Though PaLM is not an open-source model, we still include its results here. Remove universal binary option when building for AVX2, AVX on macOS. Similar to LLaMA, we trained a ~15B parameter model for 1 trillion tokens. arxiv: 1911. 5: gpt4-2023. bigcode-analysis Public Repository for analysis and experiments in. StarCoder: may the source be with you! The BigCode community, an open-scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder and StarCoderBase: 15. ShipItMind/starcoder-gptq-4bit-128g. From the GPTQ paper, it is recommended to quantized the. BigCode 是由 Hugging Face 和 ServiceNow 共同领导的开放式科学合作项目. starcoder-GPTQ. Click Download. Slightly adjusted preprocessing of C4 and PTB for more realistic evaluations (used in our updated results); can be activated via the flag --new-eval. Follow Reddit's Content Policy. Hugging Face and ServiceNow have partnered to develop StarCoder, a new open-source language model for code. Two models were trained: - StarCoderBase, trained on 1 trillion tokens from The Stack (hf. To run GPTQ-for-LLaMa, you can use the following command: "python server. StarChat-β is the second model in the series, and is a fine-tuned version of StarCoderPlus that was trained on an "uncensored" variant of the openassistant-guanaco dataset. I recommend using the huggingface-hub Python library: pip3 install huggingface-hub>=0. GPTQ clearly outperforms here. OpenLLM is an open-source platform designed to facilitate the deployment and operation of large language models (LLMs) in real-world applications. 4-bit GPTQ models for GPU inference; 4, 5, and 8-bit GGML models for CPU+GPU inference; Unquantised fp16 model in pytorch format, for GPU inference and for further conversions; Compatibilty These files are not compatible with llama. StarCoder — which is licensed to allow for royalty-free use by anyone, including corporations — was trained in over 80. It is based on llama. g. StarCoderBase: Trained on 80+ languages from The Stack. 0 468 75 8 Updated Oct 31, 2023. mayank31398 add mmodel. Quantization of SantaCoder using GPTQ. LM Studio is an easy to use desktop app for experimenting with local and open-source Large Language Models (LLMs). you can use model. Text Generation Transformers. Download the 3B, 7B, or 13B model from Hugging Face. In this blog post, we’ll show how StarCoder can be fine-tuned for chat to create a personalised coding assistant![Updated on 2023-01-24: add a small section on Distillation. It is the result of quantising to 4bit using AutoGPTQ. co/settings/token) with this command: Cmd/Ctrl+Shift+P to open VSCode command palette. cpp is the wrong address for this case. Add To Compare. The release of StarCoder by the BigCode project was a major milestone for the open LLM community:. The model will start downloading. +Patreon special mentions**: Sam, theTransient, Jonathan Leane, Steven Wood, webtim, Johann-Peter Hartmann, Geoffrey Montalvo, Gabriel Tamborski, Willem Michiel, John. We also have extensions for: neovim. `pip install auto-gptq` Then try the following example code: ```python: from transformers import AutoTokenizer, pipeline, logging: from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig: import argparse: model_name_or_path = "TheBloke/WizardCoder-15B-1. py:776 and torch. 39 tokens/s, 241 tokens, context 39, seed 1866660043) Output generated in 33. This happe. The instructions can be found here. bin, . Currently they can be used with: KoboldCpp, a powerful inference engine based on llama. Starcoder is pure code, and not instruct tuned, but they provide a couple extended preambles that kindof, sortof do the trick. However, whilst checking for what version of huggingface_hub I had installed, I decided to update my Python environment to the one suggested in the requirements. For 40b it needs an A100-40G or equivalent. Bigcode's Starcoder GPTQ These files are GPTQ 4bit model files for Bigcode's Starcoder. Currently gpt2, gptj, gptneox, falcon, llama, mpt, starcoder (gptbigcode), dollyv2, and replit are supported. . Note: The above table conducts a comprehensive comparison of our WizardCoder with other models on the HumanEval and MBPP benchmarks. How to get oobabooga/text-generation-webui running on Windows or Linux with LLaMa-30b 4bit mode via GPTQ-for-LLaMa on an RTX 3090 start to finish. You signed in with another tab or window. Completion/Chat endpoint. StarCoder using this comparison chart. Similar to LLaMA, we trained a ~15B parameter model for 1 trillion tokens. Type: Llm: Login. 5-turbo for natural language to SQL generation tasks on our sql-eval framework, and significantly outperforms all popular open-source models. Transformers or GPTQ models are made of several files and must be placed in a subfolder. line 64. Saved searches Use saved searches to filter your results more quicklyAbstract: The BigCode community, an open-scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs),. Without doing those steps, the stuff based on the new GPTQ-for-LLama will. Saved searches Use saved searches to filter your results more quicklyGGML - Large Language Models for Everyone: a description of the GGML format provided by the maintainers of the llm Rust crate, which provides Rust bindings for GGML. Our models outperform open-source chat models on most benchmarks we tested,. like 16. Click the Model tab. You can specify any of the following StarCoder models via openllm start: bigcode/starcoder;. 6: WizardLM-7B 1. LLaMA and Llama2 (Meta) Meta release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. However, I have seen interesting tests with Starcoder. What is GPTQ? GPTQ is a post-training quantziation method to compress LLMs, like GPT. 6 pass@1 on the GSM8k Benchmarks, which is 24. co/settings/token) with this command: Cmd/Ctrl+Shift+P to open VSCode command palette. I am looking at a few different examples of using PEFT on different models. Output generated in 37. StarCoder: may the source be with you! The BigCode community, an open-scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder and StarCoderBase: 15. StarChat is a series of language models that are trained to act as helpful coding assistants. StarCoder is a transformer-based LLM capable of generating code from. 3: defog-sqlcoder: 64. ; model_type: The model type. 💫 StarCoder is a language model (LM) trained on source code and natural language text. 0 model achieves the 57. Copied. 1 to use the GPTBigCode architecture. ago. TGI enables high-performance text generation using Tensor Parallelism and dynamic batching for the most popular open-source LLMs, including StarCoder, BLOOM, GPT-NeoX, Llama, and T5. Add support for batching and beam search to 🤗 model. . The program can run on the CPU - no video card is required. Text-Generation-Inference is a solution build for deploying and serving Large Language Models (LLMs). Minetest is an open source voxel game engine with easy modding and game creation. bigcode-tokenizer Public StarCoder: 最先进的代码大模型 关于 BigCode . License: bigcode-openrail-m. You signed out in another tab or window. The app leverages your GPU when possible. You can load them with the revision flag:These files are GPTQ 4bit model files for WizardLM's WizardCoder 15B 1. 0 model achieves 81. The StarCoder models, which have a context length of over 8,000 tokens, can process more input than any other open LLM, opening the door to a wide variety of exciting new uses. cpp, gptq, ggml, llama-cpp-python, bitsandbytes, qlora, gptq_for_llama, chatglm. In this video, I will demonstra. The Bloke’s WizardLM-7B-uncensored-GPTQ These files are GPTQ 4bit model files for Eric Hartford’s ‘uncensored’ version of WizardLM. Tensor library for. Note: The above table conducts a comprehensive comparison of our WizardCoder with other models on the HumanEval and MBPP benchmarks. TheBloke/starcoder-GPTQ. ] Large transformer models are mainstream nowadays, creating SoTA results for a variety of tasks. The 15B parameter model outperforms models such as OpenAI’s code-cushman-001 on popular. This is a C++ example running 💫 StarCoder inference using the ggml library. Self-hosted, community-driven and local-first. HF API token. marella/ctransformers: Python bindings for GGML models. The model created as a part of the BigCode initiative is an improved version of the StarCodeFor illustration, GPTQ can quantize the largest publicly-available mod-els, OPT-175B and BLOOM-176B, in approximately four GPU hours, with minimal increase in perplexity, known to be a very stringent accuracy metric. Links are on the above table. 