apr convert ~/Downloads/qwen2.5-coder-1.5b.safetensors --output qwen2.5-coder-1.5b-instruct.apr
=== APR Convert ===

Input:  /home/alfredo/Downloads/qwen2.5-coder-1.5b.safetensors
Output: qwen2.5-coder-1.5b-instruct.apr
Quantization: None (copy)

Converting...

=== Conversion Report ===

Original size:  5.75 GiB
Converted size: 5.75 GiB
Tensors:        338
Reduction:      -0.0% (1.00x)

⚠ Conversion completed (output larger than input)

=== Model Chat (APR Format) ===

Using APR v2 format with mmap (Native Library Mandate)

  Model: qwen2.5-coder-1.5b-instruct.apr
  Chat Template: ChatML
  Temperature: 0.7
  Top-P: 0.9
  Max Tokens: 512

Commands:
  /quit     Exit the chat
  /clear    Clear conversation history
  /system   Set system prompt
  /help     Show help

════════════════════════════════════════════════════════════

Loading model...
Loaded SafeTensors format in 0.92s (6174.9 MB)
Loaded tokenizer: tokenizer.json (151936 tokens)
Loaded 219 tensors from SafeTensors
Detected ChatML chat template
You: hey

thread 'main' (3360593) panicked at src/autograd/ops.rs:539:9:
assertion `left == right` failed: matmul dimension mismatch: 896 vs 1536
  left: 896
 right: 1536
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace