inferencelayer 0.2.3

Kortexya's engine-native inference layer — LLM generation + embedding/encoder family on wgpu (WGSL kernels, any adapter) with a pure-Rust CPU fallback
Documentation

Build #3968837 2026-07-27 17:34:43

# rustc version rustc 1.99.0-nightly (d0babd8b6 2026-07-15)

# docs.rs version docsrs 0.0.0 (f1b150f0cd76f7cabcd39e1f4790ba771b28b5c4 2026-07-19 )

# build log [INFO] running `Command { std: "docker" "exec" "-e" "SOURCE_DIR=/opt/rustwide/workdir" "-e" "CARGO_HOME=/opt/rustwide/cargo-home" "-e" "RUSTUP_HOME=/opt/rustwide/rustup-home" "-e" "CARGO_TARGET_DIR=/opt/rustwide/target" "-e" "DOCS_RS=1" "-w" "/opt/rustwide/workdir" "--user" "1001:1001" "ecacb71246d045f03be3c688626bb843e44bf57c44a679cedd7a2049eb28809c" "/opt/rustwide/cargo-home/bin/cargo" "+nightly-2026-07-16" "rustdoc" "--lib" "-Zrustdoc-map" "--config" "build.rustdocflags=[\"--cfg\", \"docsrs\", \"-Z\", \"unstable-options\", \"--emit=html-non-static-files\", \"--resource-suffix\", \"-20260715-1.99.0-nightly-d0babd8b6\", \"--static-root-path\", \"/-/rustdoc.static/\", \"--cap-lints\", \"warn\", \"--extern-html-root-takes-precedence\"]" "--offline" "-Zunstable-options" "--config=doc.extern-map.registries.crates-io=\"https://docs.rs/{pkg_name}/{version}/x86_64-unknown-linux-gnu\"" "-Zrustdoc-scrape-examples" "-j6" "--target" "x86_64-unknown-linux-gnu", kill_on_drop: false }` [INFO] [stderr] warning: target filter specified, but no targets matched; this is a no-op [INFO] [stderr] Documenting inferencelayer v0.2.3 (/opt/rustwide/workdir) [INFO] [stderr] warning: unresolved link to `cpu_gemm::gemm_packed` [INFO] [stderr] --> src/conv2d.rs:3:36 [INFO] [stderr] | [INFO] [stderr] 3 | //! One general path: **im2col → [`cpu_gemm::gemm_packed`]**, NHWC activations, single image [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^^ no item named `cpu_gemm` in scope [INFO] [stderr] | [INFO] [stderr] = note: `#[warn(rustdoc::broken_intra_doc_links)]` on by default [INFO] [stderr] [INFO] [stderr] warning: public documentation for `PackedWeight` links to private item `NR` [INFO] [stderr] --> src/cpu_gemm.rs:48:71 [INFO] [stderr] | [INFO] [stderr] 48 | /// A `[N, K]` row-major weight, packed once into K-major panels of [`NR`] columns. [INFO] [stderr] | ^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] = note: `#[warn(rustdoc::private_intra_doc_links)]` on by default [INFO] [stderr] [INFO] [stderr] warning: public documentation for `PackedWeight` links to private item `NR` [INFO] [stderr] --> src/cpu_gemm.rs:51:29 [INFO] [stderr] | [INFO] [stderr] 51 | /// zero-padded to a full [`NR`], so the microkernel never needs a ragged-N variant; the *store* [INFO] [stderr] | ^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `n,n` [INFO] [stderr] --> src/deepencoder_gpu.rs:7:27 [INFO] [stderr] | [INFO] [stderr] 7 | //! the V100 target — NO [n,n] materialization (naive dense needs an 805MB scores buffer per [INFO] [stderr] | ^^^ no item named `n,n` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `sam_attention_gpu` links to private item `crate::deepencoder::sam_attention` [INFO] [stderr] --> src/deepencoder_gpu.rs:538:41 [INFO] [stderr] | [INFO] [stderr] 538 | /// SAM attention on the GPU — iso to [`crate::deepencoder::sam_attention`]. Input `x` `[n, width]` [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `gelu` [INFO] [stderr] --> src/deepencoder_gpu.rs:640:67 [INFO] [stderr] | [INFO] [stderr] 640 | /// Full SAM **global** block on GPU (ln → attn → +res → ln → mlp[gelu] → +res), one readback. [INFO] [stderr] | ^^^^ no item named `gelu` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `layernorm_gpu` links