{ "_id": "6ad4b64cd423e4b0c576bcb8", "Package": "ggmlR", "Type": "Package", "Title": "'GGML' Tensor Operations for Machine Learning", "Version": "0.8.8", "Authors@R": "c(\nperson(\"Yuri\", \"Baramykov\",\nemail = \"lbsbmsu@mail.ru\",\nrole = c(\"aut\", \"cre\"),\ncomment = c(ORCID = \"0009-0000-7627-4217\")),\nperson(\"Georgi\", \"Gerganov\",\nrole = c(\"ctb\", \"cph\"),\ncomment = \"Author of the GGML library\"),\nperson(\"Jeffrey\", \"Quesnelle\",\nrole = c(\"ctb\", \"cph\"),\ncomment = \"Contributor to ops.cpp\"),\nperson(\"Bowen\", \"Peng\",\nrole = c(\"ctb\", \"cph\"),\ncomment = \"Contributor to ops.cpp\"),\nperson(\"Mozilla Foundation\",\nrole = c(\"ctb\", \"cph\"),\ncomment = \"Author of llamafile/sgemm.cpp\")\n)", "Description": "Provides 'R' bindings to the 'GGML' tensor library for\nmachine learning, optimized for 'Vulkan' GPU acceleration with\na transparent CPU fallback. The package features a 'Keras'-like\nsequential API and a 'PyTorch'-style 'autograd' engine for\nbuilding, training, and deploying neural networks. Key\ncapabilities include high-performance 5D tensor operations,\n'f16' precision, and efficient quantization. It supports native\n'ONNX' model import (50+ operators) and 'GGUF' weight loading\nfrom the 'llama.cpp' and 'Hugging Face' ecosystems. Designed\nfor zero-overhead inference via dedicated weight buffering, it\nintegrates seamlessly as a 'parsnip' engine for 'tidymodels'\nand provides first-class learners for the 'mlr3' framework. See\n for more information about\nthe underlying library.", "License": "MIT + file LICENSE", "URL": "https://github.com/Zabis13/ggmlR", "BugReports": "https://github.com/Zabis13/ggmlR/issues", "Encoding": "UTF-8", "SystemRequirements": "C++17, GNU make, libvulkan-dev, glslc (optional,\nfor GPU on Linux), 'Vulkan' 'SDK' (optional, for GPU on\nWindows)", "VignetteBuilder": "knitr, rmarkdown", "RoxygenNote": "7.3.3", "Config/testthat/edition": "3", "Config/build/clean-inst-doc": "false", "Config/pak/sysreqs": "make", "Repository": "https://zabis13.r-universe.dev", "Date/Publication": "2026-10-18 07:16:19 UTC", "RemoteUrl": "https://github.com/zabis13/ggmlr", "RemoteRef": "HEAD", "RemoteSha": "61c835098454641a78c8bec83a07ff06fbdc8a79", "NeedsCompilation": "yes", "Packaged": { "Date": "2026-10-18 11:34:13 UTC", "User": "root" }, "Author": "Yuri Baramykov [aut, cre] (ORCID:\n),\nGeorgi Gerganov [ctb, cph] (Author of the GGML library),\nJeffrey Quesnelle [ctb, cph] (Contributor to ops.cpp),\nBowen Peng [ctb, cph] (Contributor to ops.cpp),\nMozilla Foundation [ctb, cph] (Author of llamafile/sgemm.cpp)", "Maintainer": "Yuri Baramykov ", "_user": "zabis13", "_type": "src", "_file": "ggmlR_0.8.8.tar.gz", "_fileid": "https://r2.ropensci.org/7fe4f86fb741598ddcf2298726a632c71ccda19f0a6b945eb38853b23a784bef", "_filesize": 4306382, "_sha256": "7fe4f86fb741598ddcf2298726a632c71ccda19f0a6b945eb38853b23a784bef", "_expires": "2027-01-26T12:06:33.000Z", "_created": "2026-10-18T11:34:13.000Z", "_published": "2026-10-18T12:06:35.973Z", "_buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047", "_status": "success", "_host": "GitHub-Actions", "_jobs": [ { "job": 117664960528, "time": 275, "config": "linux-devel-arm64", "r": "4.7.0", "check": "NOTE", "artifact": "12135591435" }, { "job": 117664960544, "time": 347, "config": "linux-devel-x86_64", "r": "4.7.0", "check": "NOTE", "artifact": "12135850176" }, { "job": 117664960497, "time": 270, "config": "linux-release-arm64", "r": "4.6.1", "check": "NOTE", "artifact": "12134479267" }, { "job": 117664960494, "time": 264, "config": "linux-release-x86_64", "r": "4.6.1", "check": "NOTE", "artifact": "12135307773" }, { "job": 117664960485, "time": 212, "config": "macos-oldrel-arm64", "r": "4.5.3", "check": "NOTE", "artifact": "12135179704" }, { "job": 117664960545, "time": 375, "config": "macos-oldrel-x86_64", "r": "4.5.3", "check": "NOTE", "artifact": "12135966834" }, { "job": 117664960531, "time": 283, "config": "macos-release-arm64", "r": "4.6.1", "check": "NOTE", "artifact": "12135911274" }, { "job": 117664960547, "time": 520, "config": "macos-release-x86_64", "r": "4.6.1", "check": "NOTE", "artifact": "12136271055" }, { "job": 117663500336, "time": 436, "config": "source", "r": "4.6.1", "check": "OK", "artifact": "12135390796" }, { "job": 117664960707, "time": 194, "config": "wasm-release", "r": "4.6.0", "check": "OK", "artifact": "12135785197" }, { "job": 117664960587, "time": 440, "config": "windows-devel-arm64", "r": "4.7.0", "check": "NOTE", "artifact": "12135885265" }, { "job": 117664960576, "time": 404, "config": "windows-devel-x86_64", "r": "4.7.0", "check": "NOTE", "artifact": "12135611579" }, { "job": 117664960558, "time": 404, "config": "windows-oldrel-x86_64", "r": "4.5.3", "check": "NOTE", "artifact": "12135575850" }, { "job": 117664960596, "time": 430, "config": "windows-release-arm64", "r": "4.6.1", "check": "NOTE", "artifact": "12135925527" }, { "job": 117664960623, "time": 403, "config": "windows-release-x86_64", "r": "4.6.1", "check": "NOTE", "artifact": "12135641498" } ], "_upstream": "https://github.com/zabis13/ggmlr", "_commit": { "id": "61c835098454641a78c8bec83a07ff06fbdc8a79", "author": "Ваше Имя ", "committer": "Ваше Имя ", "message": "Upstream update\n", "time": 1792307779 }, "_maintainer": { "name": "Yuri Baramykov", "email": "lbsbmsu@mail.ru", "login": "zabis13", "description": "Yuri Baramykov", "uuid": 77751222, "orcid": "0009-0000-7627-4217" }, "_distro": "resolute", "_registered": true, "_dependencies": [ { "package": "R", "version": ">= 4.1.0", "role": "Depends" }, { "package": "generics", "role": "Imports" }, { "package": "R6", "role": "Imports" }, { "package": "methods", "role": "Imports" }, { "package": "stats", "role": "Imports" }, { "package": "utils", "role": "Imports" }, { "package": "testthat", "version": ">= 3.0.0", "role": "Suggests" }, { "package": "mlr3", "version": ">= 0.21.0", "role": "Suggests" }, { "package": "paradox", "role": "Suggests" }, { "package": "digest", "role": "Suggests" }, { "package": "parsnip", "role": "Suggests" }, { "package": "tibble", "role": "Suggests" }, { "package": "rlang", "role": "Suggests" }, { "package": "dials", "role": "Suggests" }, { "package": "lgr", "role": "Suggests" }, { "package": "knitr", "role": "Suggests" }, { "package": "rmarkdown", "role": "Suggests" }, { "package": "Matrix", "role": "Suggests" }, { "package": "Seurat", "role": "Suggests" }, { "package": "SeuratObject", "role": "Suggests" }, { "package": "RSpectra", "role": "Suggests" }, { "package": "irlba", "role": "Suggests" }, { "package": "uwot", "role": "Suggests" }, { "package": "FNN", "role": "Suggests" }, { "package": "SingleCellExperiment", "role": "Suggests" }, { "package": "SummarizedExperiment", "role": "Suggests" }, { "package": "S4Vectors", "role": "Suggests" }, { "package": "withr", "role": "Suggests" }, { "package": "hardhat", "role": "Suggests" }, { "package": "rsample", "role": "Suggests" }, { "package": "tune", "role": "Suggests" }, { "package": "workflows", "role": "Suggests" } ], "_owner": "zabis13", "_selfowned": true, "_usedby": 2, "_updates": [ { "week": "2026-06", "n": 2 }, { "week": "2026-07", "n": 15 }, { "week": "2026-08", "n": 28 }, { "week": "2026-09", "n": 1 }, { "week": "2026-11", "n": 8 }, { "week": "2026-12", "n": 3 }, { "week": "2026-13", "n": 10 }, { "week": "2026-14", "n": 11 }, { "week": "2026-15", "n": 2 }, { "week": "2026-16", "n": 6 }, { "week": "2026-17", "n": 2 }, { "week": "2026-21", "n": 7 }, { "week": "2026-22", "n": 1 }, { "week": "2026-23", "n": 2 }, { "week": "2026-24", "n": 20 }, { "week": "2026-25", "n": 15 }, { "week": "2026-27", "n": 5 }, { "week": "2026-28", "n": 26 }, { "week": "2026-29", "n": 10 }, { "week": "2026-30", "n": 3 }, { "week": "2026-31", "n": 8 }, { "week": "2026-32", "n": 2 }, { "week": "2026-34", "n": 5 }, { "week": "2026-35", "n": 2 }, { "week": "2026-36", "n": 3 }, { "week": "2026-37", "n": 5 }, { "week": "2026-38", "n": 5 }, { "week": "2026-40", "n": 3 }, { "week": "2026-41", "n": 2 }, { "week": "2026-42", "n": 2 } ], "_tags": [ { "name": "ggmlR_0.1.0", "date": "2026-01-16" }, { "name": "ggmlR_0.2.0", "date": "2026-01-16" }, { "name": "ggmlR_0.3.0", "date": "2026-01-23" }, { "name": "ggmlR_0.4.0", "date": "2026-01-25" }, { "name": "ggmlR_0.5.0", "date": "2026-01-31" }, { "name": "ggmlR_0.5.1", "date": "2026-02-01" }, { "name": "v0.5.2", "date": "2026-02-11" }, { "name": "v0.5.3", "date": "2026-02-11" }, { "name": "v0.5.4", "date": "2026-02-16" }, { "name": "v0.6.0", "date": "2026-02-20" }, { "name": "v0.6.1", "date": "2026-02-21" }, { "name": "v0.7.7", "date": "2026-05-18" }, { "name": "v0.7.8", "date": "2026-05-28" }, { "name": "v0.8.6", "date": "2026-10-02" } ], "_topics": [ "gpu-acceleration", "gpu-vulkan", "vulkan", "cpp", "openmp" ], "_stars": 20, "_contributors": [ { "user": "zabis13", "count": 129, "uuid": 77751222 }, { "user": "jeroen", "count": 1, "uuid": 216319 } ], "_userbio": { "uuid": 77751222, "type": "user", "name": "Yuri", "followers": 13, "description": "Yuri Baramykov" }, "_downloads": { "count": 625, "source": "https://cranlogs.r-pkg.org/downloads/total/last-month/ggmlR" }, "_devurl": "https://github.com/zabis13/ggmlr", "_searchresults": 161, "_rbuild": "4.6.1", "_assets": [ "extra/citation.cff", "extra/citation.html", "extra/citation.json", "extra/citation.txt", "extra/contents.json", "extra/ggmlR.html", "extra/manual.pdf", "extra/NEWS.html", "extra/NEWS.txt", "extra/readme.html", "extra/readme.md", "LICENSE" ], "_homeurl": "https://github.com/zabis13/ggmlr", "_realowner": "zabis13", "_cranurl": true, "_releases": [ { "version": "0.5.1", "date": "2026-02-16" }, { "version": "0.6.1", "date": "2026-02-22" }, { "version": "0.6.3", "date": "2026-03-18" }, { "version": "0.6.7", "date": "2026-03-29" }, { "version": "0.7.0", "date": "2026-04-06" }, { "version": "0.7.2", "date": "2026-04-15" }, { "version": "0.7.5", "date": "2026-04-20" }, { "version": "0.7.6", "date": "2026-04-22" }, { "version": "0.7.7", "date": "2026-05-25" }, { "version": "0.7.8", "date": "2026-06-04" }, { "version": "0.8.1", "date": "2026-07-14" }, { "version": "0.8.2", "date": "2026-07-22" }, { "version": "0.8.4", "date": "2026-08-21" } ], "_exports": [ "ag_add", "ag_argmax", "ag_batch_norm", "ag_buffer", "ag_buffer_free", "ag_buffer_read", "ag_buffer_write", "ag_capture", "ag_capture_free", "ag_capture_set", "ag_capture_step", "ag_checkpoint", "ag_clamp", "ag_concat_rows", "ag_cross_entropy_loss", "ag_dataloader", "ag_default_device", "ag_default_dtype", "ag_device", "ag_dropout", "ag_dtype", "ag_embedding", "ag_estimate_training_memory", "ag_eval", "ag_exp", "ag_flash_attention", "ag_gather", "ag_get_rows", "ag_grad", "ag_gradcheck", "ag_graph_mode", "ag_layer_norm", "ag_linear", "ag_load_model", "ag_load_state_dict", "ag_local_mode", "ag_log", "ag_log_softmax", "ag_matmul", "ag_matmul_precision", "ag_max", "ag_mean", "ag_mse_loss", "ag_mul", "ag_multihead_attention", "ag_param", "ag_pow", "ag_relu", "ag_reshape", "ag_sample_categorical", "ag_sample_normal", "ag_save_model", "ag_scale", "ag_sequential", "ag_sigmoid", "ag_slice_rows", "ag_softmax", "ag_softmax_cross_entropy_loss", "ag_state_dict", "ag_sub", "ag_sum", "ag_tanh", "ag_tape_memory", "ag_tensor", "ag_to_device", "ag_train", "ag_transpose", "as_gpu_matrix", "backward", "check_grad_anomaly", "clip_grad_norm", "clip_grad_value", "compile", "dequantize_row_iq1_m", "dequantize_row_iq1_s", "dequantize_row_iq2_s", "dequantize_row_iq2_xs", "dequantize_row_iq2_xxs", "dequantize_row_iq3_s", "dequantize_row_iq3_xxs", "dequantize_row_iq4_nl", "dequantize_row_iq4_xs", "dequantize_row_mxfp4", "dequantize_row_nvfp4", "dequantize_row_q1_0", "dequantize_row_q2_K", "dequantize_row_q3_K", "dequantize_row_q4_0", "dequantize_row_q4_1", "dequantize_row_q4_K", "dequantize_row_q5_0", "dequantize_row_q5_1", "dequantize_row_q5_K", "dequantize_row_q6_K", "dequantize_row_q8_0", "dequantize_row_q8_K", "dequantize_row_tq1_0", "dequantize_row_tq2_0", "dp_train", "evaluate", "fit", "ggml_abort_is_r_enabled", "ggml_abs", "ggml_abs_inplace", "ggml_add", "ggml_add_inplace", "ggml_add_rel_pos", "ggml_add1", "ggml_apply", "ggml_arange", "ggml_are_same_layout", "ggml_are_same_shape", "ggml_are_same_stride", "ggml_argmax", "ggml_argsort", "ggml_attention", "ggml_backend_alloc_ctx_tensors", "ggml_backend_buffer_clear", "ggml_backend_buffer_free", "ggml_backend_buffer_get_size", "ggml_backend_buffer_get_usage", "ggml_backend_buffer_is_host", "ggml_backend_buffer_is_multi_buffer", "ggml_backend_buffer_name", "ggml_backend_buffer_reset", "ggml_backend_buffer_set_usage", "ggml_backend_buffer_usage_any", "ggml_backend_buffer_usage_compute", "ggml_backend_buffer_usage_weights", "ggml_backend_buft_get_alignment", "ggml_backend_buft_get_max_size", "ggml_backend_buft_is_host", "ggml_backend_buft_name", "ggml_backend_cpu_init", "ggml_backend_cpu_set_n_threads", "ggml_backend_dev_buffer_type", "ggml_backend_dev_by_name", "ggml_backend_dev_by_type", "ggml_backend_dev_count", "ggml_backend_dev_description", "ggml_backend_dev_get", "ggml_backend_dev_get_props", "ggml_backend_dev_host_buffer_type", "ggml_backend_dev_init", "ggml_backend_dev_memory", "ggml_backend_dev_name", "ggml_backend_dev_offload_op", "ggml_backend_dev_supports_buft", "ggml_backend_dev_supports_op", "ggml_backend_dev_type", "ggml_backend_device_register", "ggml_backend_device_type_accel", "ggml_backend_device_type_cpu", "ggml_backend_device_type_gpu", "ggml_backend_device_type_igpu", "ggml_backend_event_free", "ggml_backend_event_new", "ggml_backend_event_record", "ggml_backend_event_synchronize", "ggml_backend_event_wait", "ggml_backend_free", "ggml_backend_get_device", "ggml_backend_graph_compute", "ggml_backend_graph_compute_async", "ggml_backend_graph_plan_compute", "ggml_backend_graph_plan_create", "ggml_backend_graph_plan_free", "ggml_backend_init_best", "ggml_backend_init_by_name", "ggml_backend_init_by_type", "ggml_backend_load", "ggml_backend_load_all", "ggml_backend_meta_device", "ggml_backend_multi_buffer_alloc_buffer", "ggml_backend_multi_buffer_set_usage", "ggml_backend_name", "ggml_backend_reg_by_name", "ggml_backend_reg_count", "ggml_backend_reg_dev_count", "ggml_backend_reg_dev_get", "ggml_backend_reg_get", "ggml_backend_reg_name", "ggml_backend_register", "ggml_backend_sched_alloc_graph", "ggml_backend_sched_free", "ggml_backend_sched_get_backend", "ggml_backend_sched_get_n_backends", "ggml_backend_sched_get_n_copies", "ggml_backend_sched_get_n_splits", "ggml_backend_sched_get_tensor_backend", "ggml_backend_sched_graph_compute", "ggml_backend_sched_graph_compute_async", "ggml_backend_sched_new", "ggml_backend_sched_reserve", "ggml_backend_sched_reset", "ggml_backend_sched_set_tensor_backend", "ggml_backend_sched_synchronize", "ggml_backend_sched_trace", "ggml_backend_synchronize", "ggml_backend_tensor_copy_async", "ggml_backend_tensor_get_and_sync", "ggml_backend_tensor_get_async", "ggml_backend_tensor_get_data", "ggml_backend_tensor_get_f32_first", "ggml_backend_tensor_set_async", "ggml_backend_tensor_set_data", "ggml_backend_unload", "ggml_batch_norm", "ggml_blck_size", "ggml_build_backward_expand", "ggml_build_forward_expand", "ggml_build_forward_expand_grads", "ggml_callback_early_stopping", "ggml_can_repeat", "ggml_cast", "ggml_cast_numeric", "ggml_ceil", "ggml_ceil_inplace", "ggml_clamp", "ggml_compile", "ggml_concat", "ggml_cont", "ggml_conv_1d", "ggml_conv_1d_dw", "ggml_conv_1d_layer", "ggml_conv_2d", "ggml_conv_2d_direct", "ggml_conv_2d_dw", "ggml_conv_2d_dw_direct", "ggml_conv_2d_layer", "ggml_conv_transpose_1d", "ggml_conv_transpose_2d_p0", "ggml_cos", "ggml_count_equal", "ggml_cpu_add", "ggml_cpu_features", "ggml_cpu_get_rvv_vlen", "ggml_cpu_get_sve_cnt", "ggml_cpu_has_amx_int8", "ggml_cpu_has_arm_fma", "ggml_cpu_has_avx", "ggml_cpu_has_avx_vnni", "ggml_cpu_has_avx2", "ggml_cpu_has_avx512", "ggml_cpu_has_avx512_bf16", "ggml_cpu_has_avx512_vbmi", "ggml_cpu_has_avx512_vnni", "ggml_cpu_has_bmi2", "ggml_cpu_has_dotprod", "ggml_cpu_has_f16c", "ggml_cpu_has_fma", "ggml_cpu_has_fp16_va", "ggml_cpu_has_llamafile", "ggml_cpu_has_matmul_int8", "ggml_cpu_has_neon", "ggml_cpu_has_riscv_v", "ggml_cpu_has_sme", "ggml_cpu_has_sse3", "ggml_cpu_has_ssse3", "ggml_cpu_has_sve", "ggml_cpu_has_vsx", "ggml_cpu_has_vxe", "ggml_cpu_has_wasm_simd", "ggml_cpu_mul", "ggml_cpy", "ggml_cross_entropy_loss", "ggml_cross_entropy_loss_back", "ggml_crossprod", "ggml_cumsum", "ggml_custom", "ggml_custom_inplace", "ggml_custom_ops", "ggml_cycles", "ggml_cycles_per_ms", "ggml_default_mlp", "ggml_dense", "ggml_diag", "ggml_diag_mask_inf", "ggml_diag_mask_inf_inplace", "ggml_diag_mask_zero", "ggml_div", "ggml_div_inplace", "ggml_dup", "ggml_dup_inplace", "ggml_dup_tensor", "ggml_element_size", "ggml_elu", "ggml_elu_inplace", "ggml_embedding", "ggml_estimate_memory", "ggml_evaluate", "ggml_exp", "ggml_exp_inplace", "ggml_extract", "ggml_fit", "ggml_fit_generator", "ggml_fit_opt", "ggml_fit_opt_gen", "ggml_fit_opt_multi", "ggml_flash_attn_back", "ggml_flash_attn_ext", "ggml_flash_attn_ext_add_sinks", "ggml_flash_attn_ext_get_prec", "ggml_flash_attn_ext_set_prec", "ggml_floor", "ggml_floor_inplace", "ggml_free", "ggml_freeze_bn_stats", "ggml_freeze_weights", "ggml_ftype_to_ggml_type", "ggml_gallocr_alloc_graph", "ggml_gallocr_free", "ggml_gallocr_get_buffer_size", "ggml_gallocr_new", "ggml_gallocr_new_buft", "ggml_gallocr_reserve", "ggml_gated_linear_attn", "ggml_gated_linear_attn_back", "ggml_geglu", "ggml_geglu_quick", "ggml_geglu_split", "ggml_gelu", "ggml_gelu_back", "ggml_gelu_erf", "ggml_gelu_inplace", "ggml_gelu_quick", "ggml_get_f32", "ggml_get_f32_nd", "ggml_get_first_tensor", "ggml_get_i32", "ggml_get_i32_nd", "ggml_get_layer", "ggml_get_max_tensor_size", "ggml_get_mem_size", "ggml_get_n_threads", "ggml_get_name", "ggml_get_next_tensor", "ggml_get_no_alloc", "ggml_get_op_params", "ggml_get_op_params_f32", "ggml_get_op_params_i32", "ggml_get_rel_pos", "ggml_get_rows", "ggml_get_rows_back", "ggml_get_unary_op", "ggml_glu", "GGML_GLU_OP_GEGLU", "GGML_GLU_OP_GEGLU_ERF", "GGML_GLU_OP_GEGLU_QUICK", "GGML_GLU_OP_REGLU", "GGML_GLU_OP_SWIGLU", "GGML_GLU_OP_SWIGLU_OAI", "ggml_glu_split", "ggml_graph_add_node", "ggml_graph_clear", "ggml_graph_compute", "ggml_graph_compute_with_ctx", "ggml_graph_cpy", "ggml_graph_dump_dot", "ggml_graph_dup", "ggml_graph_expand", "ggml_graph_get_grad", "ggml_graph_get_grad_acc", "ggml_graph_get_tensor", "ggml_graph_n_nodes", "ggml_graph_node", "ggml_graph_overhead", "ggml_graph_print", "ggml_graph_reset", "ggml_graph_view", "ggml_group_norm", "ggml_group_norm_inplace", "ggml_gru", "ggml_hardsigmoid", "ggml_hardswish", "ggml_im2col", "ggml_init", "ggml_init_auto", "ggml_inject", "ggml_input", "ggml_is_available", "ggml_is_contiguous", "ggml_is_contiguous_0", "ggml_is_contiguous_1", "ggml_is_contiguous_2", "ggml_is_contiguous_channels", "ggml_is_contiguous_rows", "ggml_is_contiguously_allocated", "ggml_is_permuted", "ggml_is_quantized", "ggml_is_transposed", "ggml_l2_norm", "ggml_l2_norm_inplace", "ggml_layer_add", "ggml_layer_attention", "ggml_layer_batch_norm", "ggml_layer_concatenate", "ggml_layer_conv_1d", "ggml_layer_conv_2d", "ggml_layer_custom", "ggml_layer_dense", "ggml_layer_dropout", "ggml_layer_embedding", "ggml_layer_flatten", "ggml_layer_global_average_pooling_2d", "ggml_layer_global_max_pooling_2d", "ggml_layer_gru", "ggml_layer_layer_norm", "ggml_layer_lstm", "ggml_layer_max_pooling_2d", "ggml_layer_permute", "ggml_layer_positional_embedding", "ggml_layer_reshape", "ggml_layer_rms_norm", "ggml_layer_sequence_pooling", "ggml_layer_transformer_block", "ggml_leaky_relu", "ggml_load_model", "ggml_load_weights", "ggml_log", "ggml_log_inplace", "ggml_log_is_r_enabled", "ggml_log_set_default", "ggml_log_set_r", "ggml_lstm", "ggml_marshal_model", "ggml_matmul", "ggml_matmul_f64", "ggml_mean", "ggml_model", "ggml_model_backend", "ggml_model_sequential", "ggml_mul", "ggml_mul_inplace", "ggml_mul_mat", "ggml_mul_mat_id", "ggml_n_dims", "GGML_N_TASKS_MAX", "ggml_nbytes", "ggml_neg", "ggml_neg_inplace", "ggml_nelements", "ggml_new_f32", "ggml_new_i32", "ggml_new_tensor", "ggml_new_tensor_1d", "ggml_new_tensor_2d", "ggml_new_tensor_3d", "ggml_new_tensor_4d", "ggml_norm", "ggml_norm_back", "ggml_norm_inplace", "ggml_nrows", "ggml_op_can_inplace", "ggml_op_desc", "ggml_op_name", "GGML_OP_POOL_AVG", "GGML_OP_POOL_MAX", "ggml_op_symbol", "ggml_ops_registry", "ggml_opt_alloc", "ggml_opt_context_optimizer_type", "ggml_opt_dataset_data", "ggml_opt_dataset_free", "ggml_opt_dataset_get_batch", "ggml_opt_dataset_get_batch_head", "ggml_opt_dataset_init", "ggml_opt_dataset_labels", "ggml_opt_dataset_ndata", "ggml_opt_dataset_shuffle", "ggml_opt_dataset_weights", "ggml_opt_default_params", "ggml_opt_epoch", "ggml_opt_eval", "ggml_opt_fit", "ggml_opt_fit_multi", "ggml_opt_free", "ggml_opt_get_lr", "ggml_opt_grad_acc", "ggml_opt_init", "ggml_opt_init_for_fit", "ggml_opt_init_for_fit_multi", "ggml_opt_inputs", "ggml_opt_labels", "ggml_opt_labels_i", "ggml_opt_loss", "ggml_opt_loss_i", "ggml_opt_loss_type_binary_cross_entropy", "ggml_opt_loss_type_cross_entropy", "ggml_opt_loss_type_huber", "ggml_opt_loss_type_mae", "ggml_opt_loss_type_mean", "ggml_opt_loss_type_mse", "ggml_opt_loss_type_sum", "ggml_opt_loss_type_weighted_mse", "ggml_opt_loss_weights_i", "ggml_opt_n_loss", "ggml_opt_ncorrect", "ggml_opt_ncorrect_i", "ggml_opt_optimizer_name", "ggml_opt_optimizer_type_adamw", "ggml_opt_optimizer_type_sgd", "ggml_opt_outputs", "ggml_opt_outputs_i", "ggml_opt_pred", "ggml_opt_pred_i", "ggml_opt_prepare_alloc", "ggml_opt_reset", "ggml_opt_result_accuracy", "ggml_opt_result_accuracy_i", "ggml_opt_result_free", "ggml_opt_result_init", "ggml_opt_result_loss", "ggml_opt_result_loss_i", "ggml_opt_result_n_loss", "ggml_opt_result_ndata", "ggml_opt_result_pred", "ggml_opt_result_reset", "ggml_opt_set_lr", "ggml_opt_static_graphs", "ggml_opt_step_adamw", "ggml_out_prod", "ggml_pad", "ggml_pad_ext", "ggml_pad_reflect_1d", "ggml_permute", "ggml_pool_1d", "ggml_pool_2d", "ggml_pop_layer", "ggml_pp_dp_forward", "ggml_pp_forward", "ggml_predict", "ggml_predict_classes", "ggml_print_mem_status", "ggml_print_objects", "ggml_quant_block_info", "ggml_quantize_chunk", "ggml_quantize_free", "ggml_quantize_init", "ggml_quantize_requires_imatrix", "ggml_reglu", "ggml_reglu_split", "ggml_relu", "ggml_relu_inplace", "ggml_repeat", "ggml_repeat_back", "ggml_reset", "ggml_reshape_1d", "ggml_reshape_2d", "ggml_reshape_3d", "ggml_reshape_4d", "ggml_result", "ggml_rms_norm", "ggml_rms_norm_back", "ggml_rms_norm_inplace", "ggml_roll", "ggml_rope", "ggml_rope_ext", "ggml_rope_ext_back", "ggml_rope_ext_inplace", "ggml_rope_inplace", "ggml_rope_multi", "ggml_rope_multi_back", "ggml_rope_multi_inplace", "GGML_ROPE_TYPE_MROPE", "GGML_ROPE_TYPE_NEOX", "GGML_ROPE_TYPE_NORM", "GGML_ROPE_TYPE_VISION", "ggml_round", "ggml_round_inplace", "ggml_run", "ggml_rwkv_output", "ggml_rwkv_state", "ggml_rwkv_wkv6", "ggml_rwkv_wkv6_back", "ggml_rwkv_wkv7", "ggml_rwkv_wkv7_back", "ggml_save_model", "ggml_save_weights", "ggml_scale", "ggml_scale_bias", "ggml_scale_inplace", "GGML_SCALE_MODE_BICUBIC", "GGML_SCALE_MODE_BILINEAR", "GGML_SCALE_MODE_NEAREST", "ggml_scatter_elements", "ggml_schedule_cosine_decay", "ggml_schedule_reduce_on_plateau", "ggml_schedule_step_decay", "ggml_set", "ggml_set_1d", "ggml_set_2d", "ggml_set_abort_callback_default", "ggml_set_abort_callback_r", "ggml_set_f32", "ggml_set_f32_nd", "ggml_set_i32", "ggml_set_i32_nd", "ggml_set_input", "ggml_set_loss", "ggml_set_n_threads", "ggml_set_name", "ggml_set_no_alloc", "ggml_set_omp_threads", "ggml_set_op_params", "ggml_set_op_params_f32", "ggml_set_op_params_i32", "ggml_set_output", "ggml_set_param", "ggml_set_seed", "ggml_set_zero", "ggml_sgn", "ggml_sigmoid", "ggml_sigmoid_inplace", "ggml_silu", "ggml_silu_back", "ggml_silu_inplace", "ggml_sin", "ggml_snapshot_free", "ggml_soft_max", "ggml_soft_max_add_sinks", "ggml_soft_max_ext", "ggml_soft_max_ext_back", "ggml_soft_max_ext_back_inplace", "ggml_soft_max_ext_inplace", "ggml_soft_max_inplace", "ggml_softplus", "ggml_softplus_inplace", "GGML_SORT_ORDER_ASC", "GGML_SORT_ORDER_DESC", "ggml_sqr", "ggml_sqr_inplace", "ggml_sqrt", "ggml_sqrt_inplace", "ggml_ssm_conv", "ggml_ssm_conv_back", "ggml_ssm_scan", "ggml_ssm_scan_back", "ggml_ssm_scan_output", "ggml_ssm_scan_state", "ggml_step", "ggml_sub", "ggml_sub_inplace", "ggml_sum", "ggml_sum_rows", "ggml_swiglu", "ggml_swiglu_split", "ggml_tanh", "ggml_tanh_inplace", "ggml_task", "ggml_tcrossprod", "ggml_tensor_copy", "ggml_tensor_data_ptr", "ggml_tensor_nb", "ggml_tensor_num", "ggml_tensor_overhead", "ggml_tensor_set_f32_scalar", "ggml_tensor_shape", "ggml_tensor_type", "ggml_test", "ggml_test_adamw_steps", "ggml_test_backend_ops", "ggml_time_init", "ggml_time_ms", "ggml_time_us", "ggml_timestep_embedding", "ggml_top_k", "ggml_tp_dp_forward", "ggml_trainer", "ggml_training_history", "ggml_transpose", "GGML_TYPE_BF16", "GGML_TYPE_F16", "GGML_TYPE_F32", "GGML_TYPE_I32", "ggml_type_name", "GGML_TYPE_Q2_K", "GGML_TYPE_Q3_K", "GGML_TYPE_Q4_0", "GGML_TYPE_Q4_1", "GGML_TYPE_Q4_K", "GGML_TYPE_Q5_K", "GGML_TYPE_Q6_K", "GGML_TYPE_Q8_0", "ggml_type_size", "ggml_type_sizef", "ggml_unary_op_name", "ggml_unfreeze_bn_stats", "ggml_unfreeze_weights", "ggml_unmarshal_model", "ggml_upscale", "ggml_used_mem", "ggml_version", "ggml_view_1d", "ggml_view_2d", "ggml_view_3d", "ggml_view_4d", "ggml_view_tensor", "ggml_vulkan_available", "ggml_vulkan_backend_name", "ggml_vulkan_device_caps", "ggml_vulkan_device_count", "ggml_vulkan_device_description", "ggml_vulkan_device_groups", "ggml_vulkan_device_memory", "ggml_vulkan_free", "ggml_vulkan_hard_exit_available", "ggml_vulkan_init", "ggml_vulkan_is_backend", "ggml_vulkan_list_devices", "ggml_vulkan_p2p_selftest", "ggml_vulkan_shutdown", "ggml_vulkan_split_buffer_type", "ggml_vulkan_split_mul_mat", "ggml_vulkan_split_row_ranges", "ggml_vulkan_status", "ggml_win_part", "ggml_win_unpart", "ggml_with_temp_ctx", "ggmlr_parsnip_fit_classif", "ggmlr_parsnip_fit_regr", "gguf_free", "gguf_load", "gguf_metadata", "gguf_tensor_data", "gguf_tensor_info", "gguf_tensor_names", "iq2xs_free_impl", "iq2xs_init_impl", "iq3xs_free_impl", "iq3xs_init_impl", "lr_scheduler_cosine", "lr_scheduler_cyclic", "lr_scheduler_onecycle", "lr_scheduler_step", "lr_scheduler_warmup_cosine", "nn_topo_sort", "onnx_device_info", "onnx_graph_info", "onnx_inputs", "onnx_load", "onnx_run", "onnx_summary", "optimizer_adam", "optimizer_sgd", "pt_cube4_layout", "pt_forward", "pt_transformer", "pt_transformer_load", "pth_catalogue", "pth_load", "quantize_iq1_m", "quantize_iq1_s", "quantize_iq2_s", "quantize_iq2_xs", "quantize_iq2_xxs", "quantize_iq3_s", "quantize_iq3_xxs", "quantize_iq4_nl", "quantize_iq4_xs", "quantize_mxfp4", "quantize_nvfp4", "quantize_q1_0", "quantize_q2_K", "quantize_q3_K", "quantize_q4_0", "quantize_q4_1", "quantize_q4_K", "quantize_q5_0", "quantize_q5_1", "quantize_q5_K", "quantize_q6_K", "quantize_q8_0", "quantize_row_iq2_s_ref", "quantize_row_iq3_s_ref", "quantize_row_iq3_xxs_ref", "quantize_row_iq4_nl_ref", "quantize_row_iq4_xs_ref", "quantize_row_mxfp4_ref", "quantize_row_nvfp4_ref", "quantize_row_q1_0_ref", "quantize_row_q2_K_ref", "quantize_row_q3_K_ref", "quantize_row_q4_0_ref", "quantize_row_q4_1_ref", "quantize_row_q4_K_ref", "quantize_row_q5_0_ref", "quantize_row_q5_1_ref", "quantize_row_q5_K_ref", "quantize_row_q6_K_ref", "quantize_row_q8_0_ref", "quantize_row_q8_1_ref", "quantize_row_q8_K_ref", "quantize_row_tq1_0_ref", "quantize_row_tq2_0_ref", "quantize_tq1_0", "quantize_tq2_0", "RunGGML", "show", "with_grad_tape" ], "_help": [ { "page": "ag_add", "title": "Element-wise addition with broadcasting", "topics": [ "ag_add" ] }, { "page": "ag_argmax", "title": "Column-wise argmax", "topics": [ "ag_argmax" ] }, { "page": "ag_batch_norm", "title": "Create a Batch Normalisation layer", "topics": [ "ag_batch_norm" ] }, { "page": "ag_buffer", "title": "Device buffer for column-indexed reads", "topics": [ "ag_buffer", "ag_buffer_free", "ag_buffer_read", "ag_buffer_write" ] }, { "page": "ag_capture", "title": "Capture an ag_* computation as a reusable device graph", "topics": [ "ag_capture" ] }, { "page": "ag_capture_free", "title": "Release the device memory held by a captured function", "topics": [ "ag_capture_free" ] }, { "page": "ag_capture_set", "title": "Change a setting of a captured training step", "topics": [ "ag_capture_set" ] }, { "page": "ag_capture_step", "title": "Capture a whole training step as reusable device graphs", "topics": [ "ag_capture_step" ] }, { "page": "ag_checkpoint", "title": "Run part of the forward pass without storing its activations", "topics": [ "ag_checkpoint" ] }, { "page": "ag_clamp", "title": "Element-wise clamp", "topics": [ "ag_clamp" ] }, { "page": "ag_concat_rows", "title": "Concatenate two tensors by rows", "topics": [ "ag_concat_rows" ] }, { "page": "ag_cross_entropy_loss", "title": "Categorical Cross-Entropy loss", "topics": [ "ag_cross_entropy_loss" ] }, { "page": "ag_dataloader", "title": "Create a mini-batch data loader", "topics": [ "ag_dataloader" ] }, { "page": "ag_default_device", "title": "Return the current default compute device", "topics": [ "ag_default_device" ] }, { "page": "ag_default_dtype", "title": "Return the current default dtype for GPU operations", "topics": [ "ag_default_dtype" ] }, { "page": "ag_device", "title": "Set the default compute device for ag_* operations", "topics": [ "ag_device" ] }, { "page": "ag_dropout", "title": "Create a Dropout layer", "topics": [ "ag_dropout" ] }, { "page": "ag_dtype", "title": "Set the default floating-point precision for ag_* GPU operations", "topics": [ "ag_dtype" ] }, { "page": "ag_embedding", "title": "Create an Embedding layer", "topics": [ "ag_embedding" ] }, { "page": "ag_estimate_training_memory", "title": "Estimate the memory a training run needs", "topics": [ "ag_estimate_training_memory" ] }, { "page": "ag_eval", "title": "Switch a layer or sequential model to eval mode", "topics": [ "ag_eval" ] }, { "page": "ag_exp", "title": "Element-wise exponential", "topics": [ "ag_exp" ] }, { "page": "ag_flash_attention", "title": "Multi-head attention in a single fused operation", "topics": [ "ag_flash_attention" ] }, { "page": "ag_gather", "title": "Pick one entry per column", "topics": [ "ag_gather" ] }, { "page": "ag_get_rows", "title": "Read buffer columns by index", "topics": [ "ag_get_rows" ] }, { "page": "ag_grad", "title": "Gradient of a tensor as an R matrix", "topics": [ "ag_grad" ] }, { "page": "ag_gradcheck", "title": "Numerical gradient check (like torch.autograd.gradcheck)", "topics": [ "ag_gradcheck" ] }, { "page": "ag_graph_mode", "title": "Graph mode for ag_* operations on the GPU", "topics": [ "ag_graph_mode" ] }, { "page": "ag_layer_norm", "title": "Create a Layer Normalization layer", "topics": [ "ag_layer_norm" ] }, { "page": "ag_linear", "title": "Create a dense layer with learnable parameters", "topics": [ "ag_linear" ] }, { "page": "ag_load_model", "title": "Load an autograd module from a saved state", "topics": [ "ag_load_model" ] }, { "page": "ag_load_state_dict", "title": "Load weights into an existing model", "topics": [ "ag_load_state_dict" ] }, { "page": "ag_local_mode", "title": "Set ag_* GPU modes for the rest of the calling function", "topics": [ "ag_local_mode" ] }, { "page": "ag_log", "title": "Element-wise natural logarithm", "topics": [ "ag_log" ] }, { "page": "ag_log_softmax", "title": "Column-wise log-softmax", "topics": [ "ag_log_softmax" ] }, { "page": "ag_matmul", "title": "Matrix multiplication", "topics": [ "ag_matmul" ] }, { "page": "ag_matmul_precision", "title": "Matrix-multiply precision for ag_* operations on the GPU", "topics": [ "ag_matmul_precision" ] }, { "page": "ag_max", "title": "Column-wise maximum", "topics": [ "ag_max" ] }, { "page": "ag_mean", "title": "Mean of elements (or along a dim)", "topics": [ "ag_mean" ] }, { "page": "ag_mse_loss", "title": "Mean Squared Error loss", "topics": [ "ag_mse_loss" ] }, { "page": "ag_mul", "title": "Element-wise multiplication", "topics": [ "ag_mul" ] }, { "page": "ag_multihead_attention", "title": "Create a Multi-Head Attention layer", "topics": [ "ag_multihead_attention" ] }, { "page": "ag_param", "title": "Create a parameter tensor (gradient tracked)", "topics": [ "ag_param" ] }, { "page": "ag_pow", "title": "Element-wise power", "topics": [ "ag_pow" ] }, { "page": "ag_relu", "title": "ReLU activation", "topics": [ "ag_relu" ] }, { "page": "ag_reshape", "title": "Reshape tensor", "topics": [ "ag_reshape" ] }, { "page": "ag_sample_categorical", "title": "Sample one index per column from logits", "topics": [ "ag_sample_categorical" ] }, { "page": "ag_sample_normal", "title": "Sample from a diagonal Gaussian", "topics": [ "ag_sample_normal" ] }, { "page": "ag_save_model", "title": "Save an autograd module's state to disk", "topics": [ "ag_save_model" ] }, { "page": "ag_scale", "title": "Scale tensor by a scalar constant", "topics": [ "ag_scale" ] }, { "page": "ag_sequential", "title": "Create a sequential container of layers", "topics": [ "ag_sequential" ] }, { "page": "ag_sigmoid", "title": "Sigmoid activation", "topics": [ "ag_sigmoid" ] }, { "page": "ag_slice_rows", "title": "Slice a range of rows", "topics": [ "ag_slice_rows" ] }, { "page": "ag_softmax", "title": "Softmax activation (column-wise)", "topics": [ "ag_softmax" ] }, { "page": "ag_softmax_cross_entropy_loss", "title": "Fused softmax + cross-entropy loss (numerically stable)", "topics": [ "ag_softmax_cross_entropy_loss" ] }, { "page": "ag_state_dict", "title": "Model weights as plain R data", "topics": [ "ag_state_dict" ] }, { "page": "ag_sub", "title": "Element-wise subtraction", "topics": [ "ag_sub" ] }, { "page": "ag_sum", "title": "Sum all elements (or along a dim): out = sum(x)", "topics": [ "ag_sum" ] }, { "page": "ag_tanh", "title": "Tanh activation", "topics": [ "ag_tanh" ] }, { "page": "ag_tape_memory", "title": "Report what the gradient tape is holding", "topics": [ "ag_tape_memory" ] }, { "page": "ag_tensor", "title": "Create a dynamic tensor (no gradient tracking)", "topics": [ "ag_tensor" ] }, { "page": "ag_to_device", "title": "Move a tensor to the specified device", "topics": [ "ag_to_device" ] }, { "page": "ag_train", "title": "Switch a layer or sequential model to training mode", "topics": [ "ag_train" ] }, { "page": "ag_transpose", "title": "Transpose a tensor", "topics": [ "ag_transpose" ] }, { "page": "as_gpu_matrix", "title": "Wrap a matrix so multiplies run on the GPU", "topics": [ "as_gpu_matrix" ] }, { "page": "as.matrix.ag_tensor", "title": "Extract the value of an ag_tensor as an R matrix", "topics": [ "as.matrix.ag_tensor" ] }, { "page": "as.matrix.ggml_matrix", "title": "Extract the underlying matrix from a ggml_matrix", "topics": [ "as.matrix.ggml_matrix" ] }, { "page": "augment.ggmlr_parsnip_model", "title": "Augment new data with predictions from a fitted ggml parsnip model", "topics": [ "augment.ggmlr_parsnip_model" ] }, { "page": "backward", "title": "Run backward pass from a scalar loss tensor", "topics": [ "backward" ] }, { "page": "check_grad_anomaly", "title": "Detect NaN / Inf gradients", "topics": [ "check_grad_anomaly" ] }, { "page": "clip_grad_norm", "title": "Clip gradients by global L2 norm", "topics": [ "clip_grad_norm" ] }, { "page": "clip_grad_value", "title": "Clip gradients element-wise by value", "topics": [ "clip_grad_value" ] }, { "page": "compile", "title": "Compile a Model", "topics": [ "compile.ggml_functional_model", "compile.ggml_sequential_model" ] }, { "page": "dequantize_row_iq2_xxs", "title": "Dequantize Row (IQ)", "concept": [ "quantization" ], "topics": [ "dequantize_row_iq1_m", "dequantize_row_iq1_s", "dequantize_row_iq2_s", "dequantize_row_iq2_xs", "dequantize_row_iq2_xxs", "dequantize_row_iq3_s", "dequantize_row_iq3_xxs", "dequantize_row_iq4_nl", "dequantize_row_iq4_xs" ] }, { "page": "dequantize_row_mxfp4", "title": "Dequantize Row (MXFP4)", "concept": [ "quantization" ], "topics": [ "dequantize_row_mxfp4" ] }, { "page": "dequantize_row_nvfp4", "title": "Dequantize NVFP4 Data", "concept": [ "quantization" ], "topics": [ "dequantize_row_nvfp4" ] }, { "page": "dequantize_row_q1_0", "title": "Dequantize Q1_0 Data", "concept": [ "quantization" ], "topics": [ "dequantize_row_q1_0" ] }, { "page": "dequantize_row_q2_K", "title": "Dequantize Row (K-quants)", "concept": [ "quantization" ], "topics": [ "dequantize_row_q2_K", "dequantize_row_q3_K", "dequantize_row_q4_K", "dequantize_row_q5_K", "dequantize_row_q6_K", "dequantize_row_q8_K" ] }, { "page": "dequantize_row_q4_0", "title": "Dequantize Row (Q4_0)", "concept": [ "quantization" ], "topics": [ "dequantize_row_q4_0", "dequantize_row_q4_1", "dequantize_row_q5_0", "dequantize_row_q5_1", "dequantize_row_q8_0" ] }, { "page": "dequantize_row_tq1_0", "title": "Dequantize Row (Ternary)", "concept": [ "quantization" ], "topics": [ "dequantize_row_tq1_0", "dequantize_row_tq2_0" ] }, { "page": "dp_train", "title": "Data-parallel training across multiple GPUs", "topics": [ "dp_train" ] }, { "page": "evaluate", "title": "Evaluate a Model", "topics": [ "evaluate", "evaluate.ggml_functional_model", "evaluate.ggml_sequential_model" ] }, { "page": "fit", "title": "Train a Model", "topics": [ "fit.ggml_functional_model", "fit.ggml_sequential_model" ] }, { "page": "ggml_abort_is_r_enabled", "title": "Check if R Abort Handler is Enabled", "concept": [ "logging" ], "topics": [ "ggml_abort_is_r_enabled" ] }, { "page": "ggml_abs", "title": "Absolute Value (Graph)", "topics": [ "ggml_abs" ] }, { "page": "ggml_abs_inplace", "title": "Absolute Value In-place (Graph)", "topics": [ "ggml_abs_inplace" ] }, { "page": "ggml_add", "title": "Add tensors", "topics": [ "ggml_add" ] }, { "page": "ggml_add_inplace", "title": "Element-wise Addition In-place (Graph)", "topics": [ "ggml_add_inplace" ] }, { "page": "ggml_add_rel_pos", "title": "Add Relative Position Bias (Graph)", "topics": [ "ggml_add_rel_pos" ] }, { "page": "ggml_add1", "title": "Add Scalar to Tensor (Graph)", "topics": [ "ggml_add1" ] }, { "page": "ggml_apply", "title": "Apply a Layer Object to a Tensor Node", "topics": [ "ggml_apply" ] }, { "page": "ggml_arange", "title": "Arange (Graph)", "topics": [ "ggml_arange" ] }, { "page": "ggml_are_same_layout", "title": "Check if Two Tensors Have the Same Layout", "concept": [ "tensor" ], "topics": [ "ggml_are_same_layout" ] }, { "page": "ggml_are_same_shape", "title": "Compare Tensor Shapes", "topics": [ "ggml_are_same_shape" ] }, { "page": "ggml_are_same_stride", "title": "Compare Tensor Strides", "concept": [ "tensor_layout" ], "topics": [ "ggml_are_same_stride" ] }, { "page": "ggml_argmax", "title": "Argmax (Graph)", "topics": [ "ggml_argmax" ] }, { "page": "ggml_argsort", "title": "Argsort - Get Sorting Indices (Graph)", "topics": [ "ggml_argsort" ] }, { "page": "ggml_attention", "title": "Create a Multi-Head Attention Layer Object", "topics": [ "ggml_attention" ] }, { "page": "ggml_backend_alloc_ctx_tensors", "title": "Allocate Context Tensors to Backend", "topics": [ "ggml_backend_alloc_ctx_tensors" ] }, { "page": "ggml_backend_buffer_clear", "title": "Clear buffer memory", "concept": [ "backend" ], "topics": [ "ggml_backend_buffer_clear" ] }, { "page": "ggml_backend_buffer_free", "title": "Free Backend Buffer", "topics": [ "ggml_backend_buffer_free" ] }, { "page": "ggml_backend_buffer_get_size", "title": "Get Backend Buffer Size", "topics": [ "ggml_backend_buffer_get_size" ] }, { "page": "ggml_backend_buffer_get_usage", "title": "Get buffer usage", "concept": [ "backend" ], "topics": [ "ggml_backend_buffer_get_usage" ] }, { "page": "ggml_backend_buffer_is_host", "title": "Check if buffer is host memory", "concept": [ "backend" ], "topics": [ "ggml_backend_buffer_is_host" ] }, { "page": "ggml_backend_buffer_is_multi_buffer", "title": "Check if buffer is a multi-buffer", "concept": [ "backend" ], "topics": [ "ggml_backend_buffer_is_multi_buffer" ] }, { "page": "ggml_backend_buffer_name", "title": "Get Backend Buffer Name", "topics": [ "ggml_backend_buffer_name" ] }, { "page": "ggml_backend_buffer_reset", "title": "Reset buffer", "concept": [ "backend" ], "topics": [ "ggml_backend_buffer_reset" ] }, { "page": "ggml_backend_buffer_set_usage", "title": "Set buffer usage hint", "concept": [ "backend" ], "topics": [ "ggml_backend_buffer_set_usage" ] }, { "page": "ggml_backend_buffer_usage_any", "title": "Buffer usage: Any", "concept": [ "backend" ], "topics": [ "ggml_backend_buffer_usage_any" ] }, { "page": "ggml_backend_buffer_usage_compute", "title": "Buffer usage: Compute", "concept": [ "backend" ], "topics": [ "ggml_backend_buffer_usage_compute" ] }, { "page": "ggml_backend_buffer_usage_weights", "title": "Buffer usage: Weights", "concept": [ "backend" ], "topics": [ "ggml_backend_buffer_usage_weights" ] }, { "page": "ggml_backend_buft_get_alignment", "title": "Get the allocation alignment of a buffer type", "concept": [ "backend" ], "topics": [ "ggml_backend_buft_get_alignment" ] }, { "page": "ggml_backend_buft_get_max_size", "title": "Get the maximum allocation size of a buffer type", "concept": [ "backend" ], "topics": [ "ggml_backend_buft_get_max_size" ] }, { "page": "ggml_backend_buft_is_host", "title": "Check whether a buffer type allocates host-accessible memory", "concept": [ "backend" ], "topics": [ "ggml_backend_buft_is_host" ] }, { "page": "ggml_backend_buft_name", "title": "Get the name of a buffer type", "concept": [ "backend" ], "topics": [ "ggml_backend_buft_name" ] }, { "page": "ggml_backend_cpu_init", "title": "Initialize CPU Backend", "topics": [ "ggml_backend_cpu_init" ] }, { "page": "ggml_backend_cpu_set_n_threads", "title": "Set CPU Backend Threads", "topics": [ "ggml_backend_cpu_set_n_threads" ] }, { "page": "ggml_backend_dev_buffer_type", "title": "Get the default buffer type of a device", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_buffer_type" ] }, { "page": "ggml_backend_dev_by_name", "title": "Get device by name", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_by_name" ] }, { "page": "ggml_backend_dev_by_type", "title": "Get device by type", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_by_type" ] }, { "page": "ggml_backend_dev_count", "title": "Get number of available devices", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_count" ] }, { "page": "ggml_backend_dev_description", "title": "Get device description", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_description" ] }, { "page": "ggml_backend_dev_get", "title": "Get device by index", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_get" ] }, { "page": "ggml_backend_dev_get_props", "title": "Get device properties", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_get_props" ] }, { "page": "ggml_backend_dev_host_buffer_type", "title": "Get the host (pinned) buffer type of a device", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_host_buffer_type" ] }, { "page": "ggml_backend_dev_init", "title": "Initialize backend from device", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_init" ] }, { "page": "ggml_backend_dev_memory", "title": "Get device memory", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_memory" ] }, { "page": "ggml_backend_dev_name", "title": "Get device name", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_name" ] }, { "page": "ggml_backend_dev_offload_op", "title": "Check if device should offload operation", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_offload_op" ] }, { "page": "ggml_backend_dev_supports_buft", "title": "Check if device supports buffer type", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_supports_buft" ] }, { "page": "ggml_backend_dev_supports_op", "title": "Check if device supports operation", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_supports_op" ] }, { "page": "ggml_backend_dev_type", "title": "Get device type", "concept": [ "backend" ], "topics": [ "ggml_backend_dev_type" ] }, { "page": "ggml_backend_device_register", "title": "Register a device", "concept": [ "backend" ], "topics": [ "ggml_backend_device_register" ] }, { "page": "ggml_backend_device_type_accel", "title": "Device type: Accelerator", "concept": [ "backend" ], "topics": [ "ggml_backend_device_type_accel" ] }, { "page": "ggml_backend_device_type_cpu", "title": "Device type: CPU", "concept": [ "backend" ], "topics": [ "ggml_backend_device_type_cpu" ] }, { "page": "ggml_backend_device_type_gpu", "title": "Device type: GPU", "concept": [ "backend" ], "topics": [ "ggml_backend_device_type_gpu" ] }, { "page": "ggml_backend_device_type_igpu", "title": "Device type: Integrated GPU", "concept": [ "backend" ], "topics": [ "ggml_backend_device_type_igpu" ] }, { "page": "ggml_backend_event_free", "title": "Free event", "concept": [ "backend" ], "topics": [ "ggml_backend_event_free" ] }, { "page": "ggml_backend_event_new", "title": "Create new event", "concept": [ "backend" ], "topics": [ "ggml_backend_event_new" ] }, { "page": "ggml_backend_event_record", "title": "Record event", "concept": [ "backend" ], "topics": [ "ggml_backend_event_record" ] }, { "page": "ggml_backend_event_synchronize", "title": "Synchronize event", "concept": [ "backend" ], "topics": [ "ggml_backend_event_synchronize" ] }, { "page": "ggml_backend_event_wait", "title": "Wait for event", "concept": [ "backend" ], "topics": [ "ggml_backend_event_wait" ] }, { "page": "ggml_backend_free", "title": "Free Backend", "topics": [ "ggml_backend_free" ] }, { "page": "ggml_backend_get_device", "title": "Get device from backend", "concept": [ "backend" ], "topics": [ "ggml_backend_get_device" ] }, { "page": "ggml_backend_graph_compute", "title": "Compute Graph with Backend", "topics": [ "ggml_backend_graph_compute" ] }, { "page": "ggml_backend_graph_compute_async", "title": "Compute graph asynchronously", "concept": [ "backend" ], "topics": [ "ggml_backend_graph_compute_async" ] }, { "page": "ggml_backend_graph_plan_compute", "title": "Execute graph plan", "concept": [ "backend" ], "topics": [ "ggml_backend_graph_plan_compute" ] }, { "page": "ggml_backend_graph_plan_create", "title": "Create graph execution plan", "concept": [ "backend" ], "topics": [ "ggml_backend_graph_plan_create" ] }, { "page": "ggml_backend_graph_plan_free", "title": "Free graph execution plan", "concept": [ "backend" ], "topics": [ "ggml_backend_graph_plan_free" ] }, { "page": "ggml_backend_init_best", "title": "Initialize best available backend", "concept": [ "backend" ], "topics": [ "ggml_backend_init_best" ] }, { "page": "ggml_backend_init_by_name", "title": "Initialize backend by name", "concept": [ "backend" ], "topics": [ "ggml_backend_init_by_name" ] }, { "page": "ggml_backend_init_by_type", "title": "Initialize backend by type", "concept": [ "backend" ], "topics": [ "ggml_backend_init_by_type" ] }, { "page": "ggml_backend_load", "title": "Load backend from dynamic library", "concept": [ "backend" ], "topics": [ "ggml_backend_load" ] }, { "page": "ggml_backend_load_all", "title": "Load all available backends", "concept": [ "backend" ], "topics": [ "ggml_backend_load_all" ] }, { "page": "ggml_backend_meta_device", "title": "Create a Meta Backend Device", "concept": [ "backend" ], "topics": [ "ggml_backend_meta_device" ] }, { "page": "ggml_backend_multi_buffer_alloc_buffer", "title": "Allocate multi-buffer", "concept": [ "backend" ], "topics": [ "ggml_backend_multi_buffer_alloc_buffer" ] }, { "page": "ggml_backend_multi_buffer_set_usage", "title": "Set usage for all buffers in a multi-buffer", "concept": [ "backend" ], "topics": [ "ggml_backend_multi_buffer_set_usage" ] }, { "page": "ggml_backend_name", "title": "Get Backend Name", "topics": [ "ggml_backend_name" ] }, { "page": "ggml_backend_reg_by_name", "title": "Get backend registry by name", "concept": [ "backend" ], "topics": [ "ggml_backend_reg_by_name" ] }, { "page": "ggml_backend_reg_count", "title": "Get number of registered backends", "concept": [ "backend" ], "topics": [ "ggml_backend_reg_count" ] }, { "page": "ggml_backend_reg_dev_count", "title": "Get number of devices in registry", "concept": [ "backend" ], "topics": [ "ggml_backend_reg_dev_count" ] }, { "page": "ggml_backend_reg_dev_get", "title": "Get device from registry", "concept": [ "backend" ], "topics": [ "ggml_backend_reg_dev_get" ] }, { "page": "ggml_backend_reg_get", "title": "Get backend registry by index", "concept": [ "backend" ], "topics": [ "ggml_backend_reg_get" ] }, { "page": "ggml_backend_reg_name", "title": "Get registry name", "concept": [ "backend" ], "topics": [ "ggml_backend_reg_name" ] }, { "page": "ggml_backend_register", "title": "Register a backend", "concept": [ "backend" ], "topics": [ "ggml_backend_register" ] }, { "page": "ggml_backend_sched_alloc_graph", "title": "Allocate graph on scheduler", "topics": [ "ggml_backend_sched_alloc_graph" ] }, { "page": "ggml_backend_sched_free", "title": "Free backend scheduler", "topics": [ "ggml_backend_sched_free" ] }, { "page": "ggml_backend_sched_get_backend", "title": "Get backend from scheduler", "topics": [ "ggml_backend_sched_get_backend" ] }, { "page": "ggml_backend_sched_get_n_backends", "title": "Get number of backends in scheduler", "topics": [ "ggml_backend_sched_get_n_backends" ] }, { "page": "ggml_backend_sched_get_n_copies", "title": "Get number of tensor copies", "topics": [ "ggml_backend_sched_get_n_copies" ] }, { "page": "ggml_backend_sched_get_n_splits", "title": "Get number of graph splits", "topics": [ "ggml_backend_sched_get_n_splits" ] }, { "page": "ggml_backend_sched_get_tensor_backend", "title": "Get tensor backend assignment", "topics": [ "ggml_backend_sched_get_tensor_backend" ] }, { "page": "ggml_backend_sched_graph_compute", "title": "Compute graph using scheduler", "topics": [ "ggml_backend_sched_graph_compute" ] }, { "page": "ggml_backend_sched_graph_compute_async", "title": "Compute graph asynchronously", "topics": [ "ggml_backend_sched_graph_compute_async" ] }, { "page": "ggml_backend_sched_new", "title": "Create a new backend scheduler", "topics": [ "ggml_backend_sched_new" ] }, { "page": "ggml_backend_sched_reserve", "title": "Reserve memory for scheduler", "topics": [ "ggml_backend_sched_reserve" ] }, { "page": "ggml_backend_sched_reset", "title": "Reset scheduler", "topics": [ "ggml_backend_sched_reset" ] }, { "page": "ggml_backend_sched_set_tensor_backend", "title": "Set tensor backend assignment", "topics": [ "ggml_backend_sched_set_tensor_backend" ] }, { "page": "ggml_backend_sched_synchronize", "title": "Synchronize scheduler", "topics": [ "ggml_backend_sched_synchronize" ] }, { "page": "ggml_backend_sched_trace", "title": "Trace Every Node a Scheduler Computes", "topics": [ "ggml_backend_sched_trace" ] }, { "page": "ggml_backend_synchronize", "title": "Synchronize backend", "concept": [ "backend" ], "topics": [ "ggml_backend_synchronize" ] }, { "page": "ggml_backend_tensor_copy_async", "title": "Copy tensor asynchronously between backends", "concept": [ "backend" ], "topics": [ "ggml_backend_tensor_copy_async" ] }, { "page": "ggml_backend_tensor_get_and_sync", "title": "Backend Tensor Get and Sync", "topics": [ "ggml_backend_tensor_get_and_sync" ] }, { "page": "ggml_backend_tensor_get_async", "title": "Get tensor data asynchronously", "concept": [ "backend" ], "topics": [ "ggml_backend_tensor_get_async" ] }, { "page": "ggml_backend_tensor_get_data", "title": "Get Tensor Data via Backend", "topics": [ "ggml_backend_tensor_get_data" ] }, { "page": "ggml_backend_tensor_get_f32_first", "title": "Get First Float from Backend Tensor", "topics": [ "ggml_backend_tensor_get_f32_first" ] }, { "page": "ggml_backend_tensor_set_async", "title": "Set tensor data asynchronously", "concept": [ "backend" ], "topics": [ "ggml_backend_tensor_set_async" ] }, { "page": "ggml_backend_tensor_set_data", "title": "Set Tensor Data via Backend", "topics": [ "ggml_backend_tensor_set_data" ] }, { "page": "ggml_backend_unload", "title": "Unload backend", "concept": [ "backend" ], "topics": [ "ggml_backend_unload" ] }, { "page": "ggml_batch_norm", "title": "Create a Batch Normalization Layer Object", "topics": [ "ggml_batch_norm" ] }, { "page": "ggml_blck_size", "title": "Get Block Size", "concept": [ "type_system" ], "topics": [ "ggml_blck_size" ] }, { "page": "ggml_build_backward_expand", "title": "Add the Backward Pass to a Computation Graph", "topics": [ "ggml_build_backward_expand" ] }, { "page": "ggml_build_forward_expand", "title": "Build forward expand", "topics": [ "ggml_build_forward_expand" ] }, { "page": "ggml_build_forward_expand_grads", "title": "Build a Forward Graph That Can Be Differentiated", "topics": [ "ggml_build_forward_expand_grads" ] }, { "page": "ggml_callback_early_stopping", "title": "Early stopping callback", "concept": [ "callbacks" ], "topics": [ "ggml_callback_early_stopping" ] }, { "page": "ggml_can_repeat", "title": "Check If Tensor Can Be Repeated", "concept": [ "tensor_layout" ], "topics": [ "ggml_can_repeat" ] }, { "page": "ggml_cast", "title": "Cast a Tensor to Another Type", "topics": [ "ggml_cast" ] }, { "page": "ggml_cast_numeric", "title": "Numeric Type Conversion", "topics": [ "ggml_cast_numeric" ] }, { "page": "ggml_ceil", "title": "Ceiling (Graph)", "topics": [ "ggml_ceil" ] }, { "page": "ggml_ceil_inplace", "title": "Ceiling In-place (Graph)", "topics": [ "ggml_ceil_inplace" ] }, { "page": "ggml_clamp", "title": "Clamp (Graph)", "topics": [ "ggml_clamp" ] }, { "page": "ggml_compile", "title": "Compile a Sequential Model", "topics": [ "ggml_compile", "ggml_compile.ggml_functional_model", "ggml_compile.ggml_sequential_model" ] }, { "page": "ggml_concat", "title": "Concatenate Tensors (Graph)", "topics": [ "ggml_concat" ] }, { "page": "ggml_cont", "title": "Make Contiguous (Graph)", "topics": [ "ggml_cont" ] }, { "page": "ggml_conv_1d", "title": "1D Convolution (Graph)", "topics": [ "ggml_conv_1d" ] }, { "page": "ggml_conv_1d_dw", "title": "Depthwise 1D Convolution (Graph)", "topics": [ "ggml_conv_1d_dw" ] }, { "page": "ggml_conv_1d_layer", "title": "Create a Reusable Conv1D Layer Object", "topics": [ "ggml_conv_1d_layer" ] }, { "page": "ggml_conv_2d", "title": "2D Convolution (Graph)", "topics": [ "ggml_conv_2d" ] }, { "page": "ggml_conv_2d_direct", "title": "Direct 2D Convolution (Graph)", "topics": [ "ggml_conv_2d_direct" ] }, { "page": "ggml_conv_2d_dw", "title": "Depthwise 2D Convolution (Graph)", "topics": [ "ggml_conv_2d_dw" ] }, { "page": "ggml_conv_2d_dw_direct", "title": "Depthwise 2D Convolution, direct (Graph)", "topics": [ "ggml_conv_2d_dw_direct" ] }, { "page": "ggml_conv_2d_layer", "title": "Create a Reusable Conv2D Layer Object", "topics": [ "ggml_conv_2d_layer" ] }, { "page": "ggml_conv_transpose_1d", "title": "Transposed 1D Convolution (Graph)", "topics": [ "ggml_conv_transpose_1d" ] }, { "page": "ggml_conv_transpose_2d_p0", "title": "Transposed 2D Convolution, zero padding (Graph)", "topics": [ "ggml_conv_transpose_2d_p0" ] }, { "page": "ggml_cos", "title": "Cosine (Graph)", "topics": [ "ggml_cos" ] }, { "page": "ggml_count_equal", "title": "Count Equal Elements (Graph)", "concept": [ "tensor_layout" ], "topics": [ "ggml_count_equal" ] }, { "page": "ggml_cpu_add", "title": "Element-wise Addition (CPU Direct)", "topics": [ "ggml_cpu_add" ] }, { "page": "ggml_cpu_features", "title": "Get All CPU Features", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_features" ] }, { "page": "ggml_cpu_get_rvv_vlen", "title": "Get RISC-V Vector Length", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_get_rvv_vlen" ] }, { "page": "ggml_cpu_get_sve_cnt", "title": "Get SVE Vector Length (ARM)", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_get_sve_cnt" ] }, { "page": "ggml_cpu_has_amx_int8", "title": "CPU Feature Detection - AMX INT8", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_amx_int8" ] }, { "page": "ggml_cpu_has_arm_fma", "title": "CPU Feature Detection - ARM FMA", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_arm_fma" ] }, { "page": "ggml_cpu_has_avx", "title": "CPU Feature Detection - AVX", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_avx" ] }, { "page": "ggml_cpu_has_avx_vnni", "title": "CPU Feature Detection - AVX-VNNI", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_avx_vnni" ] }, { "page": "ggml_cpu_has_avx2", "title": "CPU Feature Detection - AVX2", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_avx2" ] }, { "page": "ggml_cpu_has_avx512", "title": "CPU Feature Detection - AVX-512", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_avx512" ] }, { "page": "ggml_cpu_has_avx512_bf16", "title": "CPU Feature Detection - AVX-512 BF16", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_avx512_bf16" ] }, { "page": "ggml_cpu_has_avx512_vbmi", "title": "CPU Feature Detection - AVX-512 VBMI", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_avx512_vbmi" ] }, { "page": "ggml_cpu_has_avx512_vnni", "title": "CPU Feature Detection - AVX-512 VNNI", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_avx512_vnni" ] }, { "page": "ggml_cpu_has_bmi2", "title": "CPU Feature Detection - BMI2", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_bmi2" ] }, { "page": "ggml_cpu_has_dotprod", "title": "CPU Feature Detection - Dot Product (ARM)", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_dotprod" ] }, { "page": "ggml_cpu_has_f16c", "title": "CPU Feature Detection - F16C", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_f16c" ] }, { "page": "ggml_cpu_has_fma", "title": "CPU Feature Detection - FMA", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_fma" ] }, { "page": "ggml_cpu_has_fp16_va", "title": "CPU