Diffusion Language Models combining deep narrow networks, Canon layers (depthwise causal convolutions), and WSD (Warmup-Stable-Decay) training.
🔄 In a Training Loop
Asankhaya Sharma
codelion
AI & ML interests
Creator of OptiLLM, OpenEvolve, Adaptive Classifier, and Ellora. Pioneering a new category in AI infrastructure: inference-time compute for LLMs.
Recent Activity
updated a model about 1 hour ago
mlx-community/Muse-Glimmer-30B-OptiQ-4bit updated a model about 2 hours ago
mlx-community/gemma-4-31B-it-qat-OptiQ-4bit updated a model about 4 hours ago
mlx-community/Fara1.5-27B-OptiQ-4bitOrganizations
A collection of really small language models pre-trained from scratch with open-data. Ideal for use in experimentation and evaluations.
Ellora: Enhancing LLMs with LoRA - Standardized Recipes for Capability Enhancement
-
codelion/Qwen3-0.6B-accuracy-recovery-lora
Text Generation • Updated • 317 • 5 -
codelion/Qwen3-0.6B-magpie
Viewer • Updated • 735 • 11 • 2 -
codelion/gemma-3-1b-it-reasoning-grpo-lora
Text Generation • Updated • 24 • 4 -
codelion/gemma-3-1b-it-magpie-reasoning
Viewer • Updated • 131 • 18 • 2
Internal Coherence Maximization (ICM): A Label-Free, Unsupervised Training Framework for LLMs
High-quality synthetic educational datasets designed for LLM pretraining with structured pedagogical content across 9 knowledge domains.
A collection of pre-training datasets samples of sizes 10M, 100M and 1B tokens. Ideal for use in quick experimentation and ablations.
Pivotal Token Search (PTS) identifies tokens in a language model's generation that significantly impact the probability of success
-
PTS Visualizer
🔬15Visualize token impacts and reasoning graphs from model data
-
codelion/Qwen3-0.6B-pts
Viewer • Updated • 1.38k • 34 • 2 -
codelion/DeepSeek-R1-Distill-Qwen-1.5B-pts
Preview • Updated • 25 • 1 -
codelion/Qwen3-0.6B-pts-thought-anchors
Viewer • Updated • 148 • 27 • 2
All models, datasets and tools related to https://securade.ai
Diffusion Language Models combining deep narrow networks, Canon layers (depthwise causal convolutions), and WSD (Warmup-Stable-Decay) training.
High-quality synthetic educational datasets designed for LLM pretraining with structured pedagogical content across 9 knowledge domains.
A collection of really small language models pre-trained from scratch with open-data. Ideal for use in experimentation and evaluations.
A collection of pre-training datasets samples of sizes 10M, 100M and 1B tokens. Ideal for use in quick experimentation and ablations.
Ellora: Enhancing LLMs with LoRA - Standardized Recipes for Capability Enhancement
-
codelion/Qwen3-0.6B-accuracy-recovery-lora
Text Generation • Updated • 317 • 5 -
codelion/Qwen3-0.6B-magpie
Viewer • Updated • 735 • 11 • 2 -
codelion/gemma-3-1b-it-reasoning-grpo-lora
Text Generation • Updated • 24 • 4 -
codelion/gemma-3-1b-it-magpie-reasoning
Viewer • Updated • 131 • 18 • 2
Pivotal Token Search (PTS) identifies tokens in a language model's generation that significantly impact the probability of success
-
PTS Visualizer
🔬15Visualize token impacts and reasoning graphs from model data
-
codelion/Qwen3-0.6B-pts
Viewer • Updated • 1.38k • 34 • 2 -
codelion/DeepSeek-R1-Distill-Qwen-1.5B-pts
Preview • Updated • 25 • 1 -
codelion/Qwen3-0.6B-pts-thought-anchors
Viewer • Updated • 148 • 27 • 2
Internal Coherence Maximization (ICM): A Label-Free, Unsupervised Training Framework for LLMs
All models, datasets and tools related to https://securade.ai