to private item `crate::deepencoder::layernorm` [INFO] [stderr] --> src/deepencoder_gpu.rs:665:30 [INFO] [stderr] | [INFO] [stderr] 665 | /// GPU LayerNorm — iso to [`crate::deepencoder::layernorm`]. `[rows, c]` → `[rows, c]`. [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `linear_gpu` links to private item `crate::deepencoder::linear` [INFO] [stderr] --> src/deepencoder_gpu.rs:678:51 [INFO] [stderr] | [INFO] [stderr] 678 | /// GPU linear (public, for iso-tests) — iso to [`crate::deepencoder::linear`]. [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `enc_gemm2_src` links to private item `enc_gemm_src` [INFO] [stderr] --> src/encoder.rs:109:42 [INFO] [stderr] | [INFO] [stderr] 109 | /// **GEMM v2** — the same contract as [`enc_gemm_src`] (`y[m,n] = act(x[m,k] · Wᵀ + bias)`, same [INFO] [stderr] | ^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `ENC_GEMM3_SK_REDUCE` [INFO] [stderr] --> src/encoder.rs:803:51 [INFO] [stderr] | [INFO] [stderr] 803 | /// (`y_part[c][m,n]`, no bias, no activation); [`ENC_GEMM3_SK_REDUCE`] then folds the chunks in [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^ no item named `ENC_GEMM3_SK_REDUCE` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `gemm_bench` [INFO] [stderr] --> src/encoder.rs:2820:11 [INFO] [stderr] | [INFO] [stderr] 2820 | /// [`gemm_bench`]'s v3 twin — the STOP instrument for deciding whether the vec4/transposed [INFO] [stderr] | ^^^^^^^^^^ no item named `gemm_bench` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `gemm3_bench` [INFO] [stderr] --> src/encoder.rs:2881:11 [INFO] [stderr] | [INFO] [stderr] 2881 | /// [`gemm3_bench`]'s split-K twin: the partial+reduce PAIR, timed as the plan would run it [INFO] [stderr] | ^^^^^^^^^^^ no item named `gemm3_bench` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `crate::Gemv` [INFO] [stderr] --> src/forward.rs:5:39 [INFO] [stderr] | [INFO] [stderr] 5 | //! Kernels here beyond the GEMV in [`crate::Gemv`]: rmsnorm (one workgroup per vector), the gated [INFO] [stderr] | ^^^^^^^^^^^ no item named `Gemv` in module `inferencelayer` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `make_spec_plan` links to private item `SparsePick` [INFO] [stderr] --> src/forward.rs:10076:15 [INFO] [stderr] | [INFO] [stderr] 10076 | /// one [`SparsePick`] per column over the caller's grammar-mask buffer. Build once per [INFO] [stderr] | ^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `Self::read_stage_out_async` [INFO] [stderr] --> src/forward.rs:14363:53 [INFO] [stderr] | [INFO] [stderr] 14363 | /// `device.poll(Wait)`). The wasm sibling is [`Self::read_stage_out_async`]. [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^^^^^^^ the struct `Lfm2Gpu` has no field or associated item named `read_stage_out_async` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `quantize_q4_0` [INFO] [stderr] --> src/forward.rs:15579:73 [INFO] [stderr] | [INFO] [stderr] 15579 | /// vocab is clamped to the first `N` rows of the shared Q4 head ([`quantize_q4_0`] [INFO] [stderr] | ^^^^^^^^^^^^^ no item named `quantize_q4_0` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `draft_chain` links to private item `DraftRound` [INFO] [stderr] --> src/forward.rs:15662:47 [INFO] [stderr] | [INFO] [stderr] 15662 | /// ONE submit + one map per round (see [`DraftRound`]): per-step metas ride staged buffers [INFO] [stderr] | ^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `draft_chain` links to private item `Self::draft_chain_probe` [INFO] [stderr] --> src/forward.rs:15665:47 [INFO] [stderr] | [INFO] [stderr] 15665 | /// takes