14135. It is an OpenAI API-compatible wrapper ctransformers supporting GGML / GPTQ with optional CUDA/Metal acceleration. It is optimized to run 7-13B parameter LLMs on the CPU's of any computer running OSX/Windows/Linux. Token stream support. 3: Call for Feedbacks. For the first time ever, this means GGML can now outperform AutoGPTQ and GPTQ-for-LLaMa inference (though it still loses to exllama) Note: if you test this, be aware that you should now use --threads 1 as it's no longer beneficial to use. It is difficult to see what is happening without seing the trace and the content of your checkpoint folder. Note: Any StarCoder variants can be deployed with OpenLLM. In total, the training dataset contains 175B tokens, which were repeated over 3 epochs -- in total, replit-code-v1-3b has been trained on 525B tokens (~195 tokens per parameter). In particular, the model has not been aligned to human preferences with techniques like RLHF, so may generate. Supports transformers, GPTQ, AWQ, EXL2, llama. 5B parameters created by finetuning StarCoder on CommitPackFT &. Slightly adjusted preprocessing of C4 and PTB for more realistic evaluations (used in our updated results); can be activated via the flag -. You signed out in another tab or window. The GPT4-x-Alpaca is a remarkable open-source AI LLM model that operates without censorship, surpassing GPT-4 in performance. StarCoder and StarCoderBase are Large Language Models for Code (Code LLMs) trained on permissively licensed data from GitHub, including from 80+ programming languages, Git commits, GitHub issues, and Jupyter notebooks. Home of StarCoder: fine-tuning & inference! Python 6,623 Apache-2. BigCode's StarCoder Plus. jupyter. 738: 59195: BF16: 16-10. Changed to support new features proposed by GPTQ. README. ChatGPT. You signed out in another tab or window. Found the following quantized model: modelsanon8231489123_vicuna-13b-GPTQ-4bit-128gvicuna-13b-4bit-128g. Besides llama based models, LocalAI is compatible also with other architectures. Model compatibility table. Demos . Runs ggml, gguf,. The openassistant-guanaco dataset was further trimmed to within 2 standard deviations of token size for input and output pairs and all non-english data has been removed to reduce. cpp (through llama-cpp-python), ExLlama, ExLlamaV2, AutoGPTQ, GPTQ-for-LLaMa, CTransformers, AutoAWQ ; Dropdown menu for quickly switching between different modelsHi. Model Summary. So besides GPT4, I have found Codeium to be the best imo. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. New PR llama. Token stream support. The more performant GPTQ kernels from @turboderp's exllamav2 library are now available directly in AutoGPTQ, and are the default backend choice. 0 468 75 8 Updated Oct 31, 2023. two new tricks:--act-order (quantizing columns in order of decreasing activation size) and --true-sequential. 8: WizardCoder-15B 1. There's an open issue for implementing GPTQ quantization in 3-bit and 4-bit. arxiv: 2207. ), which is permissively licensed with inspection tools, deduplication and opt-out - StarCoder, a fine-tuned version of. StarCoder Bits group-size memory(MiB) wikitext2 ptb c4 stack checkpoint size(MB) FP32: 32-10. The model will start downloading. ; lib: The path to a shared library or. But for the GGML / GGUF format, it's more about having enough RAM. I tried to issue 3 requests from 3 different devices and it waits till one is finished and then continues to the next one. Multiple GPTQ parameter permutations are provided; see Provided Files below for details of the options provided, their parameters, and the software used to create them. You switched accounts on another tab or window. Repository: bigcode/Megatron-LM. py--model TheBloke_vicuna-13B-1. 