MCPcopy Create free account

hub / github.com/Luce-Org/lucebox-hub / functions

Functions2,878 in github.com/Luce-Org/lucebox-hub

⨍ Functions 2,878◇ Types & classes 420
↓ 5 callersMethodrecord
server/src/common/moe_routing_collector.cpp:26
↓ 5 callersFunctionrecord_probe
( name: str, status: int, body: dict[str, Any] | str, elapsed: float, expect_status: int =
harness/client_test_runner.py:575
↓ 5 callersFunctionrel_stddev
(mean_value: float, stddev_value: float)
server/scripts/profile.py:584
↓ 5 callersFunctionreset_laguna_target_cache
server/src/laguna/laguna_target_graph.cpp:326
↓ 5 callersMethodrestore_and_generate
RESTORE <slot> <prompt_path> <n_gen> — restore snapshot + generate. Backend handles the diff-prefill and decode internally.
server/src/common/model_backend.h:206
↓ 5 callersFunctionrun_cmd
( cmd: list[str], *, cwd: Path | None = None, env: dict[str, str] | None = None, timeout:
harness/client_test_runner.py:218
↓ 5 callersFunctionscan_and_force_transitive
server/src/qwen3/anchor_scan.cpp:62
↓ 5 callersFunctionscore_layer_count
Helper: compute n_score_layers as the fixed allocator does.
server/test/test_drafter_warm_path_regression.cpp:13
↓ 5 callersFunctionselect_common_activation_precision_policy
server/src/common/backend_precision.cpp:229
↓ 5 callersMethodset_identity_salt
Set a config/model identity salt prepended to the layout hash buffer so that layout_id encodes model identity in addition to tensor structure. Call th
server/src/server/disk_prefix_cache.h:166
↓ 5 callersFunctionshell_tools
server/test/test_server_unit.cpp:130
↓ 5 callersFunctionsock_strerror
server/src/server/http_server.cpp:59
↓ 5 callersFunctionsplit_elapsed_us
server/src/deepseek4/deepseek4_layer_split_adapter.cpp:29
↓ 5 callersFunctionst_shape2
server/src/draft/draft_safetensors_loader.cpp:195
↓ 5 callersFunctionstream_chat
POST /v1/chat/completions with stream=True. Returns dict with: n_tok, wall_s, ttft_s, decode_s, decode_tps, wall_tps, text, usage (if
server/scripts/bench_server.py:47
↓ 5 callersFunctionsummarize
server/test/test_kvflash.cpp:432
↓ 5 callersFunctiontail
(path: Path, n: int = 5000)
harness/client_test_runner.py:1084
↓ 5 callersFunctiontail_fits
server/test/test_drafter_tail_capture_guard.cpp:35
↓ 5 callersFunctionturn
(history: list[dict], user: str)
server/scripts/test_multi_turn_prefix_cache.py:107
↓ 5 callersFunctionutf8_sanitize
Sanitize a string for JSON: replace invalid/incomplete UTF-8 with U+FFFD.
server/src/server/utf8_utils.h:24
↓ 5 callersFunctionweather_tools
server/test/test_server_unit.cpp:104
↓ 5 callersFunctionwrite_int32_file
server/test/test_generate.cpp:117
↓ 5 callersFunctionwrite_json
(path: Path | None, payload: dict[str, Any])
harness/client_test_runner.py:1091
↓ 4 callersFunction_load_op
()
optimizations/megakernel/model_nvfp4.py:73
↓ 4 callersFunction_m
(key)
server/scripts/bench_server.py:484
↓ 4 callersFunction_math_equiv
Check if two math answers are equivalent.
harness/math_scoring.py:64
↓ 4 callersFunction_print_row
(name: str, r: dict, score: str = "")
server/scripts/bench_server.py:357
↓ 4 callersFunction_str
(row: dict, *names)
server/scripts/profile.py:171
↓ 4 callersFunctionalign_up_i
server/src/qwen3/qwen3_drafter.cpp:43
↓ 4 callersFunctionbuild
(reps: int)
server/scripts/phase_split_dual_gpu.py:456
↓ 4 callersFunctionbuild_cached_decode_hc_pre_graph
server/src/deepseek4/deepseek4_graph.cpp:1343
↓ 4 callersFunctionbuild_mla_attention
server/src/deepseek4/deepseek4_graph.cpp:815
↓ 4 callersFunctioncoding_prompt
()
harness/client_test_runner.py:501
↓ 4 callersMethodcold_prefix_boundary
server/src/server/disk_prefix_cache.cpp:535
↓ 4 callersFunctioncopy_activation_to_host
server/src/common/target_shard_ipc.cpp:99
↓ 4 callersFunctioncount_set
Helper: count set entries in forced.