Feature Detection - FP16 Vector Arithmetic (ARM)", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_fp16_va" ] }, { "page": "ggml_cpu_has_llamafile", "title": "CPU Feature Detection - Llamafile", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_llamafile" ] }, { "page": "ggml_cpu_has_matmul_int8", "title": "CPU Feature Detection - INT8 Matrix Multiply (ARM)", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_matmul_int8" ] }, { "page": "ggml_cpu_has_neon", "title": "CPU Feature Detection - NEON (ARM)", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_neon" ] }, { "page": "ggml_cpu_has_riscv_v", "title": "CPU Feature Detection - RISC-V Vector", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_riscv_v" ] }, { "page": "ggml_cpu_has_sme", "title": "CPU Feature Detection - SME (ARM)", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_sme" ] }, { "page": "ggml_cpu_has_sse3", "title": "CPU Feature Detection - SSE3", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_sse3" ] }, { "page": "ggml_cpu_has_ssse3", "title": "CPU Feature Detection - SSSE3", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_ssse3" ] }, { "page": "ggml_cpu_has_sve", "title": "CPU Feature Detection - SVE (ARM)", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_sve" ] }, { "page": "ggml_cpu_has_vsx", "title": "CPU Feature Detection - VSX (PowerPC)", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_vsx" ] }, { "page": "ggml_cpu_has_vxe", "title": "CPU Feature Detection - VXE (IBM z/Architecture)", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_vxe" ] }, { "page": "ggml_cpu_has_wasm_simd", "title": "CPU Feature Detection - WebAssembly SIMD", "concept": [ "cpu_features" ], "topics": [ "ggml_cpu_has_wasm_simd" ] }, { "page": "ggml_cpu_mul", "title": "Element-wise Multiplication (CPU Direct)", "topics": [ "ggml_cpu_mul" ] }, { "page": "ggml_cpy", "title": "Copy Tensor with Type Conversion (Graph)", "topics": [ "ggml_cpy" ] }, { "page": "ggml_cross_entropy_loss", "title": "Cross-Entropy Loss (Graph)", "topics": [ "ggml_cross_entropy_loss" ] }, { "page": "ggml_cross_entropy_loss_back", "title": "Cross-Entropy Loss Backward (Graph)", "topics": [ "ggml_cross_entropy_loss_back" ] }, { "page": "ggml_crossprod", "title": "GPU cross product (drop-in for 'crossprod')", "topics": [ "ggml_crossprod" ] }, { "page": "ggml_cumsum", "title": "Cumulative Sum (Graph)", "topics": [ "ggml_cumsum" ] }, { "page": "ggml_custom", "title": "Custom Operation Node", "topics": [ "ggml_custom" ] }, { "page": "ggml_custom_inplace", "title": "In-Place Custom Operation Node", "topics": [ "ggml_custom_inplace" ] }, { "page": "ggml_custom_ops", "title": "List Registered Custom Ops", "topics": [ "ggml_custom_ops" ] }, { "page": "ggml_cycles", "title": "Get CPU Cycles", "topics": [ "ggml_cycles" ] }, { "page": "ggml_cycles_per_ms", "title": "Get CPU Cycles per Millisecond", "topics": [ "ggml_cycles_per_ms" ] }, { "page": "ggml_default_mlp", "title": "Default MLP builder for classification and regression", "topics": [ "ggml_default_mlp" ] }, { "page": "ggml_dense", "title": "Create a Dense Layer Object", "topics": [ "ggml_dense" ] }, { "page": "ggml_diag", "title": "Diagonal Matrix (Graph)", "topics": [ "ggml_diag" ] }, { "page": "ggml_diag_mask_inf", "title": "Diagonal Mask with -Inf (Graph)", "topics": [ "ggml_diag_mask_inf" ] }, { "page": "ggml_diag_mask_inf_inplace", "title": "Diagonal Mask with -Inf In-place (Graph)", "topics": [ "ggml_diag_mask_inf_inplace" ] }, { "page": "ggml_diag_mask_zero", "title": "Diagonal Mask with Zero (Graph)", "topics": [ "ggml_diag_mask_zero" ] }, { "page": "ggml_div", "title": "Element-wise Division (Graph)", "topics": [ "ggml_div" ] }, { "page": "ggml_div_inplace", "title": "Element-wise Division In-place (Graph)", "topics": [ "ggml_div_inplace" ] }, { "page": "ggml_dup", "title": "Duplicate Tensor (Graph)", "topics": [ "ggml_dup" ] }, { "page": "ggml_dup_inplace", "title": "Duplicate Tensor In-place (Graph)", "topics": [ "ggml_dup_inplace" ] }, { "page": "ggml_dup_tensor", "title": "Duplicate Tensor", "topics": [ "ggml_dup_tensor" ] }, { "page": "ggml_element_size", "title": "Get Element Size", "topics": [ "ggml_element_size" ] }, { "page": "ggml_elu", "title": "ELU Activation (Graph)", "topics": [ "ggml_elu" ] }, { "page": "ggml_elu_inplace", "title": "ELU Activation In-place (Graph)", "topics": [ "ggml_elu_inplace" ] }, { "page": "ggml_embedding", "title": "Create an Embedding Layer Object", "topics": [ "ggml_embedding" ] }, { "page": "ggml_estimate_memory", "title": "Estimate Required Memory", "topics": [ "ggml_estimate_memory" ] }, { "page": "ggml_evaluate", "title": "Evaluate a Trained Model", "topics": [ "ggml_evaluate", "ggml_evaluate.ggml_functional_model", "ggml_evaluate.ggml_sequential_model" ] }, { "page": "ggml_exp", "title": "Exponential (Graph)", "topics": [ "ggml_exp" ] }, { "page": "ggml_exp_inplace", "title": "Exponential In-place (Graph)", "topics": [ "ggml_exp_inplace" ] }, { "page": "ggml_extract", "title": "Extract a feature-by-cell matrix from a single-cell container", "topics": [ "ggml_extract", "ggml_extract.dgCMatrix", "ggml_extract.matrix", "ggml_extract.Seurat", "ggml_extract.SingleCellExperiment" ] }, { "page": "ggml_fit_generator", "title": "Train a Sequential Model from a Generator", "topics": [ "ggml_fit_generator" ] }, { "page": "ggml_fit_opt", "title": "Fit model with R-side epoch loop and callbacks", "concept": [ "optimization" ], "topics": [ "ggml_fit_opt" ] }, { "page": "ggml_fit_opt_gen", "title": "Train from a generator, one batch at a time", "concept": [ "optimization" ], "topics": [ "ggml_fit_opt_gen" ] }, { "page": "ggml_fit_opt_multi", "title": "Fit a multi-output model with an R-side epoch loop and callbacks", "concept": [ "optimization" ], "topics": [ "ggml_fit_opt_multi" ] }, { "page": "ggml_fit", "title": "Train a Model (dispatcher)", "topics": [ "ggml_fit", "ggml_fit.default", "ggml_fit.ggml_functional_model", "ggml_fit.ggml_sequential_model" ] }, { "page": "ggml_flash_attn_back", "title": "Flash Attention Backward (Graph)", "topics": [ "ggml_flash_attn_back" ] }, { "page": "ggml_flash_attn_ext", "title": "Flash Attention (Graph)", "topics": [ "ggml_flash_attn_ext" ] }, { "page": "ggml_flash_attn_ext_add_sinks", "title": "Attach Attention Sinks to a Flash-Attention Node", "topics": [ "ggml_flash_attn_ext_add_sinks" ] }, { "page": "ggml_flash_attn_ext_get_prec", "title": "Get Flash-Attention Precision", "topics": [ "ggml_flash_attn_ext_get_prec" ] }, { "page": "ggml_flash_attn_ext_set_prec", "title": "Set Flash-Attention Precision", "topics": [ "ggml_flash_attn_ext_set_prec" ] }, { "page": "ggml_floor", "title": "Floor (Graph)", "topics": [ "ggml_floor" ] }, { "page": "ggml_floor_inplace", "title": "Floor In-place (Graph)", "topics": [ "ggml_floor_inplace" ] }, { "page": "ggml_free", "title": "Free GGML context", "topics": [ "ggml_free" ] }, { "page": "ggml_freeze_bn_stats", "title": "Fix or Release Batch-Norm Statistics", "topics": [ "ggml_freeze_bn_stats", "ggml_unfreeze_bn_stats" ] }, { "page": "ggml_freeze_weights", "title": "Freeze Layer Weights", "topics": [ "ggml_freeze_weights" ] }, { "page": "ggml_ftype_to_ggml_type", "title": "Convert ftype to ggml_type", "concept": [ "type_system" ], "topics": [ "ggml_ftype_to_ggml_type" ] }, { "page": "ggml_gallocr_alloc_graph", "title": "Allocate Memory for Graph", "topics": [ "ggml_gallocr_alloc_graph" ] }, { "page": "ggml_gallocr_free", "title": "Free Graph Allocator", "topics": [ "ggml_gallocr_free" ] }, { "page": "ggml_gallocr_get_buffer_size", "title": "Get Graph Allocator Buffer Size", "topics": [ "ggml_gallocr_get_buffer_size" ] }, { "page": "ggml_gallocr_new", "title": "Create Graph Allocator", "topics": [ "ggml_gallocr_new" ] }, { "page": "ggml_gallocr_new_buft", "title": "Create a Graph Allocator for a Specific Buffer Type", "topics": [ "ggml_gallocr_new_buft" ] }, { "page": "ggml_gallocr_reserve", "title": "Reserve Memory for Graph", "topics": [ "ggml_gallocr_reserve" ] }, { "page": "ggml_gated_linear_attn", "title": "Gated Linear Attention", "concept": [ "state-space" ], "topics": [ "ggml_gated_linear_attn" ] }, { "page": "ggml_gated_linear_attn_back", "title": "Backward Pass of Gated Linear Attention", "concept": [ "state-space" ], "topics": [ "ggml_gated_linear_attn_back" ] }, { "page": "ggml_geglu", "title": "GeGLU (GELU Gated Linear Unit) (Graph)", "topics": [ "ggml_geglu" ] }, { "page": "ggml_geglu_quick", "title": "GeGLU Quick (Fast GeGLU) (Graph)", "topics": [ "ggml_geglu_quick" ] }, { "page": "ggml_geglu_split", "title": "GeGLU Split (Graph)", "topics": [ "ggml_geglu_split" ] }, { "page": "ggml_gelu", "title": "GELU Activation (Graph)", "topics": [ "ggml_gelu" ] }, { "page": "ggml_gelu_back", "title": "GELU Backward (Graph)", "topics": [ "ggml_gelu_back" ] }, { "page": "ggml_gelu_erf", "title": "Exact GELU Activation (Graph)", "topics": [ "ggml_gelu_erf" ] }, { "page": "ggml_gelu_inplace", "title": "GELU Activation In-place (Graph)", "topics": [ "ggml_gelu_inplace" ] }, { "page": "ggml_gelu_quick", "title": "GELU Quick Activation (Graph)", "topics": [ "ggml_gelu_quick" ] }, { "page": "ggml_get_f32", "title": "Get F32 data", "topics": [ "ggml_get_f32" ] }, { "page": "ggml_get_f32_nd", "title": "Get Single Float Value by N-D Index", "topics": [ "ggml_get_f32_nd" ] }, { "page": "ggml_get_first_tensor", "title": "Get First Tensor from Context", "topics": [ "ggml_get_first_tensor" ] }, { "page": "ggml_get_i32", "title": "Get I32 Data", "topics": [ "ggml_get_i32" ] }, { "page": "ggml_get_i32_nd", "title": "Get Single Int32 Value by N-D Index", "topics": [ "ggml_get_i32_nd" ] }, { "page": "ggml_get_layer", "title": "Get a Layer from a Sequential Model", "topics": [ "ggml_get_layer" ] }, { "page": "ggml_get_max_tensor_size", "title": "Get Maximum Tensor Size", "topics": [ "ggml_get_max_tensor_size" ] }, { "page": "ggml_get_mem_size", "title": "Get Context Memory Size", "topics": [ "ggml_get_mem_size" ] }, { "page": "ggml_get_n_threads", "title": "Get Number of Threads", "topics": [ "ggml_get_n_threads" ] }, { "page": "ggml_get_name", "title": "Get Tensor Name", "topics": [ "ggml_get_name" ] }, { "page": "ggml_get_next_tensor", "title": "Get Next Tensor from Context", "topics": [ "ggml_get_next_tensor" ] }, { "page": "ggml_get_no_alloc", "title": "Get No Allocation Mode", "topics": [ "ggml_get_no_alloc" ] }, { "page": "ggml_get_op_params", "title": "Get Tensor Operation Parameters", "concept": [ "tensor" ], "topics": [ "ggml_get_op_params" ] }, { "page": "ggml_get_op_params_f32", "title": "Get Float Op Parameter", "concept": [ "tensor" ], "topics": [ "ggml_get_op_params_f32" ] }, { "page": "ggml_get_op_params_i32", "title": "Get Integer Op Parameter", "concept": [ "tensor" ], "topics": [ "ggml_get_op_params_i32" ] }, { "page": "ggml_get_rel_pos", "title": "Get Relative Position (Graph)", "topics": [ "ggml_get_rel_pos" ] }, { "page": "ggml_get_rows", "title": "Get Rows by Indices (Graph)", "topics": [ "ggml_get_rows" ] }, { "page": "ggml_get_rows_back", "title": "Get Rows Backward (Graph)", "topics": [ "ggml_get_rows_back" ] }, { "page": "ggml_get_unary_op", "title": "Get Unary Operation from Tensor", "concept": [ "op_info" ], "topics": [ "ggml_get_unary_op" ] }, { "page": "ggml_glu", "title": "Generic GLU (Gated Linear Unit) (Graph)", "topics": [ "ggml_glu" ] }, { "page": "GGML_GLU_OP_REGLU", "title": "GLU Operation Types", "topics": [ "GGML_GLU_OP_GEGLU", "GGML_GLU_OP_GEGLU_ERF", "GGML_GLU_OP_GEGLU_QUICK", "GGML_GLU_OP_REGLU", "GGML_GLU_OP_SWIGLU", "GGML_GLU_OP_SWIGLU_OAI" ] }, { "page": "ggml_glu_split", "title": "Generic GLU Split (Graph)", "topics": [ "ggml_glu_split" ] }, { "page": "ggml_graph_add_node", "title": "Append a Node to a Computation Graph", "topics": [ "ggml_graph_add_node" ] }, { "page": "ggml_graph_clear", "title": "Clear a Computation Graph", "topics": [ "ggml_graph_clear" ] }, { "page": "ggml_graph_compute", "title": "Compute graph", "topics": [ "ggml_graph_compute" ] }, { "page": "ggml_graph_compute_with_ctx", "title": "Compute Graph with Context (Alternative Method)", "topics": [ "ggml_graph_compute_with_ctx" ] }, { "page": "ggml_graph_cpy", "title": "Copy One Graph's Nodes Into Another", "topics": [ "ggml_graph_cpy" ] }, { "page": "ggml_graph_dump_dot", "title": "Export Graph to DOT Format", "topics": [ "ggml_graph_dump_dot" ] }, { "page": "ggml_graph_dup", "title": "Duplicate a Computation Graph", "topics": [ "ggml_graph_dup" ] }, { "page": "ggml_graph_expand", "title": "Add Another Root to an Existing Computation Graph", "topics": [ "ggml_graph_expand" ] }, { "page": "ggml_graph_get_grad", "title": "Get the Gradient Tensor of a Node", "topics": [ "ggml_graph_get_grad", "ggml_graph_get_grad_acc" ] }, { "page": "ggml_graph_get_tensor", "title": "Get Tensor from Graph by Name", "topics": [ "ggml_graph_get_tensor" ] }, { "page": "ggml_graph_n_nodes", "title": "Get Number of Nodes in Graph", "topics": [ "ggml_graph_n_nodes" ] }, { "page": "ggml_graph_node", "title": "Get Graph Node", "topics": [ "ggml_graph_node" ] }, { "page": "ggml_graph_overhead", "title": "Get Graph Overhead", "topics": [ "ggml_graph_overhead" ] }, { "page": "ggml_graph_print", "title": "Print Graph Information", "topics": [ "ggml_graph_print" ] }, { "page": "ggml_graph_reset", "title": "Reset Graph (for backpropagation)", "topics": [ "ggml_graph_reset" ] }, { "page": "ggml_graph_view", "title": "Create a View of a Subgraph", "concept": [ "graph" ], "topics": [ "ggml_graph_view" ] }, { "page": "ggml_group_norm", "title": "Group Normalization (Graph)", "topics": [ "ggml_group_norm" ] }, { "page": "ggml_group_norm_inplace", "title": "Group Normalization In-place (Graph)", "topics": [ "ggml_group_norm_inplace" ] }, { "page": "ggml_gru", "title": "Create a GRU Layer Object", "topics": [ "ggml_gru" ] }, { "page": "ggml_hardsigmoid", "title": "Hard Sigmoid Activation (Graph)", "topics": [ "ggml_hardsigmoid" ] }, { "page": "ggml_hardswish", "title": "Hard Swish Activation (Graph)", "topics": [ "ggml_hardswish" ] }, { "page": "ggml_im2col", "title": "Image to Column (Graph)", "topics": [ "ggml_im2col" ] }, { "page": "ggml_init", "title": "Initialize GGML context", "topics": [ "ggml_init" ] }, { "page": "ggml_init_auto", "title": "Create Context with Auto-sizing", "topics": [ "ggml_init_auto" ] }, { "page": "ggml_inject", "title": "Inject a single-cell result back into its