the old per-step-submit path ([`Self::draft_chain_probe`]) — measurement only. [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `draft_chain_round` links to private item `DraftRound` [INFO] [stderr] --> src/forward.rs:16454:84 [INFO] [stderr] | [INFO] [stderr] 16454 | /// The serving draft chain as ONE submission — the batched twin of the mono [`DraftRound`] [INFO] [stderr] | ^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `weights::Weights` [INFO] [stderr] --> src/gguf.rs:701:16 [INFO] [stderr] | [INFO] [stderr] 701 | /// engine's [`weights::Weights`], keeping the weights BINARY (packed into the `Q4` containers [INFO] [stderr] | ^^^^^^^^^^^^^^^^ no item named `weights` in scope [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `get_p1_at` [INFO] [stderr] --> src/gguf.rs:708:64 [INFO] [stderr] | [INFO] [stderr] 708 | /// `1+w` delta at conversion, unlike the safetensors path's [`get_p1_at`]). [INFO] [stderr] | ^^^^^^^^^ no item named `get_p1_at` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `weights::Weights` [INFO] [stderr] --> src/gguf.rs:919:7 [INFO] [stderr] | [INFO] [stderr] 919 | /// [`weights::Weights`], REPACKING every projection to `wdtype` (`Q4` or `F16`). [INFO] [stderr] | ^^^^^^^^^^^^^^^^ no item named `weights` in scope [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `elems_per_entry` [INFO] [stderr] --> src/iq_tables.rs:3:29 [INFO] [stderr] | [INFO] [stderr] 3 | //! Row layout: [n_entries][elems_per_entry] flattened; indices select whole rows. [INFO] [stderr] | ^^^^^^^^^^^^^^^ no item named `elems_per_entry` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `3,4,6,3` [INFO] [stderr] --> src/rtdetr.rs:14:15 [INFO] [stderr] | [INFO] [stderr] 14 | //! stages [3,4,6,3]; stride-2 lives in the 3×3 (`downsample_in_bottleneck=false`); stride-2 [INFO] [stderr] | ^^^^^^^ no item named `3,4,6,3` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `W,H,W,H` [INFO] [stderr] --> src/rtdetr.rs:32:57 [INFO] [stderr] | [INFO] [stderr] 32 | //! emit several labels), threshold 0.3, cxcywh→xyxy·[W,H,W,H], canonical docling label map [INFO] [stderr] | ^^^^^^^ no item named `W,H,W,H` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `0,1` [INFO] [stderr] --> src/tableformer.rs:78:29 [INFO] [stderr] | [INFO] [stderr] 78 | /// cxcywh, NORMALIZED [0,1] in CROP space, post-mergebboxes. [INFO] [stderr] | ^^^ no item named `0,1` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `k,n` [INFO] [stderr] --> src/rtdetr_gpu.rs:14:19 [INFO] [stderr] | [INFO] [stderr] 14 | //! transpose-to-[k,n], the same from-cpu construction whisper_gpu uses. The CPU path is the [INFO] [stderr] | ^^^ no item named `k,n` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `parakeet_gpu` links to private item `REL_POS_ATTN` [INFO] [stderr] --> src/parakeet_gpu.rs:6:40 [INFO] [stderr] | [INFO] [stderr] 6 | //! BERT/DeBERTa do not: this file's [`REL_POS_ATTN`] (Transformer-XL relative-position attention) [INFO] [stderr] | ^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `relpos_attention_gpu` links to private item `REL_POS_ATTN` [INFO] [stderr] --> src/parakeet_gpu.rs:610:69 [INFO] [stderr] | [INFO] [stderr] 610 | /// GPU rel-pos attention: uploads `q,k,v,p` and the biases, runs [`REL_POS_ATTN`], reads back the [INFO] [stderr] | ^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `vision_glm_gpu` links to private item `crate::deepencoder_gpu::flash_reg_src` [INFO] [stderr] --> src/vision_glm_gpu.rs:7:8 [INFO] [stderr] | [INFO] [stderr] 7 | //! ([`crate::deepencoder_gpu::flash_reg_src`], 4.74× vs naive on Vulkan, hd=64 = GLM's head