code wiki / (root) / nx_f32_llm_load.nx

nx_f32_llm_load.nx

buildroot/runtime/nx_f32_llm_load.nx

5121 B129 linesdepth 11pulls 33 transitivereach 4 importersview sourcekind librarytopic f32
docsdependenciesstructsconstsfunctions

about

nx_f32_llm_load.nx -- model-level GGUF binder. Populates an NxF32LlamaModel from a parsed GGUF: - 3 top-level tensors: token_embd.weight -> embed_weights output_norm.weight -> gamma_out output.weight -> lm_head (tied to embed if missing) - N per-layer NxF32LlamaLayer structs via nx_f32_llama_layer_load_from_gguf Caller responsibility: pre-fill model dimension fields (n_layers, hidden_dim, n_heads, n_kv_heads, head_dim, ffn_dim, vocab_size). v1 does not parse GGUF metadata to derive them -- a separate metadata-walker brick is queued for that. This keeps the binder focused on tensor binding only. Tied-embed detection: if "output.weight" is not found, lm_head points to the same storage as embed_weights (the canonical Llama memory-saving trick used by Qwen, Llama-3, etc.). Caller must not free either pointer independently. genealogy_id: llama_gguf_naming_gerganov_2024 + standard_model_binding + tied_embed_qwen_llama3 lineage_id: substrate_f32_llm_load_v1

dependencies 7 imports · 4 importers

nx_syscalls.nx nx_tier.nx nx_gguf.nx nx_gguf_load_f32.nx nx_f32_llama_block.nx nx_f32_llama_layer_load.nx nx_f32_llm.nx nx_f32_llm_load.nx nx_f32_llm_e2e_test.nx nx_f32_llm_load_test.nx nx_f32_llm_run_test.nx nx_f32_llm_run_v2_test.nx

imports: nx_syscalls.nxnx_tier.nxnx_gguf.nxnx_gguf_load_f32.nxnx_f32_llama_block.nxnx_f32_llama_layer_load.nxnx_f32_llm.nx

imported by: nx_f32_llm_e2e_test.nxnx_f32_llm_load_test.nxnx_f32_llm_run_test.nxnx_f32_llm_run_v2_test.nx

structs

none

consts

33const NX_FLM_OK: nx_int = 0
34const NX_FLM_ERR_NULL: nx_int = 1
35const NX_FLM_ERR_BAD_DIM: nx_int = 2
36const NX_FLM_ERR_NOT_FOUND: nx_int = 3
37const NX_FLM_ERR_LAYER: nx_int = 4
38const NX_FLM_N_VERDICTS: nx_int = 5

functions

40func nx_flm_verdict_is_valid(v: nx_int) -> nx_int
called by 1: main
52func nx_f32_llm_load_weights_from_gguf(buf: *u8, hdr: *NxGgufHeader,