angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k
Viewer • Updated • 38.5k • 921 • 45
Aptronym
Aptronym
AI & ML interests
Emergent Gaming
Recent Activity
liked a dataset about 8 hours ago
angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k liked a dataset 12 days ago
storytracer/LoC-PD-Books liked a dataset 12 days ago
storytracer/US-PD-BooksOrganizations
None yet
liked a
dataset about 8 hours ago
liked 3
datasets 12 days ago
liked a
dataset 24 days ago
reacted to eaddario's post with 👍 25 days ago
Post
179
Experimental global target bits‑per‑weight quantization of Qwen/Qwen3.5-4B and Qwen/Qwen3.5-9B
Unlike standard llama.cpp quantizations that rely on fixed type heuristics (e.g., Q4_K_M), the Target BPW approach optimizes per-tensor precision where it matters the most, and produces high quality models that meet a precise global file size target.
Key Advantages:
- VRAM Maximization: Can generate high quality models sized exactly to fit hardware constraints (e.g., fitting the model into exactly 24GB VRAM).
- Data-Driven Precision: Quantization mix is determined by actual weight error sensitivity rather than hardcoded rules, often yielding better PPL/KLD size trade-offs.
Full benchmarks (PPL, KLD, ARC, MMLU, etc.) and methodology in the models' cards
eaddario/Qwen3.5-4B-GGUF
eaddario/Qwen3.5-9B-GGUF
Unlike standard llama.cpp quantizations that rely on fixed type heuristics (e.g., Q4_K_M), the Target BPW approach optimizes per-tensor precision where it matters the most, and produces high quality models that meet a precise global file size target.
Key Advantages:
- VRAM Maximization: Can generate high quality models sized exactly to fit hardware constraints (e.g., fitting the model into exactly 24GB VRAM).
- Data-Driven Precision: Quantization mix is determined by actual weight error sensitivity rather than hardcoded rules, often yielding better PPL/KLD size trade-offs.
Full benchmarks (PPL, KLD, ARC, MMLU, etc.) and methodology in the models' cards
eaddario/Qwen3.5-4B-GGUF
eaddario/Qwen3.5-9B-GGUF
- 4 replies
liked a
dataset 25 days ago
Is 40% possible?
#4 opened 29 days ago
by Aptronym
liked a
model about 1 month ago
liked 2
datasets about 1 month ago
Problem w/ tokenizer_config.json
#2 opened about 1 month ago
by Aptronym
liked 5
datasets about 1 month ago
Crownelius/Opus-4.6-Reasoning-3300x
Viewer • Updated • 2.16k • 3.3k • 295
Crownelius/High-Coder-SFT-Medium
Preview • Updated • 123 • 10
Crownelius/Crow-8B-Training-Data
Viewer • Updated • 190k • 68 • 1
Crownelius/Creative-Writing-Reasoning-KimiK2.5-600x
Viewer • Updated • 655 • 1.03k • 5
reedmayhew/LittleMonster-reasoning
Viewer • Updated • 416 • 15 • 1
liked a
model about 1 month ago
liked a
dataset about 1 month ago