container", "topics": [ "ggml_inject", "ggml_inject.Seurat", "ggml_inject.SingleCellExperiment" ] }, { "page": "ggml_input", "title": "Declare a Functional API Input Tensor", "topics": [ "ggml_input" ] }, { "page": "ggml_is_available", "title": "Check if GGML is available", "topics": [ "ggml_is_available" ] }, { "page": "ggml_is_contiguous", "title": "Check if Tensor is Contiguous", "topics": [ "ggml_is_contiguous" ] }, { "page": "ggml_is_contiguous_0", "title": "Check Tensor Contiguity (Dimension 0)", "concept": [ "tensor_layout" ], "topics": [ "ggml_is_contiguous_0" ] }, { "page": "ggml_is_contiguous_1", "title": "Check Tensor Contiguity (Dimensions >= 1)", "concept": [ "tensor_layout" ], "topics": [ "ggml_is_contiguous_1" ] }, { "page": "ggml_is_contiguous_2", "title": "Check Tensor Contiguity (Dimensions >= 2)", "concept": [ "tensor_layout" ], "topics": [ "ggml_is_contiguous_2" ] }, { "page": "ggml_is_contiguous_channels", "title": "Check Channel-wise Contiguity", "concept": [ "tensor_layout" ], "topics": [ "ggml_is_contiguous_channels" ] }, { "page": "ggml_is_contiguous_rows", "title": "Check Row-wise Contiguity", "concept": [ "tensor_layout" ], "topics": [ "ggml_is_contiguous_rows" ] }, { "page": "ggml_is_contiguously_allocated", "title": "Check If Tensor is Contiguously Allocated", "concept": [ "tensor_layout" ], "topics": [ "ggml_is_contiguously_allocated" ] }, { "page": "ggml_is_permuted", "title": "Check if Tensor is Permuted", "topics": [ "ggml_is_permuted" ] }, { "page": "ggml_is_quantized", "title": "Check If Type is Quantized", "concept": [ "type_system" ], "topics": [ "ggml_is_quantized" ] }, { "page": "ggml_is_transposed", "title": "Check if Tensor is Transposed", "topics": [ "ggml_is_transposed" ] }, { "page": "ggml_l2_norm", "title": "L2 Normalization (Graph)", "topics": [ "ggml_l2_norm" ] }, { "page": "ggml_l2_norm_inplace", "title": "L2 Normalization In-place (Graph)", "topics": [ "ggml_l2_norm_inplace" ] }, { "page": "ggml_layer_add", "title": "Element-wise Addition of Two Tensor Nodes", "topics": [ "ggml_layer_add" ] }, { "page": "ggml_layer_attention", "title": "Add a Multi-Head Attention Layer", "topics": [ "ggml_layer_attention" ] }, { "page": "ggml_layer_batch_norm", "title": "Add Batch Normalization Layer", "topics": [ "ggml_layer_batch_norm" ] }, { "page": "ggml_layer_concatenate", "title": "Concatenate Tensor Nodes Along an Axis", "topics": [ "ggml_layer_concatenate" ] }, { "page": "ggml_layer_conv_1d", "title": "Add 1D Convolution Layer", "topics": [ "ggml_layer_conv_1d" ] }, { "page": "ggml_layer_conv_2d", "title": "Add 2D Convolution Layer", "topics": [ "ggml_layer_conv_2d" ] }, { "page": "ggml_layer_custom", "title": "Define a Custom Layer From an R Function", "topics": [ "ggml_layer_custom" ] }, { "page": "ggml_layer_dense", "title": "Add Dense (Fully Connected) Layer", "topics": [ "ggml_layer_dense" ] }, { "page": "ggml_layer_dropout", "title": "Add Dropout Layer", "topics": [ "ggml_layer_dropout" ] }, { "page": "ggml_layer_embedding", "title": "Add Embedding Layer", "topics": [ "ggml_layer_embedding" ] }, { "page": "ggml_layer_flatten", "title": "Add Flatten Layer", "topics": [ "ggml_layer_flatten" ] }, { "page": "ggml_layer_global_average_pooling_2d", "title": "Global Average Pooling for 2D Feature Maps", "topics": [ "ggml_layer_global_average_pooling_2d" ] }, { "page": "ggml_layer_global_max_pooling_2d", "title": "Global Max Pooling for 2D Feature Maps", "topics": [ "ggml_layer_global_max_pooling_2d" ] }, { "page": "ggml_layer_gru", "title": "Add a GRU Layer", "topics": [ "ggml_layer_gru" ] }, { "page": "ggml_layer_layer_norm", "title": "Add Layer Normalization Layer", "topics": [ "ggml_layer_layer_norm" ] }, { "page": "ggml_layer_lstm", "title": "Add an LSTM Layer", "topics": [ "ggml_layer_lstm" ] }, { "page": "ggml_layer_max_pooling_2d", "title": "Add 2D Max Pooling Layer", "topics": [ "ggml_layer_max_pooling_2d" ] }, { "page": "ggml_layer_permute", "title": "Add Permute Layer", "topics": [ "ggml_layer_permute" ] }, { "page": "ggml_layer_positional_embedding", "title": "Add a Learned Positional Embedding", "topics": [ "ggml_layer_positional_embedding" ] }, { "page": "ggml_layer_reshape", "title": "Add Reshape Layer", "topics": [ "ggml_layer_reshape" ] }, { "page": "ggml_layer_rms_norm", "title": "Add RMS Normalization Layer", "topics": [ "ggml_layer_rms_norm" ] }, { "page": "ggml_layer_sequence_pooling", "title": "Pool a Sequence Down to One Vector", "topics": [ "ggml_layer_sequence_pooling" ] }, { "page": "ggml_layer_transformer_block", "title": "Add a Transformer Encoder Block", "topics": [ "ggml_layer_transformer_block" ] }, { "page": "ggml_leaky_relu", "title": "Leaky ReLU Activation (Graph)", "topics": [ "ggml_leaky_relu" ] }, { "page": "ggml_load_model", "title": "Load a Full Model (Architecture + Weights)", "topics": [ "ggml_load_model" ] }, { "page": "ggml_load_weights", "title": "Load Model Weights from File", "topics": [ "ggml_load_weights" ] }, { "page": "ggml_log", "title": "Natural Logarithm (Graph)", "topics": [ "ggml_log" ] }, { "page": "ggml_log_inplace", "title": "Natural Logarithm In-place (Graph)", "topics": [ "ggml_log_inplace" ] }, { "page": "ggml_log_is_r_enabled", "title": "Check if R Logging is Enabled", "concept": [ "logging" ], "topics": [ "ggml_log_is_r_enabled" ] }, { "page": "ggml_log_set_default", "title": "Restore Default GGML Logging", "concept": [ "logging" ], "topics": [ "ggml_log_set_default" ] }, { "page": "ggml_log_set_r", "title": "Enable R-compatible GGML Logging", "concept": [ "logging" ], "topics": [ "ggml_log_set_r" ] }, { "page": "ggml_lstm", "title": "Create an LSTM Layer Object", "topics": [ "ggml_lstm" ] }, { "page": "ggml_marshal_model", "title": "Marshal a ggmlR model to an in-memory container", "topics": [ "ggml_marshal_model" ] }, { "page": "ggml_matmul", "title": "GPU matrix multiply (drop-in for '%*%')", "topics": [ "ggml_matmul" ] }, { "page": "ggml_matmul_f64", "title": "GPU double-precision matrix multiply", "topics": [ "ggml_matmul_f64" ] }, { "page": "ggml_matrix-class", "title": "A GPU-backed matrix wrapper", "topics": [ "%*%,ANY,ggml_matrix-method", "%*%,ggml_matrix,ANY-method", "%*%,ggml_matrix,ggml_matrix-method", "crossprod,ggml_matrix-method", "dim,ggml_matrix-method", "ggml_matrix-class", "show,ggml_matrix-method", "tcrossprod,ggml_matrix-method" ] }, { "page": "ggml_mean", "title": "Mean (Graph)", "topics": [ "ggml_mean" ] }, { "page": "ggml_model", "title": "Create a Functional Model", "topics": [ "ggml_model" ] }, { "page": "ggml_model_backend", "title": "Backend a fitted ggml model actually ran on", "topics": [ "ggml_model_backend" ] }, { "page": "ggml_model_sequential", "title": "Create a Sequential Neural Network Model", "topics": [ "ggml_model_sequential" ] }, { "page": "ggml_mul", "title": "Multiply tensors", "topics": [ "ggml_mul" ] }, { "page": "ggml_mul_inplace", "title": "Element-wise Multiplication In-place (Graph)", "topics": [ "ggml_mul_inplace" ] }, { "page": "ggml_mul_mat", "title": "Matrix Multiplication (Graph)", "topics": [ "ggml_mul_mat" ] }, { "page": "ggml_mul_mat_id", "title": "Matrix Multiplication with Expert Selection (Graph)", "topics": [ "ggml_mul_mat_id" ] }, { "page": "ggml_n_dims", "title": "Get Number of Dimensions", "topics": [ "ggml_n_dims" ] }, { "page": "GGML_N_TASKS_MAX", "title": "Maximum Task Count for Custom Ops", "topics": [ "GGML_N_TASKS_MAX" ] }, { "page": "ggml_nbytes", "title": "Get number of bytes", "topics": [ "ggml_nbytes" ] }, { "page": "ggml_neg", "title": "Negation (Graph)", "topics": [ "ggml_neg" ] }, { "page": "ggml_neg_inplace", "title": "Negation In-place (Graph)", "topics": [ "ggml_neg_inplace" ] }, { "page": "ggml_nelements", "title": "Get number of elements", "topics": [ "ggml_nelements" ] }, { "page": "ggml_new_f32", "title": "Create Scalar F32 Tensor", "topics": [ "ggml_new_f32" ] }, { "page": "ggml_new_i32", "title": "Create Scalar I32 Tensor", "topics": [ "ggml_new_i32" ] }, { "page": "ggml_new_tensor", "title": "Create Tensor with Arbitrary Dimensions", "topics": [ "ggml_new_tensor" ] }, { "page": "ggml_new_tensor_1d", "title": "Create 1D tensor", "topics": [ "ggml_new_tensor_1d" ] }, { "page": "ggml_new_tensor_2d", "title": "Create 2D tensor", "topics": [ "ggml_new_tensor_2d" ] }, { "page": "ggml_new_tensor_3d", "title": "Create 3D Tensor", "topics": [ "ggml_new_tensor_3d" ] }, { "page": "ggml_new_tensor_4d", "title": "Create 4D Tensor", "topics": [ "ggml_new_tensor_4d" ] }, { "page": "ggml_norm", "title": "Layer Normalization (Graph)", "topics": [ "ggml_norm" ] }, { "page": "ggml_norm_back", "title": "Layer Norm Backward (Graph)", "topics": [ "ggml_norm_back" ] }, { "page": "ggml_norm_inplace", "title": "Layer Normalization In-place (Graph)", "topics": [ "ggml_norm_inplace" ] }, { "page": "ggml_nrows", "title": "Get Number of Rows", "topics": [ "ggml_nrows" ] }, { "page": "ggml_op_can_inplace", "title": "Check if Operation Can Be Done In-place", "concept": [ "graph" ], "topics": [ "ggml_op_can_inplace" ] }, { "page": "ggml_op_desc", "title": "Get Operation Description from Tensor", "concept": [ "op_info" ], "topics": [ "ggml_op_desc" ] }, { "page": "ggml_op_name", "title": "Get Operation Name", "concept": [ "op_info" ], "topics": [ "ggml_op_name" ] }, { "page": "ggml_op_symbol", "title": "Get Operation Symbol", "concept": [ "op_info" ], "topics": [ "ggml_op_symbol" ] }, { "page": "ggml_ops_registry", "title": "Supported single-cell operations", "topics": [ "ggml_ops_registry" ] }, { "page": "ggml_opt_alloc", "title": "Allocate graph for evaluation", "concept": [ "optimization" ], "topics": [ "ggml_opt_alloc" ] }, { "page": "ggml_opt_context_optimizer_type", "title": "Get optimizer type from context", "concept": [ "optimization" ], "topics": [ "ggml_opt_context_optimizer_type" ] }, { "page": "ggml_opt_dataset_data", "title": "Get data tensor from dataset", "concept": [ "optimization" ], "topics": [ "ggml_opt_dataset_data" ] }, { "page": "ggml_opt_dataset_free", "title": "Free optimization dataset", "concept": [ "optimization" ], "topics": [ "ggml_opt_dataset_free" ] }, { "page": "ggml_opt_dataset_get_batch", "title": "Get batch from dataset", "concept": [ "optimization" ], "topics": [ "ggml_opt_dataset_get_batch" ] }, { "page": "ggml_opt_dataset_get_batch_head", "title": "Get one output head's slice of a batch's labels", "concept": [ "optimization" ], "topics": [ "ggml_opt_dataset_get_batch_head" ] }, { "page": "ggml_opt_dataset_init", "title": "Create a new optimization dataset", "concept": [ "optimization" ], "topics": [ "ggml_opt_dataset_init" ] }, { "page": "ggml_opt_dataset_labels", "title": "Get labels tensor from dataset", "concept": [ "optimization" ], "topics": [ "ggml_opt_dataset_labels" ] }, { "page": "ggml_opt_dataset_ndata", "title": "Get number of datapoints in dataset", "concept": [ "optimization" ], "topics": [ "ggml_opt_dataset_ndata" ] }, { "page": "ggml_opt_dataset_shuffle", "title": "Shuffle dataset", "concept": [ "optimization" ], "topics": [ "ggml_opt_dataset_shuffle" ] }, { "page": "ggml_opt_dataset_weights", "title": "Get dataset loss weights tensor", "concept": [ "optimization" ], "topics": [ "ggml_opt_dataset_weights" ] }, { "page": "ggml_opt_default_params", "title": "Get default optimizer parameters", "concept": [ "optimization" ], "topics": [ "ggml_opt_default_params" ] }, { "page": "ggml_opt_epoch", "title": "Run one training epoch", "concept": [ "optimization" ], "topics": [ "ggml_opt_epoch" ] }, { "page": "ggml_opt_eval", "title": "Evaluate model", "concept": [ "optimization" ], "topics": [ "ggml_opt_eval" ] }, { "page": "ggml_opt_fit", "title": "Fit model to dataset", "concept": [ "optimization" ], "topics": [ "ggml_opt_fit" ] }, { "page": "ggml_opt_fit_multi", "title": "Fit a multi-output model", "concept": [ "optimization" ], "topics": [ "ggml_opt_fit_multi" ] }, { "page": "ggml_opt_free", "title": "Free optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_free" ] }, { "page": "ggml_opt_get_lr", "title": "Get current learning rate from optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_get_lr" ] }, { "page": "ggml_opt_grad_acc", "title": "Get gradient accumulator for a tensor", "concept": [ "optimization" ], "topics": [ "ggml_opt_grad_acc" ] }, { "page": "ggml_opt_init", "title": "Initialize optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_init" ] }, { "page": "ggml_opt_init_for_fit", "title": "Initialize optimizer context for R-side epoch loop", "concept": [ "optimization" ], "topics": [ "ggml_opt_init_for_fit" ] }, { "page": "ggml_opt_init_for_fit_multi", "title": "Initialize a multi-output optimizer context for an R-side epoch loop", "concept": [ "optimization" ], "topics": [ "ggml_opt_init_for_fit_multi" ] }, { "page": "ggml_opt_inputs", "title": "Get inputs tensor from optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_inputs" ] }, { "page": "ggml_opt_labels", "title": "Get labels tensor from optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_labels" ] }, { "page": "ggml_opt_loss", "title": "Get loss tensor from optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_loss" ] }, { "page": "ggml_opt_loss_i", "title": "Get one head's loss scalar, before weighting", "concept": [ "optimization" ], "topics": [ "ggml_opt_loss_i" ] }, { "page": "ggml_opt_loss_type_binary_cross_entropy", "title": "Loss type: Binary Cross-Entropy", "concept": [ "optimization" ], "topics": [ "ggml_opt_loss_type_binary_cross_entropy" ] }, { "page": "ggml_opt_loss_type_cross_entropy", "title": "Loss type: Cross Entropy", "concept": [ "optimization" ], "topics": [ "ggml_opt_loss_type_cross_entropy" ] }, { "page": "ggml_opt_loss_type_huber", "title": "Loss type: Huber (smooth L1)", "concept": [ "optimization" ], "topics": [ "ggml_opt_loss_type_huber" ] }, { "page": "ggml_opt_loss_type_mae", "title": "Loss type: Mean Absolute Error", "concept": [ "optimization" ], "topics": [ "ggml_opt_loss_type_mae" ] }, { "page": "ggml_opt_loss_type_mean", "title": "Loss type: Mean", "concept": [ "optimization" ], "topics": [ "ggml_opt_loss_type_mean" ] }, { "page": "ggml_opt_loss_type_mse", "title": "Loss type: Mean Squared Error", "concept": [ "optimization" ], "topics": [ "ggml_opt_loss_type_mse" ] }, { "page": "ggml_opt_loss_type_sum", "title": "Loss type: Sum", "concept": [ "optimization" ], "topics": [ "ggml_opt_loss_type_sum" ] }, { "page": "ggml_opt_loss_type_weighted_mse", "title": "Loss