nx_f32_llm_embed_diag_test.nx
buildroot/runtime/nx_f32_llm_embed_diag_test.nx
about
nx_f32_llm_embed_diag_test.nx -- check Q5_0-dequanted embed values
on real Qwen2.5. Reports the max absolute value (as f32 bit pattern
truncated to lower 8 bits via exit code).
Token embeddings in well-trained models are typically in [-1, 1].
Huge values (>10) would indicate dequant is wrong.
dependencies 12 imports · 0 importers
diagram shows first 10 each side; +2 more imports, +0 more importers in the complete lists below.
imports: nx_syscalls.nxnx_tier.nxnx_gguf.nxnx_gguf_load.nxnx_f32.nxnx_f32_lazy_weight.nxnx_f32_llama_block.nxnx_f32_llama_block_v4.nxnx_f32_llama_layer_lazy_load.nxnx_f32_llm.nxnx_f32_llm_v4.nxnx_f32_llm_read_dims.nx
imported by: nobody (leaf or entry point)
call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown
structs
| none |
consts
| none |
functions
| 21 | func main() -> i64 |