server/src/qwen3/anchor_scan.cpp:56
↓ 4 callersFunctioncreate_gemma4_cache_partial
server/src/gemma4/gemma4_loader.cpp:530
↓ 4 callersFunctioncreate_laguna_target_cache_partial
server/src/laguna/laguna_target_graph.cpp:56
↓ 4 callersFunctiondetect_arch
server/src/common/backend_factory.cpp:23
↓ 4 callersFunctiondraft_feature_mirror_sync_tail
server/src/common/dflash_feature_ring.cpp:355
↓ 4 callersFunctiondraft_residency_policy_name
server/src/placement/draft_residency.h:36
↓ 4 callersFunctionds4_comp_rows_used
server/src/deepseek4/deepseek4_graph.cpp:721
↓ 4 callersFunctionds4_parallel_for_tokens
server/src/deepseek4/deepseek4_graph.cpp:50
↓ 4 callersFunctionelapsed_ms
server/tests/test_deepseek4_unit.cpp:71
↓ 4 callersFunctionelapsed_s
server/src/deepseek4/deepseek4_backend.cpp:23
↓ 4 callersFunctionelapsed_us
server/src/qwen35moe/qwen35moe_backend.cpp:33
↓ 4 callersFunctionelapsed_us
server/src/deepseek4/deepseek4_backend.cpp:27
↓ 4 callersMethodembed_tokens
server/src/qwen35/qwen35_dflash_target.cpp:644
↓ 4 callersMethodempty
server/src/common/moe_hybrid_routing_stats.cpp:44
↓ 4 callersFunctionenable_peer_access_pair
server/src/common/peer_access.cpp:28
↓ 4 callersFunctionenv_path
(name: str, default: Path)
server/scripts/phase_split_dual_gpu.py:33
↓ 4 callersMethodepoch
Bumped on every residency change (alloc / page_out / page_in). Callers cache the slot mask and refill only when the epoch moves.
server/src/common/kvflash_pager.h:372
↓ 4 callersFunctioneval_moe_hybrid_ffn_single
server/src/common/moe_hybrid_ffn_eval.cpp:804
↓ 4 callersMethodfinish_reason
server/src/server/sse_emitter.cpp:768
↓ 4 callersFunctionflash_prefill_forward_bf16
server/src/flashprefill.cpp:232
↓ 4 callersMethodforward
server/src/common/target_shard_ipc.cpp:207
↓ 4 callersMethodfree
server/src/common/moe_expert_compute_ipc.cpp:154
↓ 4 callersFunctionfree_gemma4_weights
server/src/gemma4/gemma4_loader.cpp:515
↓ 4 callersFunctionfree_hip_chunk_graph_b
server/src/qwen3/qwen3_graph.cpp:115
↓ 4 callersFunctionfree_qwen3_drafter_model
server/src/qwen3/qwen3_loader.cpp:293
↓ 4 callersFunctiongemma4_step
server/src/gemma4/gemma4_graph.cpp:636
↓ 4 callersFunctiongenerate_id
server/src/server/http_server.cpp:357
↓ 4 callersMethodget_ema
server/src/server/adaptive_keep_ratio.h:77
↓ 4 callersFunctionget_f32_or
server/src/gemma4/gemma4_loader.cpp:106
↓ 4 callersFunctioninit_gpu_resident_state
server/src/common/moe_hybrid_ffn_eval.cpp:2096
↓ 4 callersFunctioninit_layer_split_runtime
server/src/common/layer_split_runtime.h:29
↓ 4 callersFunctionis_letter
server/src/server/tokenizer.cpp:60
↓ 4 callersMethodis_ready
server/src/common/moe_hybrid_stream.cpp:87
↓ 4 callersFunctionkvflash_test_sync_identity
server/test/test_server_unit.cpp:1972
↓ 4 callersFunctionlaguna_shexp_ffn
Shared-expert SwiGLU. With the loader's fused gate|up weight this is one matmul + one in-place split-activation (bit-identical to the two-matmul form:
server/src/laguna/laguna_target_graph.cpp:363
↓ 4 callersFunctionlaguna_snapshot_save
Save cache → snapshot. Handles alloc/realloc internally.