type: Weighted Mean Squared Error", "concept": [ "optimization" ], "topics": [ "ggml_opt_loss_type_weighted_mse" ] }, { "page": "ggml_opt_n_loss", "title": "Number of output heads in an optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_n_loss" ] }, { "page": "ggml_opt_ncorrect", "title": "Get number of correct predictions tensor", "concept": [ "optimization" ], "topics": [ "ggml_opt_ncorrect" ] }, { "page": "ggml_opt_optimizer_name", "title": "Get optimizer name", "concept": [ "optimization" ], "topics": [ "ggml_opt_optimizer_name" ] }, { "page": "ggml_opt_optimizer_type_adamw", "title": "Optimizer type: AdamW", "concept": [ "optimization" ], "topics": [ "ggml_opt_optimizer_type_adamw" ] }, { "page": "ggml_opt_optimizer_type_sgd", "title": "Optimizer type: SGD", "concept": [ "optimization" ], "topics": [ "ggml_opt_optimizer_type_sgd" ] }, { "page": "ggml_opt_outputs", "title": "Get outputs tensor from optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_outputs" ] }, { "page": "ggml_opt_outputs_i", "title": "Get one output head's tensors from an optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_labels_i", "ggml_opt_loss_weights_i", "ggml_opt_ncorrect_i", "ggml_opt_outputs_i", "ggml_opt_pred_i" ] }, { "page": "ggml_opt_pred", "title": "Get predictions tensor from optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_pred" ] }, { "page": "ggml_opt_prepare_alloc", "title": "Prepare allocation for non-static graphs", "concept": [ "optimization" ], "topics": [ "ggml_opt_prepare_alloc" ] }, { "page": "ggml_opt_reset", "title": "Reset optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_reset" ] }, { "page": "ggml_opt_result_accuracy", "title": "Get accuracy from result", "concept": [ "optimization" ], "topics": [ "ggml_opt_result_accuracy" ] }, { "page": "ggml_opt_result_free", "title": "Free optimization result", "concept": [ "optimization" ], "topics": [ "ggml_opt_result_free" ] }, { "page": "ggml_opt_result_init", "title": "Initialize optimization result", "concept": [ "optimization" ], "topics": [ "ggml_opt_result_init" ] }, { "page": "ggml_opt_result_loss", "title": "Get loss from result", "concept": [ "optimization" ], "topics": [ "ggml_opt_result_loss" ] }, { "page": "ggml_opt_result_loss_i", "title": "Get one output head's loss or accuracy from a result", "concept": [ "optimization" ], "topics": [ "ggml_opt_result_accuracy_i", "ggml_opt_result_loss_i" ] }, { "page": "ggml_opt_result_n_loss", "title": "Number of output heads accumulated in a result", "concept": [ "optimization" ], "topics": [ "ggml_opt_result_n_loss" ] }, { "page": "ggml_opt_result_ndata", "title": "Get number of datapoints from result", "concept": [ "optimization" ], "topics": [ "ggml_opt_result_ndata" ] }, { "page": "ggml_opt_result_pred", "title": "Get predictions from result", "concept": [ "optimization" ], "topics": [ "ggml_opt_result_pred" ] }, { "page": "ggml_opt_result_reset", "title": "Reset optimization result", "concept": [ "optimization" ], "topics": [ "ggml_opt_result_reset" ] }, { "page": "ggml_opt_set_lr", "title": "Set learning rate in optimizer context", "concept": [ "optimization" ], "topics": [ "ggml_opt_set_lr" ] }, { "page": "ggml_opt_static_graphs", "title": "Check if using static graphs", "concept": [ "optimization" ], "topics": [ "ggml_opt_static_graphs" ] }, { "page": "ggml_opt_step_adamw", "title": "AdamW Optimizer Step (Graph Node)", "topics": [ "ggml_opt_step_adamw" ] }, { "page": "ggml_out_prod", "title": "Outer Product (Graph)", "topics": [ "ggml_out_prod" ] }, { "page": "ggml_pad", "title": "Pad Tensor with Zeros (Graph)", "topics": [ "ggml_pad" ] }, { "page": "ggml_pad_ext", "title": "Pad a tensor with zeros on both sides", "topics": [ "ggml_pad_ext" ] }, { "page": "ggml_pad_reflect_1d", "title": "Reflective 1D Padding (Graph)", "topics": [ "ggml_pad_reflect_1d" ] }, { "page": "ggml_permute", "title": "Permute Tensor Dimensions (Graph)", "topics": [ "ggml_permute" ] }, { "page": "ggml_pool_1d", "title": "1D Pooling (Graph)", "topics": [ "GGML_OP_POOL_AVG", "GGML_OP_POOL_MAX", "ggml_pool_1d" ] }, { "page": "ggml_pool_2d", "title": "2D Pooling (Graph)", "topics": [ "ggml_pool_2d" ] }, { "page": "ggml_pop_layer", "title": "Remove the Last Layer from a Sequential Model", "topics": [ "ggml_pop_layer" ] }, { "page": "ggml_pp_dp_forward", "title": "Pipeline + data parallel forward pass (PP x DP)", "topics": [ "ggml_pp_dp_forward" ] }, { "page": "ggml_pp_forward", "title": "Pipeline-parallel forward pass across per-device layer stages", "topics": [ "ggml_pp_forward" ] }, { "page": "ggml_predict_classes", "title": "Predict Classes from a Trained Model", "topics": [ "ggml_predict_classes" ] }, { "page": "ggml_predict", "title": "Get Predictions from a Trained Model", "topics": [ "ggml_predict", "ggml_predict.ggml_functional_model", "ggml_predict.ggml_sequential_model" ] }, { "page": "ggml_print_mem_status", "title": "Print Context Memory Status", "topics": [ "ggml_print_mem_status" ] }, { "page": "ggml_print_objects", "title": "Print Objects in Context", "topics": [ "ggml_print_objects" ] }, { "page": "ggml_quant_block_info", "title": "Get Quantization Block Info", "concept": [ "quantization" ], "topics": [ "ggml_quant_block_info" ] }, { "page": "ggml_quantize_chunk", "title": "Quantize Data Chunk", "topics": [ "ggml_quantize_chunk" ] }, { "page": "ggml_quantize_free", "title": "Free Quantization Resources", "topics": [ "ggml_quantize_free" ] }, { "page": "ggml_quantize_init", "title": "Initialize Quantization Tables", "topics": [ "ggml_quantize_init" ] }, { "page": "ggml_quantize_requires_imatrix", "title": "Check if Quantization Requires Importance Matrix", "topics": [ "ggml_quantize_requires_imatrix" ] }, { "page": "ggml_reglu", "title": "ReGLU (ReLU Gated Linear Unit) (Graph)", "topics": [ "ggml_reglu" ] }, { "page": "ggml_reglu_split", "title": "ReGLU Split (Graph)", "topics": [ "ggml_reglu_split" ] }, { "page": "ggml_relu", "title": "ReLU Activation (Graph)", "topics": [ "ggml_relu" ] }, { "page": "ggml_relu_inplace", "title": "ReLU Activation In-place (Graph)", "topics": [ "ggml_relu_inplace" ] }, { "page": "ggml_repeat", "title": "Repeat (Graph)", "topics": [ "ggml_repeat" ] }, { "page": "ggml_repeat_back", "title": "Repeat Backward (Graph)", "topics": [ "ggml_repeat_back" ] }, { "page": "ggml_reset", "title": "Reset GGML Context", "topics": [ "ggml_reset" ] }, { "page": "ggml_reshape_1d", "title": "Reshape to 1D (Graph)", "topics": [ "ggml_reshape_1d" ] }, { "page": "ggml_reshape_2d", "title": "Reshape to 2D (Graph)", "topics": [ "ggml_reshape_2d" ] }, { "page": "ggml_reshape_3d", "title": "Reshape to 3D (Graph)", "topics": [ "ggml_reshape_3d" ] }, { "page": "ggml_reshape_4d", "title": "Reshape to 4D (Graph)", "topics": [ "ggml_reshape_4d" ] }, { "page": "ggml_result", "title": "Construct a single-cell result", "topics": [ "ggml_result" ] }, { "page": "ggml_rms_norm", "title": "RMS Normalization (Graph)", "topics": [ "ggml_rms_norm" ] }, { "page": "ggml_rms_norm_back", "title": "RMS Norm Backward (Graph)", "topics": [ "ggml_rms_norm_back" ] }, { "page": "ggml_rms_norm_inplace", "title": "RMS Normalization In-place (Graph)", "topics": [ "ggml_rms_norm_inplace" ] }, { "page": "ggml_roll", "title": "Roll (Graph)", "topics": [ "ggml_roll" ] }, { "page": "ggml_rope", "title": "Rotary Position Embedding (Graph)", "topics": [ "ggml_rope" ] }, { "page": "ggml_rope_ext", "title": "Extended RoPE with Frequency Scaling (Graph)", "topics": [ "ggml_rope_ext" ] }, { "page": "ggml_rope_ext_back", "title": "RoPE Extended Backward (Graph)", "topics": [ "ggml_rope_ext_back" ] }, { "page": "ggml_rope_ext_inplace", "title": "Extended RoPE Inplace (Graph)", "concept": [ "rope" ], "topics": [ "ggml_rope_ext_inplace" ] }, { "page": "ggml_rope_inplace", "title": "Rotary Position Embedding In-place (Graph)", "topics": [ "ggml_rope_inplace" ] }, { "page": "ggml_rope_multi", "title": "Multi-RoPE for Vision Models (Graph)", "concept": [ "rope" ], "topics": [ "ggml_rope_multi" ] }, { "page": "ggml_rope_multi_back", "title": "Multi-RoPE Backward (Graph)", "topics": [ "ggml_rope_multi_back" ] }, { "page": "ggml_rope_multi_inplace", "title": "Multi-RoPE Inplace (Graph)", "concept": [ "rope" ], "topics": [ "ggml_rope_multi_inplace" ] }, { "page": "ggml_round", "title": "Round (Graph)", "topics": [ "ggml_round" ] }, { "page": "ggml_round_inplace", "title": "Round In-place (Graph)", "topics": [ "ggml_round_inplace" ] }, { "page": "ggml_run", "title": "Run a single-cell task on the GGML backend", "topics": [ "ggml_run" ] }, { "page": "ggml_rwkv_output", "title": "Views Into a Packed RWKV / GLA Result", "concept": [ "state-space" ], "topics": [ "ggml_rwkv_output", "ggml_rwkv_state" ] }, { "page": "ggml_rwkv_wkv6", "title": "RWKV-6 WKV Recurrence", "concept": [ "state-space" ], "topics": [ "ggml_rwkv_wkv6" ] }, { "page": "ggml_rwkv_wkv6_back", "title": "Backward Pass of the RWKV-6 WKV Recurrence", "concept": [ "state-space" ], "topics": [ "ggml_rwkv_wkv6_back" ] }, { "page": "ggml_rwkv_wkv7", "title": "RWKV-7 WKV Recurrence", "concept": [ "state-space" ], "topics": [ "ggml_rwkv_wkv7" ] }, { "page": "ggml_rwkv_wkv7_back", "title": "Backward Pass of the RWKV-7 WKV Recurrence", "concept": [ "state-space" ], "topics": [ "ggml_rwkv_wkv7_back" ] }, { "page": "ggml_save_model", "title": "Save a Full Model (Architecture + Weights)", "topics": [ "ggml_save_model" ] }, { "page": "ggml_save_weights", "title": "Save Model Weights to File", "topics": [ "ggml_save_weights" ] }, { "page": "ggml_scale", "title": "Scale (Graph)", "topics": [ "ggml_scale" ] }, { "page": "ggml_scale_bias", "title": "Scale and Shift by Constants (Graph)", "topics": [ "ggml_scale_bias" ] }, { "page": "ggml_scale_inplace", "title": "Scale Tensor In-place (Graph)", "topics": [ "ggml_scale_inplace" ] }, { "page": "ggml_scatter_elements", "title": "Scatter elements into a copy of a tensor (ONNX ScatterElements)", "topics": [ "ggml_scatter_elements" ] }, { "page": "ggml_schedule_cosine_decay", "title": "Cosine annealing LR scheduler", "concept": [ "callbacks" ], "topics": [ "ggml_schedule_cosine_decay" ] }, { "page": "ggml_schedule_reduce_on_plateau", "title": "Reduce on plateau LR scheduler", "concept": [ "callbacks" ], "topics": [ "ggml_schedule_reduce_on_plateau" ] }, { "page": "ggml_schedule_step_decay", "title": "Step decay LR scheduler", "concept": [ "callbacks" ], "topics": [ "ggml_schedule_step_decay" ] }, { "page": "ggml_set", "title": "Set Tensor Region (Graph)", "topics": [ "ggml_set" ] }, { "page": "ggml_set_1d", "title": "Set 1D Tensor Region (Graph)", "topics": [ "ggml_set_1d" ] }, { "page": "ggml_set_2d", "title": "Set 2D Tensor Region (Graph)", "topics": [ "ggml_set_2d" ] }, { "page": "ggml_set_abort_callback_default", "title": "Restore Default Abort Behavior", "concept": [ "logging" ], "topics": [ "ggml_set_abort_callback_default" ] }, { "page": "ggml_set_abort_callback_r", "title": "Enable R-compatible Abort Handling", "concept": [ "logging" ], "topics": [ "ggml_set_abort_callback_r" ] }, { "page": "ggml_set_f32", "title": "Set F32 data", "topics": [ "ggml_set_f32" ] }, { "page": "ggml_set_f32_nd", "title": "Set Single Float Value by N-D Index", "topics": [ "ggml_set_f32_nd" ] }, { "page": "ggml_set_i32", "title": "Set I32 Data", "topics": [ "ggml_set_i32" ] }, { "page": "ggml_set_i32_nd", "title": "Set Single Int32 Value by N-D Index", "topics": [ "ggml_set_i32_nd" ] }, { "page": "ggml_set_input", "title": "Mark Tensor as Input", "topics": [ "ggml_set_input" ] }, { "page": "ggml_set_loss", "title": "Mark a Tensor as the Loss", "topics": [ "ggml_set_loss" ] }, { "page": "ggml_set_n_threads", "title": "Set Number of Threads", "topics": [ "ggml_set_n_threads" ] }, { "page": "ggml_set_name", "title": "Set Tensor Name", "topics": [ "ggml_set_name" ] }, { "page": "ggml_set_no_alloc", "title": "Set No Allocation Mode", "topics": [ "ggml_set_no_alloc" ] }, { "page": "ggml_set_omp_threads", "title": "Set OpenMP Thread Count", "topics": [ "ggml_set_omp_threads" ] }, { "page": "ggml_set_op_params", "title": "Set Tensor Operation Parameters", "concept": [ "tensor" ], "topics": [ "ggml_set_op_params" ] }, { "page": "ggml_set_op_params_f32", "title": "Set Float Op Parameter", "concept": [ "tensor" ], "topics": [ "ggml_set_op_params_f32" ] }, { "page": "ggml_set_op_params_i32", "title": "Set Integer Op Parameter", "concept": [ "tensor" ], "topics": [ "ggml_set_op_params_i32" ] }, { "page": "ggml_set_output", "title": "Mark Tensor as Output", "topics": [ "ggml_set_output" ] }, { "page": "ggml_set_param", "title": "Set Tensor as Trainable Parameter", "topics": [ "ggml_set_param" ] }, { "page": "ggml_set_seed", "title": "Set the random seed for reproducible ggmlR runs", "topics": [ "ggml_set_seed" ] }, { "page": "ggml_set_zero", "title": "Set Tensor to Zero", "topics": [ "ggml_set_zero" ] }, { "page": "ggml_sgn", "title": "Sign Function (Graph)", "topics": [ "ggml_sgn" ] }, { "page": "ggml_sigmoid", "title": "Sigmoid Activation (Graph)", "topics": [ "ggml_sigmoid" ] }, { "page": "ggml_sigmoid_inplace", "title": "Sigmoid Activation In-place (Graph)", "topics": [ "ggml_sigmoid_inplace" ] }, { "page": "ggml_silu", "title": "SiLU Activation (Graph)", "topics": [ "ggml_silu" ] }, { "page": "ggml_silu_back", "title": "SiLU Backward (Graph)", "topics": [ "ggml_silu_back" ] }, { "page": "ggml_silu_inplace", "title": "SiLU Activation In-place (Graph)", "topics": [ "ggml_silu_inplace" ] }, { "page": "ggml_sin", "title": "Sine (Graph)", "topics": [ "ggml_sin" ] }, { "page": "ggml_snapshot_free", "title": "Release a Snapshot Made by a Trainer", "concept": [ "optimization" ], "topics": [ "ggml_snapshot_free" ] }, { "page": "ggml_soft_max", "title": "Softmax (Graph)", "topics": [ "ggml_soft_max" ] }, { "page": "ggml_soft_max_add_sinks", "title": "Attach Attention Sinks to a Softmax Node", "topics": [ "ggml_soft_max_add_sinks" ] }, { "page": "ggml_soft_max_ext", "title": "Extended Softmax with Masking and Scaling (Graph)", "topics": [ "ggml_soft_max_ext" ] }, { "page": "ggml_soft_max_ext_back", "title": "Softmax Backward Extended (Graph)", "topics": [ "ggml_soft_max_ext_back" ] }, { "page": "ggml_soft_max_ext_back_inplace", "title": "Extended Softmax Backward Inplace (Graph)", "concept": [ "softmax" ], "topics": [ "ggml_soft_max_ext_back_inplace" ] }, { "page": "ggml_soft_max_ext_inplace", "title": "Extended