0 2 0 0 Updated Oct 24, 2023. The model will start downloading. Depending on your operating system, follow the appropriate commands below: M1 Mac/OSX: Execute the following command: . Codeium currently provides AI-generated autocomplete in more than 20 programming languages (including Python and JS, Java, TS, Java and Go) and integrates directly to the developer's IDE (VSCode, JetBrains or Jupyter notebooks. Repositories available 4-bit GPTQ models for GPU inference; 4, 5, and 8-bit GGML models for CPU+GPU inference; Unquantised fp16 model in pytorch format, for GPU inference and for further. This guide actually works well for linux too. StarCoder using this comparison chart. If you are still getting issues with multi-gpu you need to update the file modulesGPTQ_Loader. HumanEval is a widely used benchmark for Python that checks whether or not a. 3 Python GPTQ-for-LLaMa VS starcoder Home of StarCoder: fine-tuning & inference! serge. 05/08/2023. 5B parameter models trained on 80+ programming languages from The Stack (v1. Bigcode's Starcoder GPTQ These files are GPTQ 4bit model files for Bigcode's Starcoder. - GitHub - oobabooga/text-generation-webui: A Gradio web UI for Large Language Models. I tried with tiny_starcoder_py model as the weight size were quite small to fit without mem64, and tried to see the performance/accuracy. arxiv: 2305. Single GPU for. Add AutoGPTQ's cpu kernel. (it also works on GPU) Conversion is usually quite slim and the 8. Visit the HuggingFace Model Hub to see more StarCoder-compatible models. Testing. Example:. mayank31398 already made GPTQ versions of it both in 8 and 4 bits but,. Supercharger has the model build unit tests, and then uses the unit test to score the code it generated, debug/improve the code based off of the unit test quality score, and then run it. Saved searches Use saved searches to filter your results more quicklyStarCoder presents a quantized version as well as a quantized 1B version. Contribution. GPTQ-for-SantaCoder-and-StarCoder Quantization of SantaCoder using GPTQ GPTQ is SOTA one-shot weight quantization method This code is based on GPTQ Changed to. 17323. Supported models. StarChat Alpha is the first of these models, and as an alpha release is only intended for educational or research purpopses. Streaming outputs. [!NOTE] When using the Inference API, you will probably encounter some limitations. SQLCoder is a 15B parameter model that slightly outperforms gpt-3. arxiv: 2210. SQLCoder is fine-tuned on a base StarCoder model. 🚂 State-of-the-art LLMs: Integrated support for a wide. 1. 0-GPTQ. I don't quite understand where the values of the target modules come from. Under Download custom model or LoRA, enter TheBloke/WizardCoder-15B-1. It is not llama based, therefore llama. Starcoder itself isn't instruction tuned, and I have found to be very fiddly with prompts. exllamav2 integration by @SunMarc in #349; CPU inference support. Doesnt require using specific prompt format like starcoder. StarCoder. It doesn’t just predict code; it can also help you review code and solve issues using metadata, thanks to being trained with special tokens. Please refer to their papers for the same. Model type of pre-quantized model. Changed to support new features proposed by GPTQ. 9%: 2023. The GTX 1660 or 2060, AMD 5700 XT, or RTX 3050 or 3060 would all work nicely. md. main_custom: Packaged. Runs ggml, gguf, GPTQ, onnx, TF compatible models: llama, llama2, rwkv, whisper, vicuna, koala, cerebras, falcon, dolly, starcoder, and many others llama_index - LlamaIndex (formerly GPT Index) is a data framework for your LLM applications GPTQ-for-LLaMa - 4 bits quantization of LLaMA using GPTQI tried to use the gptq models such as Bloke 33b with the new changes to TGI regarding gptq. StarCoder and StarCoderBase are Large Language Models for Code (Code LLMs) trained on permissively licensed data from GitHub, including from 80+ programming languages,. intellij. To run GPTQ-for-LLaMa, you'll need to use the "--loader" parameter with the value "gptq-for-llama". In this paper, we present a new post-training quantization method, called GPTQ,1 The StarCoder models, which have a context length of over 8,000 tokens, can process more input than any other open LLM, opening the door to a wide variety of exciting new uses. It uses llm-ls as its backend. 