veyra-ai/Veyra2-Mango-15M-Base-ONNX
Text Generation • Updated • 1
🔄 In a Training Loop
Jdudeo - Veyra AI
Jdudeo
AI & ML interests
None yet
Recent Activity
liked a model about 5 hours ago
veyra-ai/Veyra2-Mango-15M-Base-ONNX liked a model about 5 hours ago
veyra-ai/Veyra2-Mango-30M-Base-ONNX liked a model about 5 hours ago
SupraLabs/Supra2-NanoOrganizations
liked 3
models about 5 hours ago
reacted to Banaxi-Tech's post with 🔥 about 23 hours ago
Post
990
Today we wanted to release BananaMind 2 Pico, our smallest model yet at ~0.9M parameters. Instead, we accidentally ran a very expensive experiment on what happens when you push a tiny model way past its useful token budget.
Short version: we trained on 200B tokens (~222K:1 tokens-per-parameter). The model peaked at 20B tokens with an INT Index of 4.55, then degraded monotonically over the next 160B to 3.31 — a 27% regression. Three of four Open SLM benchmarks were worse at the end of training than they were at 10% through.
The useful compute-optimal range for Pico-tier models looks like ~22K–30K tokens per parameter. Ratios like 7K:1, 15K:1, and 22K:1 all work fine — TinyStories and most sub-3M community models sit in this range. Push much further and benchmarks start rotting.
Follow us for more:
BananaMind
@vovaRL
@Banaxi-Tech
Full writeup with all checkpoints, the Chinchilla-ratio control run, and the schedule-vs-overtraining analysis: https://huggingface.co/blog/Banaxi-Tech/ovdadadadd
And if anyone, i dont know the reason why you would, wants the 20B token checkpoint reply and ill upload it as BananaMind 2.1 Pico EXP
Short version: we trained on 200B tokens (~222K:1 tokens-per-parameter). The model peaked at 20B tokens with an INT Index of 4.55, then degraded monotonically over the next 160B to 3.31 — a 27% regression. Three of four Open SLM benchmarks were worse at the end of training than they were at 10% through.
The useful compute-optimal range for Pico-tier models looks like ~22K–30K tokens per parameter. Ratios like 7K:1, 15K:1, and 22K:1 all work fine — TinyStories and most sub-3M community models sit in this range. Push much further and benchmarks start rotting.
Follow us for more:
@vovaRL
@Banaxi-Tech
Full writeup with all checkpoints, the Chinchilla-ratio control run, and the schedule-vs-overtraining analysis: https://huggingface.co/blog/Banaxi-Tech/ovdadadadd
And if anyone, i dont know the reason why you would, wants the 20B token checkpoint reply and ill upload it as BananaMind 2.1 Pico EXP
- 2 replies
upvoted an article about 23 hours ago
Article
Extreme Overtraining in Tiny Language Models
Banaxi-Tech
• • 8 liked a
model 1 day ago
updated 4
models 1 day ago
updated a
model 2 days ago
liked a
model 2 days ago
reacted to danielhanchen's post with 🔥 2 days ago
Post
3956
Introducing Unsloth Desktop 🦥
The first desktop app to run and train models locally.
• Open-source. Runs on Mac, Windows and Linux
• Supports MLX, diffusion image/video, audio, GGUF
• Connect Claude Code and Codex to local LLMs
• 50% more accurate, self-healing tool calls + sandboxed code exec
• Works for CPU + multiGPU setups - NVIDIA, AMD, Intel, Mac
• Train models 2× faster with 70% less VRAM
• Private web search, deep research, RAG, MCP and exports (NVFP4, GGUF)
• Use Unsloth’s OpenAI-compatible API and cloud models
• Securely deploy LLMs remotely and access anywhere
Unsloth Desktop is now available on http://unsloth.ai
and GitHub.
GitHub: https://github.com/unslothai/unsloth
Blog and Guide: https://unsloth.ai/docs/desktop
The first desktop app to run and train models locally.
• Open-source. Runs on Mac, Windows and Linux
• Supports MLX, diffusion image/video, audio, GGUF
• Connect Claude Code and Codex to local LLMs
• 50% more accurate, self-healing tool calls + sandboxed code exec
• Works for CPU + multiGPU setups - NVIDIA, AMD, Intel, Mac
• Train models 2× faster with 70% less VRAM
• Private web search, deep research, RAG, MCP and exports (NVFP4, GGUF)
• Use Unsloth’s OpenAI-compatible API and cloud models
• Securely deploy LLMs remotely and access anywhere
Unsloth Desktop is now available on http://unsloth.ai
and GitHub.
GitHub: https://github.com/unslothai/unsloth
Blog and Guide: https://unsloth.ai/docs/desktop
- 6 replies
reacted to LH-Tech-AI's post with 🚀🔥 2 days ago
Post
2071
Hey community and sponsors!
We are announcing the Supra3 family with four core SLM models:
- Supra3 Flash Lite: 25M parameters, ~60B pretraining tokens
- Supra3 Flash: 50M parameters, ~100B pretraining tokens
- Supra3 Pro: 75M parameters, ~150B pretraining tokens
- Supra3 Ultra: 100M parameters, ~200B pretraining tokens
For Supra3 Pro and Ultra, we search for sponsors who give us free access to compute like RTX 5090 32GB or so.
We estimate the total cost of the pro and ultra models at around $600.
For Supra3 Flash Lite and Flash, we do not need sponsors.
If anyone would apply for helping us, we would be really thankful and this person would get early access to new modele, insider information, credit and more!
Contact: here or on discord: lh_tech_ai
We are announcing the Supra3 family with four core SLM models:
- Supra3 Flash Lite: 25M parameters, ~60B pretraining tokens
- Supra3 Flash: 50M parameters, ~100B pretraining tokens
- Supra3 Pro: 75M parameters, ~150B pretraining tokens
- Supra3 Ultra: 100M parameters, ~200B pretraining tokens
For Supra3 Pro and Ultra, we search for sponsors who give us free access to compute like RTX 5090 32GB or so.
We estimate the total cost of the pro and ultra models at around $600.
For Supra3 Flash Lite and Flash, we do not need sponsors.
If anyone would apply for helping us, we would be really thankful and this person would get early access to new modele, insider information, credit and more!
Contact: here or on discord: lh_tech_ai
- 14 replies
Add Veyra2 Mango
1
#82 opened 2 days ago
by Jdudeo
updated a
collection 2 days ago
published a
model 2 days ago
liked a
model 2 days ago
