powerpudu/Web-Petals-SmolLM2-1.7B

Web-Petals SmolLM2-1.7B ONNX Layers

SmolLM2-1.7B-Instruct split into individual ONNX transformer layer files for distributed P2P inference.

Files

• embeds.onnx - Token embedding table (~384 MB)
• layer_0.onnx ... layer_23.onnx - 24 transformer blocks (~256 MB each)
• lm_head.onnx - Final LayerNorm + LM head (~384 MB)

Architecture

• Hidden size: 2048
• Layers: 24
• Attention heads: 32
• FFN size: 8192
• Vocab size: 49152
• Total: ~6.75 GB (FP32)

License

Apache 2.0 (same as SmolLM2)