server/src/laguna/laguna_target_graph.cpp:209
↓ 4 callersFunctionlayer_split_kvflash_save_history_snapshot
server/src/common/layer_split_kvflash.h:87
↓ 4 callersFunctionload_dataset
()
server/scripts/quality_humaneval_plus.py:132
↓ 4 callersFunctionload_gemma4_gguf_partial
server/src/gemma4/gemma4_loader.cpp:184
↓ 4 callersFunctionmake_test_adapter
server/tests/test_deepseek4_unit.cpp:514
↓ 4 callersFunctionmask_align_up
server/src/common/dflash_draft_graph.cpp:16
↓ 4 callersMethodmaybe_store_continued
server/src/server/disk_prefix_cache.cpp:506
↓ 4 callersFunctionmoe_expert_apply_scale2
server/src/common/moe_expert_compute_ipc.cpp:132
↓ 4 callersFunctionmoe_expert_ipc_is_supported_input_type
server/src/common/moe_expert_compute_ipc.cpp:412
↓ 4 callersMethodn_chunks
server/src/common/kvflash_qk.h:114
↓ 4 callersMethodnext_path
server/src/common/backend_ipc.cpp:286
↓ 4 callersMethodopen_ro
server/src/deepseek4/deepseek4_loader.cpp:56
↓ 4 callersFunctionparse_nonnegative_env
server/src/common/moe_expert_compute.cpp:30
↓ 4 callersFunctionparse_session_id_from_body
─── Parse session_id from a chat-completion JSON body ────────────────── Returns empty string when session_id is absent or not a string (int/null/arra
server/src/server/http_server.h:389
↓ 4 callersMethodpinned_bytes
Total pinned buffer size.
server/src/common/moe_hybrid_stream.h:72
↓ 4 callersFunctionpooled_within_stddev
Run-to-run (within-prompt) stddev, pooled across prompts. The headline ± must reflect measurement *noise* — how much a single prompt's number
server/scripts/profile.py:588
↓ 4 callersFunctionprefill
(ids)
optimizations/megakernel/bench_pp_tg.py:60
↓ 4 callersFunctionquery_gpu_compute_sm
server/src/common/moe_hybrid_storage.cpp:29
↓ 4 callersMethodranked_experts
server/src/common/moe_hybrid_routing_stats.cpp:99
↓ 4 callersFunctionraxIteratorDelChars
Remove the specified number of chars from the right of the current * iterator key. */
server/src/server/rax.c:1553
↓ 4 callersFunctionraxIteratorEnterChild
Descend from 'parent' into child 'childidx', updating the iterator key. * If the child is inline we stop on the virtual leaf without changing the *
server/src/server/rax.c:1578
↓ 4 callersFunctionraxLowWalk
Low level function that walks the tree looking for the string * 's' of 'len' bytes. The function returns the number of characters * of the key that
server/src/server/rax.c:559
↓ 4 callersFunctionraxStackFree
Free the stack in case we used heap allocation. */
server/src/server/rax.c:145
↓ 4 callersFunctionraxStackPush
Push an item into the stack, returns 1 on success, 0 on out of memory. */
server/src/server/rax.c:102
↓ 4 callersFunctionread
(path: Path)
harness/clients/summarize_backend_pair.py:37
↓ 4 callersFunctionread_int32_file
server/test/test_generate.cpp:107
↓ 4 callersFunctionread_line_tail
server/src/common/io_utils.h:74
↓ 4 callersFunctionread_u64
server/src/server/disk_prefix_cache.cpp:183
↓ 4 callersMethodrelease
server/src/common/gguf_mmap.h:224
↓ 4 callersFunctionrender_chat_template
server/src/server/chat_template.cpp:55
← previousnext →401–500 of 2,878, ranked by callers