dim), [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `PocketBackboneGpu` links to private item `TrShape` [INFO] [stderr] --> src/pocket_tts_gpu.rs:311:63 [INFO] [stderr] | [INFO] [stderr] 311 | /// Two differences from the codec's stack, both handled by [`TrShape`]: there is **no LayerScale** [INFO] [stderr] | ^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `2,2` [INFO] [stderr] --> src/qwen3tts.rs:94:28 [INFO] [stderr] | [INFO] [stderr] 94 | /// upsampling_ratios [2,2] (ConvTr+ConvNeXt, 12.5→50 Hz) then SEANet blocks ×[8,5,4,3]. [INFO] [stderr] | ^^^ no item named `2,2` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `8,5,4,3` [INFO] [stderr] --> src/qwen3tts.rs:94:84 [INFO] [stderr] | [INFO] [stderr] 94 | /// upsampling_ratios [2,2] (ConvTr+ConvNeXt, 12.5→50 Hz) then SEANet blocks ×[8,5,4,3]. [INFO] [stderr] | ^^^^^^^ no item named `8,5,4,3` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `load` links to private item `MAX_KV` [INFO] [stderr] --> src/qwen3tts_gpu.rs:335:73 [INFO] [stderr] | [INFO] [stderr] 335 | /// the CPU stack reads). `max_pos` bounds the KV cache length (≤ [`MAX_KV`]); `max_m` bounds the [INFO] [stderr] | ^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `step` links to private item `StepPlan` [INFO] [stderr] --> src/qwen3tts_gpu.rs:466:74 [INFO] [stderr] | [INFO] [stderr] 466 | /// length), return its final-norm hidden `[h]`. Uses the resident [`StepPlan`] (built once on [INFO] [stderr] | ^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `vision` links to private item `patchify` [INFO] [stderr] --> src/vision.rs:20:10 [INFO] [stderr] | [INFO] [stderr] 20 | //! [`patchify`]. This is the *whole* reason the merger may simply concatenate 4 CONSECUTIVE [INFO] [stderr] | ^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `vision` links to private item `resample_taps` [INFO] [stderr] --> src/vision.rs:34:10 [INFO] [stderr] | [INFO] [stderr] 34 | //! [`resample_taps`]. [INFO] [stderr] | ^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `resize_rgb8_bilinear` links to private item `resize_rgb8` [INFO] [stderr] --> src/vision.rs:364:28 [INFO] [stderr] | [INFO] [stderr] 364 | /// PIL-BILINEAR twin of [`resize_rgb8`] — the `RTDetrImageProcessor` resize (`resample=2`). [INFO] [stderr] | ^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `vision_gpu` links to private item `vis_attn_src` [INFO] [stderr] --> src/vision_gpu.rs:31:52 [INFO] [stderr] | [INFO] [stderr] 31 | //! 2. **Attention: no shared-memory staging** — [`vis_attn_src`]. It was 59% of the tower. TWO [INFO] [stderr] | ^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `vision_gpu` links to private item `vis_attn_coop_src` [INFO] [stderr] --> src/vision_gpu.rs:35:50 [INFO] [stderr] | [INFO] [stderr] 35 | //! 3. **Attention on the matrix units too** — [`vis_attn_coop_src`], 40 ms -> ~17. Its Q.K^T and P.V [INFO] [stderr] | ^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `flat_step_byte` [INFO] [stderr] --> src/grammar/gpu.rs:1:67 [INFO] [stderr] | [INFO] [stderr] 1 | //! GPU-resident JSON-schema grammar: the WGSL transcription of [`flat_step_byte`] + a wgpu harness. [INFO] [stderr] | ^^^^^^^^^^^^^^ no item named `flat_step_byte` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `gpu` links to private item `super::flat_ref::flat_step_byte` [INFO] [stderr] --> src/grammar/gpu.rs:4:8 [INFO] [stderr] | [INFO] [stderr] 4 | //! ([`super::flat_ref::flat_step_byte`]): identical 64-word state, identical `(kind, phase)` [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `inferencelayer::SpecGrammar` [INFO] [stderr] --> src/grammar/harness.rs:14:58 [INFO] [stderr] | [INFO] [stderr] 14 | //! `make_mask_bg`/`make_advance_bg`) and producing an [`inferencelayer::SpecGrammar`] for the speculative [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^^^^^^^^ no item named `inferencelayer` in scope [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `FsmState` [INFO] [stderr] --> src/grammar/harness.rs:26:9 [INFO] [stderr] | [INFO] [stderr] 26 | //! [`FsmState`] into `var<workgroup> sh_state: array<u32,64>` ONCE per workgroup (`if lid.x < 64u`, [INFO] [stderr] | ^^^^^^^^ no item named `FsmState` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `FsmState` [INFO] [stderr] --> src/grammar/harness.rs:29:13 [INFO] [stderr] | [INFO] [stderr] 29 | //! the [`FsmState`]/table port of the OSFQL `mask_main` at gpu_grammar.rs:711-751. [INFO] [stderr] | ^^^^^^^^ no item named `FsmState` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `inferencelayer::SpecGrammar` [INFO] [stderr] --> src/grammar/harness.rs:32:29 [INFO] [stderr] | [INFO] [stderr] 32 | //! `drafts[p]`; drives [`inferencelayer::SpecGrammar`]'s `snap_adv_bgs[p]` (state stride = 64·4 bytes). [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^^^^^^^^ no item named `inferencelayer` in scope [INFO] [stderr] [INFO] [stderr] warning: public documentation for `json_fsm` links to private item `FRAME_WORDS` [INFO] [stderr] --> src/grammar/json_fsm.rs:13:51 [INFO] [stderr] | [INFO] [stderr] 13 | //! - words 4..64: [`MAX_STACK`] = 15 frames of [`FRAME_WORDS`] = 4 words each, frame `i` at [INFO] [stderr] | ^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `Self::NODE_WORDS` [INFO] [stderr] --> src/grammar/json_schema.rs:17:36 [INFO] [stderr] | [INFO] [stderr] 17 | //! - `nodes: Vec<u32>` — stride [`Self::NODE_WORDS`] = 5: `[kind, f1, f2, f3, f4]`. `kind` is a [INFO] [stderr] | ^^^^^^^^^^^^^^^^ no item named `NODE_WORDS` in module `json_schema` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `Self::UNBOUNDED` [INFO] [stderr] --> src/grammar/json_schema.rs:20:77 [INFO] [stderr] | [INFO] [stderr] 20 | //! - `Array` → `f1 = item_node`, `f2 = min_items`, `f3 = max_items` ([`Self::UNBOUNDED`] = none). [INFO] [stderr] | ^^^^^^^^^^^^^^^ no item named `UNBOUNDED` in module `json_schema` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `Self::UNBOUNDED` [INFO] [stderr] --> src/grammar/json_schema.rs:21:55 [INFO] [stderr] | [INFO] [stderr] 21 | //! - `String` → `f1 = min_len`, `f2 = max_len` ([`Self::UNBOUNDED`] = none). [INFO] [stderr] | ^^^^^^^^^^^^^^^ no item named `UNBOUNDED` in module `json_schema` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `NODE_WORDS` links to private item `Self::nodes` [INFO] [stderr] --> src/grammar/json_schema.rs:88:36 [INFO] [stderr] | [INFO] [stderr] 88 | /// Words per node record in [`Self::nodes`]. [INFO] [stderr] | ^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `Qwen2` [INFO] [stderr] --> src/weights.rs:36:70 [INFO] [stderr] | [INFO] [stderr] 36 | /// Qwen2-MoE (`Qwen2MoeForCausalLM`, e.g. Qwen1.5-MoE-A2.7B): [`Qwen2`] attention (q/k/v **bias**, [INFO] [stderr] | ^^^^^ no item named `Qwen2` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `detect_model_kind` links to private item `decoder_arch` [INFO] [stderr] --> src/weights.rs:261:61 [INFO] [stderr] | [INFO] [stderr] 261 | /// `architectures[0]`; decoder families dispatch through [`decoder_arch`], which REFUSES [INFO] [stderr] | ^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `layers` links to private item `OP_LOAD` [INFO] [stderr] --> src/shard.rs:445:29 [INFO] [stderr] | [INFO] [stderr] 445 | /// the coordinator's [`OP_LOAD`] assignment (the auto-split path). [INFO] [stderr] | ^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `run_worker` links to private item `OP_LOAD` [INFO] [stderr] --> src/shard.rs:646:85 [INFO] [stderr] | [INFO] [stderr] 646 | /// the worker starts empty (AUTO) and loads whatever the coordinator assigns via [`OP_LOAD`]. [INFO] [stderr] | ^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `mtp_draft` links to private item `OP_MTP` [INFO] [stderr] --> src/shard.rs:2019:67 [INFO] [stderr] | [INFO] [stderr] 2019 | /// hidden and `first_tok` at absolute position `pos0` (see [`OP_MTP`]). [INFO] [stderr] | ^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `1` [INFO] [stderr] --> src/shard.rs:2117:44 [INFO] [stderr] | [INFO] [stderr] 2117 | /// Serving-MTP flags pack into header[1]: `ncols(16) | group(8)<<16 | k(6)<<24 | [INFO] [stderr] | ^ no item named `1` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `accept_fleet` links to private item `OP_JOIN` [INFO] [stderr] --> src/shard.rs:2486:46 [INFO] [stderr] | [INFO] [stderr] 2486 | /// Accept `n` reverse-registered workers ([`OP_JOIN`]) on `listen` — the fleet REGISTRY. [INFO] [stderr] | ^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `PAD` [INFO] [stderr] --> src/lib.rs:318:6 [INFO] [stderr] | [INFO] [stderr] 318 | /// [PAD] tokens — and the ragged encoder, which has no padding concept BY DESIGN (masking is [INFO] [stderr] | ^^^ no item named `PAD` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: unresolved link to `gemv_q4_src` [INFO] [stderr] --> src/lib.rs:2066:15 [INFO] [stderr] | [INFO] [stderr] 2066 | /// Batched [`gemv_q4_src`]: grid (⌈M/NR⌉, k). Same NR=2 body; col = wid.y. [INFO] [stderr] | ^^^^^^^^^^^ no item named `gemv_q4_src` in scope [INFO] [stderr] | [INFO] [stderr] = help: to escape `[` and `]` characters, add '\' before them like `\[` or `\]` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `mlp_gate_f16_lcpp_src` links to private item `mlp_gate_q4_lcpp_src` [INFO] [stderr] --> src/lib.rs:3189:7 [INFO] [stderr] | [INFO] [stderr] 3189 | /// [`mlp_gate_q4_lcpp_src`]'s f16-STORAGE twin: the SAME fused gate+up discipline (one WG = [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `mlp_gate_q1_f16x_src` links to private item `mlp_gate_q4_lcpp_src` [INFO] [stderr] --> src/lib.rs:3334:7 [INFO] [stderr] | [INFO] [stderr] 3334 | /// [`mlp_gate_q4_lcpp_src`]'s BINARY-weight twin (the Q1 class): the SAME fused gate+up [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `q1_lmhead_lcpp_src` links to private item `q4_lmhead_lcpp_src` [INFO] [stderr] --> src/lib.rs:5058:7 [INFO] [stderr] | [INFO] [stderr] 5058 | /// [`q4_lmhead_lcpp_src`]'s BINARY-weight twin: the same vocab-scale row-band folding over the [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: public documentation for `Q4LmHeadSparse` links to private item `q4_lmhead_sparse_src` [INFO] [stderr] --> src/lib.rs:5822:39 [INFO] [stderr] | [INFO] [stderr] 5822 | /// Sparse Q4 lm_head pipeline (see [`q4_lmhead_sparse_src`]); rows per workgroup = 8, matching the [INFO] [stderr] | ^^^^^^^^^^^^^^^^^^^^ this item is private [INFO] [stderr] | [INFO] [stderr] = note: this link will resolve properly if you pass `--document-private-items` [INFO] [stderr] [INFO] [stderr] warning: unclosed HTML tag `f16` [INFO] [stderr] --> src/encoder.rs:461:16 [INFO] [stderr] | [INFO] [stderr] 461 | /// tile's vec4<f16> accumulator folds into a persistent vec4<f32> accumulator at the tile [INFO] [stderr] | ^^^^^ [INFO] [stderr] | [INFO] [stderr] = note: `#[warn(rustdoc::invalid_html_tags)]` on by default [INFO] [stderr] help: try marking as source code [INFO] [stderr] | [INFO] [stderr] 461 | /// tile's `vec4<f16>` accumulator folds into a persistent vec4<f32> accumulator at the tile [INFO] [stderr] | + + [INFO] [stderr] [INFO] [stderr] warning: unclosed HTML tag `f32` [INFO] [stderr] --> src/encoder.rs:461:62 [INFO] [stderr] | [INFO] [stderr] 461 | /// tile's vec4<f16> accumulator folds into a persistent vec4<f32> accumulator at the tile [INFO] [stderr] | ^^^^^ [INFO] [stderr] | [INFO] [stderr] help: try marking as source code [INFO] [stderr] | [INFO] [stderr] 461 | /// tile's vec4<f16> accumulator folds into a persistent `vec4<f32>` accumulator at the tile [INFO] [stderr] | + + [INFO] [stderr] [INFO] [stderr] warning: redundant explicit link target [INFO] [stderr] --> src/encoder_weights.rs:6:17 [INFO] [stderr] | [INFO] [stderr] 6 | //! [`Pooling`](crate::pooling::Pooling) mode. Supporting a new encoder architecture is therefore [INFO] [stderr] | --------- ^^^^^^^^^^^^^^^^^^^^^^^ explicit target is redundant [INFO] [stderr] | | [INFO] [stderr] | because label contains path that resolves to same destination [INFO] [stderr] | [INFO] [stderr] = note: when a link's destination is not specified, [INFO] [stderr] the label is used to resolve intra-doc links [INFO] [stderr] = note: `#[warn(rustdoc::redundant_explicit_links)]` on by default [INFO] [stderr] help: remove explicit link target [INFO] [stderr] | [INFO] [stderr] 6 - //! [`Pooling`](crate::pooling::Pooling) mode. Supporting a new encoder architecture is therefore [INFO] [stderr] 6 + //! [`Pooling`] mode. Supporting a new encoder architecture is therefore [INFO] [stderr] | [INFO] [stderr] [INFO] [stderr] warning: unclosed HTML tag `Step` [INFO] [stderr] --> src/mimi_gpu.rs:9:19 [INFO] [stderr] | [INFO] [stderr] 9 | //! prebuilt Vec<Step> replayed into one command encoder — encode submits once and reads 32 [INFO] [stderr] | ^^^^^^ [INFO] [stderr] | [INFO] [stderr] help: try marking as source code [INFO] [stderr] | [INFO] [stderr] 9 | //! prebuilt `Vec<Step>` replayed into one command encoder — encode submits once and reads 32 [INFO] [stderr] | + + [INFO] [stderr] [INFO] [stderr] warning: `cfg` is both a module and a macro [INFO] [stderr] --> src/qwen3tts.rs:523:62 [INFO] [stderr] | [INFO] [stderr] 523 | /// fail loudly on any deviation from the landmarks in [`cfg`]. This is the P0 gate: a checkpoint [INFO] [stderr] | ^^^ ambiguous link [INFO] [stderr] | [INFO] [stderr] help: to link to the module, prefix with `mod@` [INFO] [stderr] | [INFO] [stderr] 523 | /// fail loudly on any deviation from the landmarks in [`mod@cfg`]. This is the P0 gate: a checkpoint [INFO] [stderr] | ++++ [INFO] [stderr] help: to link to the macro, add an exclamation mark [INFO] [stderr] | [INFO] [stderr] 523 | /// fail loudly on any deviation from the landmarks in [`cfg!`]. This is the P0 gate: a checkpoint [INFO] [stderr] | + [INFO] [stderr] [INFO] [stderr] warning: `inferencelayer` (lib doc) generated 69 warnings [INFO] [stderr] Finished `dev` profile [unoptimized + debuginfo] target(s) in 10.32s [INFO] [stderr] warning: the following packages contain code that will be rejected by a future version of Rust: naga v30.0.0, wgpu v30.0.0, wgpu-core v30.0.0, wgpu-hal v30.0.0 [INFO] [stderr] note: to see what the problems were, use the option `--future-incompat-report`, or run `cargo report future-incompatibilities --id 1` [INFO] [stderr] Generated /opt/rustwide/target/x86_64-unknown-linux-gnu/doc/inferencelayer/index.html [INFO] running `Command { std: "docker" "inspect" "ecacb71246d045f03be3c688626bb843e44bf57c44a679cedd7a2049eb28809c", kill_on_drop: false }`