比如， WizardLM，vicuna 和 gpt4all 模型的 model_type 皆为 llama， 因此这些模型皆被 auto_gptq 所. You signed in with another tab or window. GPT4All FAQ What models are supported by the GPT4All ecosystem? Currently, there are six different model architectures that are supported: GPT-J - Based off of the GPT-J architecture with examples found here; LLaMA - Based off of the LLaMA architecture with examples found here; MPT - Based off of Mosaic ML's MPT architecture with examples. If that fails then you've got other fish to fry before poking the wizard variant. GPTQ is a type of quantization (mainly used for models that run on a GPU). 4. Requires the bigcode fork of transformers. 1 6,600 8. It's a 15. 5B parameter models trained on 80+ programming languages from The Stack (v1. Expected behavior. Load it with AutoGPTQ and it. . py --listen --chat --model GodRain_WizardCoder-15B-V1. We welcome everyone to use your professional and difficult instructions to evaluate WizardLM, and show us examples of poor performance and your suggestions in the issue discussion area. It is the result of quantising to 4bit using AutoGPTQ. 1k • 34. alpaca-lora-65B-GPTQ-4bit-1024g. 0-GGUF wizardcoder. Download and install miniconda (Windows Only) Download and install. In the world of deploying and serving Large Language Models (LLMs), two notable frameworks have emerged as powerful solutions: Text Generation Interface (TGI) and vLLM. In the Model dropdown, choose the model you just downloaded: stablecode-completion-alpha-3b-4k-GPTQ. Exllama v2 GPTQ kernel support. Doesnt require using specific prompt format like starcoder. StarCoder LLM is out! 100% coding specialized Really hope to see more specialized models becoming more common than general use ones, like one that is a math expert, history expert. Models; Datasets; Spaces; Docs示例 提供了大量示例脚本以将 auto_gptq 用于不同领域。 支持的模型 . , 2022; Dettmers et al. You can supply your HF API token ( hf. we address this challenge, and propose GPTQ, a new one-shot weight quantiza-tion method based on approximate second-order information, that is both highly-accurate and highly. The dataset was created as part of the BigCode Project, an open scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs). 801. Project Starcoder programming from beginning to end. 1-4bit --loader gptq-for-llama". com Hi folks, back with an update to the HumanEval+ programming ranking I posted the other day incorporating your feedback - and some closed models for comparison! Now has improved generation params, new models: Falcon, Starcoder, Codegen, Claude+, Bard, OpenAssistant and more : r/LocalLLaMA. Note: Any StarCoder variants can be deployed with OpenLLM. update no_split_module_classes=["LLaMADecoderLayer"] to no_split_module_classes=["LlamaDecoderLayer"]. A less hyped framework compared to ggml/gptq is CTranslate2. The StarCoder models are 15. The open‑access, open‑science, open‑governance 15 billion parameter StarCoder LLM makes generative AI more transparent and accessible to enable responsible innovation. Develop. Click the Model tab. Besides llama based models, LocalAI is compatible also with other architectures. 2), with opt-out requests excluded. Embeddings support. The WizardCoder-Guanaco-15B-V1. Here are step-by-step instructions on how I managed to get the latest GPTQ models to work with runpod. A Gradio web UI for Large Language Models. py ShipItMind/starcoder-gptq-4bit-128g Downloading the model to models/ShipItMind_starcoder-gptq-4bit-128g. cpp, gpt4all, rwkv. If you want 8-bit weights, visit starcoderbase-GPTQ-8bit-128g. GPT4All Chat UI. TF compatible models: llama, llama2, rwkv, whisper, vicuna, koala, cerebras, falcon, dolly, starcoder, and many others llama_index - LlamaIndex (formerly GPT Index) is a data framework for your LLM. Further, we show that our model can also provide robust results in the extreme quantization regime,Bigcode's StarcoderPlus GPTQ These files are GPTQ 4bit model files for Bigcode's StarcoderPlus. SQLCoder is fine-tuned on a base StarCoder model.