Etherll

This is a binary audio classification model that determines if a speech recording is clean or if it is degraded by environmental noise. It is specifically trained to be robust and understand the difference between clean audio and audio that has actual background noise (like cars, music, or other people talking). - LABEL0: `clean`: The audio contains speech with no significant environmental noise. This includes high-quality recordings as well as recordings with source artifacts like hiss, clipping, or "bad microphone" quality. - LABEL1: `noisy`: The audio contains speech that is obscured by external, environmental background noise. This model is ideal for: - Pre-processing a large audio dataset to filter for clean samples. - Automatically tagging audio clips for quality control. - As a gate for ASR (Automatic Speech Recognition) systems that perform better on clean audio. Limitations: - This model is a classifier, not a noise-reduction tool. It only tells you if environmental noise is present. - Its definition of "noisy" is based on environmental sounds. It is trained to classify audio with only source artifacts (like microphone hum or pure static) as `clean`. The easiest way to use this model is with a `pipeline`. > Note: The model outputs a confidence score for each label. In my use case, I consider audio to be clean if the score for the `clean` label is greater than 0.7. Training Data This model was trained on a sophisticated, custom-built dataset of ~55,000 audio clips, specifically designed to teach the nuances of audio quality. This whisper model was trained 2x faster with Unsloth and Huggingface's TRL library.

Arabic Diacritization Model | نَمُوذَجُ تَشْكِيلِ النُّصُوصِ الْعَرَبِيَّةِ نموذج بحجم 350 مليون بارامتر مخصص لتشكيل النصوص العربية. تم تدريب هذا النموذج بضبط نموذج - النموذج الأساسي: ibm-granite/granite-4.0-h-350m - مجموعة البيانات: Misraj/SadeedTashkeela مثال النص المدخل: `السلام عليكم` الناتج: `اَلسَلَامُ عَلَيْكُمْ` --- --- A 350M parameter model for Arabic diacritization (Tashkeel). This model is a fine-tune of `ibm-granite/granite-4.0-h-350m` on the `Misraj/SadeedTashkeela` dataset. - Base Model: ibm-granite/granite-4.0-h-350m - Dataset: Misraj/SadeedTashkeela How to Use The Python code for usage is the same as listed in the Arabic section above. Example Input: `السلام عليكم` Output: `اَلسَلَامُ عَلَيْكُمْ` This lfm2 model was trained 2x faster with Unsloth and Huggingface's TRL library.

- Developed by: Etherll - License: apache-2.0 - Finetuned from model : unsloth/whisper-small This whisper model was trained 2x faster with Unsloth and Huggingface's TRL library.

Arabic Diacritization Model | نَمُوذِجٌ تَشْكِيلُ النُّصُوصِ الْعَرَبِيَّةِ نموذج بحجم 700 مليون بارامتر مخصص لتشكيل النصوص العربية. تم تدريب هذا النموذج بضبط نموذج - النموذج الأساسي: LiquidAI/LFM2-700M - مجموعة البيانات: arbml/tashkeela مثال النص المدخل: `السلام عليكم` الناتج: `السَّلَامُ عَلَيْكُمْ` A 700M parameter model for Arabic diacritization (Tashkeel). This model is a fine-tune of `LiquidAI/LFM2-700M` on the `arbml/tashkeela` dataset. - Base Model: LiquidAI/LFM2-700M - Dataset: arbml/tashkeela How to Use The Python code for usage is the same as listed in the Arabic section above. Example Input: `السلام عليكم` Output: `السَّلَامُ عَلَيْكُمْ` This lfm2 model was trained 2x faster with Unsloth and Huggingface's TRL library.

Base model: NousResearch Hermes 3 Llama 3.1 8B, Replete-AI Replete LLM V2 Llama 3.1 8B.

Base model: Qwen 2.5 Coder 7B Instruct, Qwen 2.5 Coder 7B.

Base model: rombodawg/Meta Llama 3.1 8B reuploaded, Joseph717171/Llama 3.1 SuperNova 8B Lite TIES with Base.

Base model: rombodawg/Meta-Llama-3.1-8B-reuploaded, Etherll/Herplete-LLM-Llama-3.1-8b.

Etherll/Mellum-4b-sft-rust is a large language model (LLM) fine-tuned specifically for Rust code Fill-in-the-Middle (FIM) tasks. It is built upon `JetBrains/Mellum-4b-base` model. This model has been fine-tuned on the `Etherll/CodeFIM-Rust-Mellum` dataset, which comprises approximately 57,000 Rust-specific FIM examples, to enhance its proficiency in completing Rust code snippets accurately and contextually. A GGUF version for CPU inference is also available: Etherll/Mellum-4b-sft-rust-GGUF. This model leverages the LLaMA-style architecture of `Mellum-4b-base` (4 billion parameters) and its extensive pre-training on over 4 trillion tokens. The fine-tuning process focused on adapting the model to the nuances of Rust syntax and common coding patterns for FIM tasks. Key Features: Specialized for Rust: Optimized for Fill-in-the-Middle tasks in Rust. Based on Mellum-4b-base: Benefits from JetBrains' robust base model. Efficient: Suitable for both cloud and local deployment. IDE Integration Ready: Designed for use in developer tooling, and works particularly well with Continue.dev for an enhanced coding assistant experience. Fine-tuning Data Dataset: `Etherll/CodeFIM-Rust-Mellum` Size: ~57,000 rows Focus: Rust code Fill-in-the-Middle This model is trained to recognize a specific format for Fill-in-the-Middle tasks. When providing input for FIM, please use the following structure: For the best integrated development experience, it's highly recommended to use this model with Continue.dev. Refer to the Continue.dev documentation for instructions on how to add custom LLMs. A GGUF version is available at Etherll/Mellum-4b-sft-rust-GGUF. This format is suitable for local inference on CPU (and GPU with appropriate llama.cpp/Ollama builds) using tools like: llama.cpp Ollama LM Studio Support & Community If you need any help, have questions, or just want to chat, feel free to message me on Discord: etherl

library_name: transformers tags: mergekit