-
AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
Paper • 2306.00978 • Published • 17 -
GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers
Paper • 2210.17323 • Published • 13 -
LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale
Paper • 2208.07339 • Published • 6 -
QLoRA: Efficient Finetuning of Quantized LLMs
Paper • 2305.14314 • Published • 65
Guilherme Monteiro
guicybercode
AI & ML interests
smol-llms, llm-training, chinese-history, brazilian-history, production-ml, terminal-ui, data-curation, edge-ai
Recent Activity
commentedon a paper about 3 hours ago
Same Bytes, Different Authority: Reserved-Token Representations in Chat-Template Prompt Injection updated a dataset about 4 hours ago
guicybercode/cybercode-bench-ptbr commentedon a paper about 4 hours ago
Specter: Fast MoE Inference via Quantization-Enabled Multi-Step Speculative OffloadingOrganizations
Key papers on small LLMs, data-centric training, and multilingual NLP.
-
The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale
Paper • 2406.17557 • Published • 108 -
SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model
Paper • 2502.02737 • Published • 261 -
The Common Pile v0.1: An 8TB Dataset of Public Domain and Openly Licensed Text
Paper • 2506.05209 • Published • 66
Multilingual models featuring Portuguese, Chinese, and other languages essential for cross-cultural historical research. Curated for fine-tuning on Ch
Datasets and an interactive explorer connecting Japan, Iceland, mathematics, technology, philosophy, Asian culture and Christian ethics.
Resources, datasets and base models for training small LLMs on consumer hardware. Focus on pre-training pipelines and data curation.
Small language models (≤3B params) that can be fine-tuned or run locally for history-focused tasks. Curated for training on Chinese and Brazilian hist
-
HuggingFaceTB/SmolLM2-135M
Text Generation • 0.1B • Updated • 1.75M • 256 -
HuggingFaceTB/SmolLM2-135M-Instruct
Text Generation • 0.1B • Updated • 1.25M • 456 -
HuggingFaceTB/SmolLM2-360M
Text Generation • 0.4B • Updated • 179k • 143 -
HuggingFaceTB/SmolLM2-360M-Instruct
Text Generation • 0.4B • Updated • 206k • 234
-
AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
Paper • 2306.00978 • Published • 17 -
GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers
Paper • 2210.17323 • Published • 13 -
LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale
Paper • 2208.07339 • Published • 6 -
QLoRA: Efficient Finetuning of Quantized LLMs
Paper • 2305.14314 • Published • 65
Datasets and an interactive explorer connecting Japan, Iceland, mathematics, technology, philosophy, Asian culture and Christian ethics.
Key papers on small LLMs, data-centric training, and multilingual NLP.
-
The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale
Paper • 2406.17557 • Published • 108 -
SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model
Paper • 2502.02737 • Published • 261 -
The Common Pile v0.1: An 8TB Dataset of Public Domain and Openly Licensed Text
Paper • 2506.05209 • Published • 66
Resources, datasets and base models for training small LLMs on consumer hardware. Focus on pre-training pipelines and data curation.
Multilingual models featuring Portuguese, Chinese, and other languages essential for cross-cultural historical research. Curated for fine-tuning on Ch
Small language models (≤3B params) that can be fine-tuned or run locally for history-focused tasks. Curated for training on Chinese and Brazilian hist
-
HuggingFaceTB/SmolLM2-135M
Text Generation • 0.1B • Updated • 1.75M • 256 -
HuggingFaceTB/SmolLM2-135M-Instruct
Text Generation • 0.1B • Updated • 1.25M • 456 -
HuggingFaceTB/SmolLM2-360M
Text Generation • 0.4B • Updated • 179k • 143 -
HuggingFaceTB/SmolLM2-360M-Instruct
Text Generation • 0.4B • Updated • 206k • 234