singulariti/Kimi-K3-bucket
1.56 TB
Leo A
singulariti
AI & ML interests
None yet
Recent Activity
updated a bucket about 9 hours ago
singulariti/Kimi-K3-bucket published a bucket about 9 hours ago
singulariti/Kimi-K3-bucket liked a model 1 day ago
moonshotai/Kimi-K3Organizations
None yet
updated a
bucket about 9 hours ago
published a
bucket about 9 hours ago
liked a
model 1 day ago
reacted to Quazim0t0's post with β€οΈ 15 days ago
Post
5034
πΌ DaisyChain-Web: train a language model with friends or by yourself with multiple devices, in the browser, no install
Open a webpage, share a room link, and every device that joins becomes part of the training cluster. Phones, laptops, old PCs: they connect peer-to-peer over WebRTC and train one shared transformer together, entirely in the browser.
What's actually happening under the hood:
π§ A mini transformer LM trains on FineWeb-Edu, streamed live from the HuggingFace Hub. Each device pulls its own slice (data parallelism), tokenized with our 16.5k-token Spikewhale tokenizer
β‘ Every single multiply runs through verified INT8 neural units, no float fallback. On WebGPU browsers it uses the GPU's DP4A integer dot-product hardware, admitted only after proving bit-identical results against the verified units, with a 3ΓINT8 fast-accurate scheme (CUTLASS's 3xTF32 trick, ported to 8-bit)
π Devices average gradients every step under a sync guard: a per-step roster protocol plus weight-hash verification keeps every device's model bit-identical. If anything drifts, training stops instead of silently forking
π Live logs show exactly what every device contributes, step by step
πΎ When you're done: test generations right on the page, download a checkpoint, or grab the inference kit, a single self-contained HTML file with the weights baked in that runs generations offline, anywhere
Works solo too. Every extra device just grows the effective batch.
π Try it: Quazim0t0/DaisyChain-Web
π Training framework: DaisyChainAI/DaisyChain-Train
Updates:
- Block-scaled INT8 quantization
- Batched attention GEMM
- Fused dequant+ReLU epilogue
- Weight-tied unembedding
- WebSocket relay fallback
- Server keepalive ping/pong every 30s
- disconnected-state redial
- Visibility/network-change reconnect
Open a webpage, share a room link, and every device that joins becomes part of the training cluster. Phones, laptops, old PCs: they connect peer-to-peer over WebRTC and train one shared transformer together, entirely in the browser.
What's actually happening under the hood:
π§ A mini transformer LM trains on FineWeb-Edu, streamed live from the HuggingFace Hub. Each device pulls its own slice (data parallelism), tokenized with our 16.5k-token Spikewhale tokenizer
β‘ Every single multiply runs through verified INT8 neural units, no float fallback. On WebGPU browsers it uses the GPU's DP4A integer dot-product hardware, admitted only after proving bit-identical results against the verified units, with a 3ΓINT8 fast-accurate scheme (CUTLASS's 3xTF32 trick, ported to 8-bit)
π Devices average gradients every step under a sync guard: a per-step roster protocol plus weight-hash verification keeps every device's model bit-identical. If anything drifts, training stops instead of silently forking
π Live logs show exactly what every device contributes, step by step
πΎ When you're done: test generations right on the page, download a checkpoint, or grab the inference kit, a single self-contained HTML file with the weights baked in that runs generations offline, anywhere
Works solo too. Every extra device just grows the effective batch.
π Try it: Quazim0t0/DaisyChain-Web
π Training framework: DaisyChainAI/DaisyChain-Train
Updates:
- Block-scaled INT8 quantization
- Batched attention GEMM
- Fused dequant+ReLU epilogue
- Weight-tied unembedding
- WebSocket relay fallback
- Server keepalive ping/pong every 30s
- disconnected-state redial
- Visibility/network-change reconnect
- 11 replies
upvoted a collection about 1 month ago
liked 2
models about 1 month ago
upvoted a collection about 1 month ago
updated a
bucket about 1 month ago
published a
bucket about 1 month ago
V2? π
π 3
5
#2 opened about 2 months ago
by singulariti
minimax 3 δ»δΉζΆεεΌζΊοΌ
25
#33 opened about 2 months ago
by 
wpfnnnns
Add IQ3 for 96Gb VRAM
3
#4 opened about 2 months ago
by 
FBykov
updated a
bucket about 2 months ago
published a
bucket 2 months ago
liked a
model 3 months ago
reacted to HannesVonEssen's post with π₯ 3 months ago
Post
277
π£ I made a visualizer for Hugging Face models: https://hfviewer.com
β¨ Simply paste a Hugging Face URL to get an interactive visualization of the architecture!
π The recent Qwen3.6-27B model as an example: https://hfviewer.com/Qwen/Qwen3.6-27B
Feel free to try it out and give me feedback on how it can be improved! β€οΈ
β¨ Simply paste a Hugging Face URL to get an interactive visualization of the architecture!
π The recent Qwen3.6-27B model as an example: https://hfviewer.com/Qwen/Qwen3.6-27B
Feel free to try it out and give me feedback on how it can be improved! β€οΈ
- 1 reply
liked 2
models 3 months ago