Softmax Inplace (Graph)", "concept": [ "softmax" ], "topics": [ "ggml_soft_max_ext_inplace" ] }, { "page": "ggml_soft_max_inplace", "title": "Softmax In-place (Graph)", "topics": [ "ggml_soft_max_inplace" ] }, { "page": "ggml_softplus", "title": "Softplus Activation (Graph)", "topics": [ "ggml_softplus" ] }, { "page": "ggml_softplus_inplace", "title": "Softplus Activation In-place (Graph)", "topics": [ "ggml_softplus_inplace" ] }, { "page": "GGML_SORT_ORDER_ASC", "title": "Sort Order Constants", "topics": [ "GGML_SORT_ORDER_ASC", "GGML_SORT_ORDER_DESC" ] }, { "page": "ggml_sqr", "title": "Square (Graph)", "topics": [ "ggml_sqr" ] }, { "page": "ggml_sqr_inplace", "title": "Square In-place (Graph)", "topics": [ "ggml_sqr_inplace" ] }, { "page": "ggml_sqrt", "title": "Square Root (Graph)", "topics": [ "ggml_sqrt" ] }, { "page": "ggml_sqrt_inplace", "title": "Square Root In-place (Graph)", "topics": [ "ggml_sqrt_inplace" ] }, { "page": "ggml_ssm_conv", "title": "State-Space Convolution (Mamba)", "concept": [ "state-space" ], "topics": [ "ggml_ssm_conv" ] }, { "page": "ggml_ssm_conv_back", "title": "Backward Pass of the State-Space Convolution", "concept": [ "state-space" ], "topics": [ "ggml_ssm_conv_back" ] }, { "page": "ggml_ssm_scan", "title": "Selective State-Space Scan (Mamba / Mamba-2)", "concept": [ "state-space" ], "topics": [ "ggml_ssm_scan" ] }, { "page": "ggml_ssm_scan_back", "title": "Backward Pass of the Selective State-Space Scan", "concept": [ "state-space" ], "topics": [ "ggml_ssm_scan_back" ] }, { "page": "ggml_ssm_scan_output", "title": "Views Into a Packed ssm_scan Result", "concept": [ "state-space" ], "topics": [ "ggml_ssm_scan_output", "ggml_ssm_scan_state" ] }, { "page": "ggml_step", "title": "Step Function (Graph)", "topics": [ "ggml_step" ] }, { "page": "ggml_sub", "title": "Element-wise Subtraction (Graph)", "topics": [ "ggml_sub" ] }, { "page": "ggml_sub_inplace", "title": "Element-wise Subtraction In-place (Graph)", "topics": [ "ggml_sub_inplace" ] }, { "page": "ggml_sum", "title": "Sum (Graph)", "topics": [ "ggml_sum" ] }, { "page": "ggml_sum_rows", "title": "Sum Rows (Graph)", "topics": [ "ggml_sum_rows" ] }, { "page": "ggml_swiglu", "title": "SwiGLU (Swish/SiLU Gated Linear Unit) (Graph)", "topics": [ "ggml_swiglu" ] }, { "page": "ggml_swiglu_split", "title": "SwiGLU Split (Graph)", "topics": [ "ggml_swiglu_split" ] }, { "page": "ggml_tanh", "title": "Tanh Activation (Graph)", "topics": [ "ggml_tanh" ] }, { "page": "ggml_tanh_inplace", "title": "Tanh Activation In-place (Graph)", "topics": [ "ggml_tanh_inplace" ] }, { "page": "ggml_task", "title": "Construct a single-cell compute task", "topics": [ "ggml_task" ] }, { "page": "ggml_tcrossprod", "title": "GPU transposed cross product (drop-in for 'tcrossprod')", "topics": [ "ggml_tcrossprod" ] }, { "page": "ggml_tensor_copy", "title": "Copy Tensor Data", "topics": [ "ggml_tensor_copy" ] }, { "page": "ggml_tensor_data_ptr", "title": "Raw Data Pointer of a Tensor", "topics": [ "ggml_tensor_data_ptr" ] }, { "page": "ggml_tensor_nb", "title": "Get Tensor Strides (nb)", "topics": [ "ggml_tensor_nb" ] }, { "page": "ggml_tensor_num", "title": "Count Tensors in Context", "topics": [ "ggml_tensor_num" ] }, { "page": "ggml_tensor_overhead", "title": "Get Tensor Overhead", "topics": [ "ggml_tensor_overhead" ] }, { "page": "ggml_tensor_set_f32_scalar", "title": "Fill Tensor with Scalar", "topics": [ "ggml_tensor_set_f32_scalar" ] }, { "page": "ggml_tensor_shape", "title": "Get Tensor Shape", "topics": [ "ggml_tensor_shape" ] }, { "page": "ggml_tensor_type", "title": "Get Tensor Type", "topics": [ "ggml_tensor_type" ] }, { "page": "ggml_test", "title": "Test GGML", "topics": [ "ggml_test" ] }, { "page": "ggml_test_adamw_steps", "title": "Compare the AdamW Optimizer Step Across Backends", "topics": [ "ggml_test_adamw_steps" ] }, { "page": "ggml_test_backend_ops", "title": "Compare Backend Operations Against the CPU", "topics": [ "ggml_test_backend_ops" ] }, { "page": "ggml_time_init", "title": "Initialize GGML Timer", "topics": [ "ggml_time_init" ] }, { "page": "ggml_time_ms", "title": "Get Time in Milliseconds", "topics": [ "ggml_time_ms" ] }, { "page": "ggml_time_us", "title": "Get Time in Microseconds", "topics": [ "ggml_time_us" ] }, { "page": "ggml_timestep_embedding", "title": "Timestep Embedding (Graph Operation)", "topics": [ "ggml_timestep_embedding" ] }, { "page": "ggml_top_k", "title": "Top-K Indices (Graph)", "topics": [ "ggml_top_k" ] }, { "page": "ggml_tp_dp_forward", "title": "TPxDP hybrid matrix multiply across replicas of Vulkan device groups", "topics": [ "ggml_tp_dp_forward" ] }, { "page": "ggml_trainer", "title": "A Training Context You Step Yourself", "concept": [ "optimization" ], "topics": [ "ggml_trainer" ] }, { "page": "ggml_training_history", "title": "Training history of a fitted ggml model", "topics": [ "ggml_training_history" ] }, { "page": "ggml_transpose", "title": "Transpose (Graph)", "topics": [ "ggml_transpose" ] }, { "page": "GGML_TYPE_F32", "title": "GGML Data Types", "topics": [ "GGML_TYPE_BF16", "GGML_TYPE_F16", "GGML_TYPE_F32", "GGML_TYPE_I32", "GGML_TYPE_Q2_K", "GGML_TYPE_Q3_K", "GGML_TYPE_Q4_0", "GGML_TYPE_Q4_1", "GGML_TYPE_Q4_K", "GGML_TYPE_Q5_K", "GGML_TYPE_Q6_K", "GGML_TYPE_Q8_0" ] }, { "page": "ggml_type_name", "title": "Get Type Name", "concept": [ "type_system" ], "topics": [ "ggml_type_name" ] }, { "page": "ggml_type_size", "title": "Get Type Size in Bytes", "topics": [ "ggml_type_size" ] }, { "page": "ggml_type_sizef", "title": "Get Type Size as Float", "concept": [ "type_system" ], "topics": [ "ggml_type_sizef" ] }, { "page": "ggml_unary_op_name", "title": "Get Unary Operation Name", "concept": [ "op_info" ], "topics": [ "ggml_unary_op_name" ] }, { "page": "ggml_unfreeze_weights", "title": "Unfreeze Layer Weights", "topics": [ "ggml_unfreeze_weights" ] }, { "page": "ggml_unmarshal_model", "title": "Unmarshal a ggmlR model from an in-memory container", "topics": [ "ggml_unmarshal_model" ] }, { "page": "ggml_upscale", "title": "Upscale Tensor (Graph)", "topics": [ "GGML_SCALE_MODE_BICUBIC", "GGML_SCALE_MODE_BILINEAR", "GGML_SCALE_MODE_NEAREST", "ggml_upscale" ] }, { "page": "ggml_used_mem", "title": "Get Used Memory", "topics": [ "ggml_used_mem" ] }, { "page": "ggml_version", "title": "Get GGML version", "topics": [ "ggml_version" ] }, { "page": "ggml_view_1d", "title": "1D View with Byte Offset (Graph)", "topics": [ "ggml_view_1d" ] }, { "page": "ggml_view_2d", "title": "2D View with Byte Offset (Graph)", "topics": [ "ggml_view_2d" ] }, { "page": "ggml_view_3d", "title": "3D View with Byte Offset (Graph)", "topics": [ "ggml_view_3d" ] }, { "page": "ggml_view_4d", "title": "4D View with Byte Offset (Graph)", "topics": [ "ggml_view_4d" ] }, { "page": "ggml_view_tensor", "title": "View Tensor", "topics": [ "ggml_view_tensor" ] }, { "page": "ggml_vulkan_available", "title": "Check if Vulkan support is available", "topics": [ "ggml_vulkan_available" ] }, { "page": "ggml_vulkan_backend_name", "title": "Get Vulkan backend name", "topics": [ "ggml_vulkan_backend_name" ] }, { "page": "ggml_vulkan_device_caps", "title": "Get Vulkan device capabilities", "topics": [ "ggml_vulkan_device_caps" ] }, { "page": "ggml_vulkan_device_count", "title": "Get number of Vulkan devices", "topics": [ "ggml_vulkan_device_count" ] }, { "page": "ggml_vulkan_device_description", "title": "Get Vulkan device description", "topics": [ "ggml_vulkan_device_description" ] }, { "page": "ggml_vulkan_device_groups", "title": "Probe Vulkan device groups (NVLink / multi-GPU peer access)", "topics": [ "ggml_vulkan_device_groups" ] }, { "page": "ggml_vulkan_device_memory", "title": "Get Vulkan device memory", "topics": [ "ggml_vulkan_device_memory" ] }, { "page": "ggml_vulkan_free", "title": "Free Vulkan backend", "topics": [ "ggml_vulkan_free" ] }, { "page": "ggml_vulkan_hard_exit_available", "title": "Is the Vulkan hard-exit path compiled in?", "topics": [ "ggml_vulkan_hard_exit_available" ] }, { "page": "ggml_vulkan_init", "title": "Initialize Vulkan backend", "topics": [ "ggml_vulkan_init" ] }, { "page": "ggml_vulkan_is_backend", "title": "Check if backend is Vulkan", "topics": [ "ggml_vulkan_is_backend" ] }, { "page": "ggml_vulkan_list_devices", "title": "List all Vulkan devices", "topics": [ "ggml_vulkan_list_devices" ] }, { "page": "ggml_vulkan_p2p_selftest", "title": "Opaque-fd device-to-device P2P self-test", "topics": [ "ggml_vulkan_p2p_selftest" ] }, { "page": "ggml_vulkan_shutdown", "title": "Explicitly tear down the Vulkan instance and devices", "topics": [ "ggml_vulkan_shutdown" ] }, { "page": "ggml_vulkan_split_buffer_type", "title": "Create a Vulkan tensor-split buffer type", "topics": [ "ggml_vulkan_split_buffer_type" ] }, { "page": "ggml_vulkan_split_mul_mat", "title": "Tensor-parallel matrix multiply across Vulkan devices", "topics": [ "ggml_vulkan_split_mul_mat" ] }, { "page": "ggml_vulkan_split_row_ranges", "title": "Compute tensor-parallel row-split ranges", "topics": [ "ggml_vulkan_split_row_ranges" ] }, { "page": "ggml_vulkan_status", "title": "Print Vulkan status", "topics": [ "ggml_vulkan_status" ] }, { "page": "ggml_win_part", "title": "Window Partition (Graph)", "topics": [ "ggml_win_part" ] }, { "page": "ggml_win_unpart", "title": "Window Un-partition (Graph)", "topics": [ "ggml_win_unpart" ] }, { "page": "ggml_with_temp_ctx", "title": "Execute with Temporary Context", "topics": [ "ggml_with_temp_ctx" ] }, { "page": "gguf_free", "title": "Free GGUF Resources", "topics": [ "gguf_free" ] }, { "page": "gguf_load", "title": "Load a GGUF File", "topics": [ "gguf_load" ] }, { "page": "gguf_metadata", "title": "Get GGUF Metadata", "topics": [ "gguf_metadata" ] }, { "page": "gguf_tensor_data", "title": "Extract Tensor Data", "topics": [ "gguf_tensor_data" ] }, { "page": "gguf_tensor_info", "title": "Get Tensor Info", "topics": [ "gguf_tensor_info" ] }, { "page": "gguf_tensor_names", "title": "List Tensor Names in a GGUF File", "topics": [ "gguf_tensor_names" ] }, { "page": "glance.ggmlr_parsnip_model", "title": "One-row summary of a fitted ggml parsnip model", "topics": [ "glance.ggmlr_parsnip_model" ] }, { "page": "iq2xs_free_impl", "title": "Free IQ2 Quantization Tables", "concept": [ "quantization" ], "topics": [ "iq2xs_free_impl" ] }, { "page": "iq2xs_init_impl", "title": "Initialize IQ2 Quantization Tables", "concept": [ "quantization" ], "topics": [ "iq2xs_init_impl" ] }, { "page": "iq3xs_free_impl", "title": "Free IQ3 Quantization Tables", "concept": [ "quantization" ], "topics": [ "iq3xs_free_impl" ] }, { "page": "iq3xs_init_impl", "title": "Initialize IQ3 Quantization Tables", "concept": [ "quantization" ], "topics": [ "iq3xs_init_impl" ] }, { "page": "lr_scheduler_cosine", "title": "Cosine-annealing learning rate scheduler", "topics": [ "lr_scheduler_cosine" ] }, { "page": "lr_scheduler_cyclic", "title": "Cyclical learning rate scheduler", "topics": [ "lr_scheduler_cyclic" ] }, { "page": "lr_scheduler_onecycle", "title": "One-cycle learning rate scheduler", "topics": [ "lr_scheduler_onecycle" ] }, { "page": "lr_scheduler_step", "title": "Step-decay learning rate scheduler", "topics": [ "lr_scheduler_step" ] }, { "page": "lr_scheduler_warmup_cosine", "title": "Linear warmup followed by cosine annealing", "topics": [ "lr_scheduler_warmup_cosine" ] }, { "page": "nn_topo_sort", "title": "Topologically sort nodes reachable from output nodes", "topics": [ "nn_topo_sort" ] }, { "page": "onnx_device_info", "title": "ONNX model device/scheduler diagnostics", "topics": [ "onnx_device_info" ] }, { "page": "onnx_graph_info", "title": "ONNX graph structure", "topics": [ "onnx_graph_info" ] }, { "page": "onnx_inputs", "title": "List ONNX model inputs", "topics": [ "onnx_inputs" ] }, { "page": "onnx_load", "title": "Load an ONNX model", "topics": [ "onnx_load" ] }, { "page": "onnx_run", "title": "Run ONNX model inference", "topics": [ "onnx_run" ] }, { "page": "onnx_summary", "title": "ONNX model summary", "topics": [ "onnx_summary" ] }, { "page": "optimizer_adam", "title": "Create an Adam optimizer", "topics": [ "optimizer_adam" ] }, { "page": "optimizer_sgd", "title": "Create an SGD optimizer", "topics": [ "optimizer_sgd" ] }, { "page": "plot.ggml_history", "title": "Plot training history", "topics": [ "plot.ggml_history" ] }, { "page": "predict.ggml_sequential_model", "title": "Predict with a Trained Model", "topics": [ "predict.ggml_functional_model", "predict.ggml_sequential_model" ] }, { "page": "predict.onnx_model", "title": "Predict with an ONNX Model", "topics": [ "predict.onnx_model" ] }, { "page": "print.ag_tensor", "title": "Print method for ag_tensor", "topics": [ "print.ag_tensor" ] }, { "page": "print.ggml_functional_model", "title": "Print method for ggml_functional_model", "topics": [ "print.ggml_functional_model" ] }, { "page": "print.ggml_history", "title": "Print method for ggml_history", "topics": [ "print.ggml_history" ] }, { "page": "print.ggml_sequential_model", "title": "Print method for ggml_sequential_model", "topics": [ "print.ggml_sequential_model" ] }, { "page": "print.onnx_model", "title": "Print ONNX model summary", "topics": [ "print.onnx_model" ] }, { "page": "pt_cube4_layout", "title": "The piece layout of a 4x4x4 cube", "topics": [ "pt_cube4_layout" ] }, { "page": "pt_forward", "title": "Run a piece transformer", "topics": [ "pt_forward" ] }, { "page": "pt_transformer", "title": "Build a piece transformer", "topics": [ "pt_transformer" ] }, { "page": "pt_transformer_load", "title": "Build a piece transformer and load a checkpoint into it", "topics": [ "pt_transformer_load" ] }, { "page": "pth_catalogue", "title": "List the tensors in a PyTorch checkpoint", "topics": [ "pth_catalogue" ] }, { "page": "pth_load", "title": "Read the tensors of a PyTorch checkpoint", "topics": [ "pth_load" ] }, { "page": "quantize_iq2_xxs", "title": "Quantize Data (IQ)", "concept": [ "quantization" ], "topics": [ "quantize_iq1_m", "quantize_iq1_s", "quantize_iq2_s", "quantize_iq2_xs", "quantize_iq2_xxs", "quantize_iq3_s", "quantize_iq3_xxs", "quantize_iq4_nl", "quantize_iq4_xs" ] }, { "page": "quantize_mxfp4", "title": "Quantize Data (MXFP4)", "concept": [ "quantization" ], "topics": [ "quantize_mxfp4" ] }, { "page": "quantize_nvfp4", "title": "Quantize Data (NVFP4)", "concept": [ "quantization" ], "topics": [ "quantize_nvfp4" ] }, { "page": "quantize_q1_0", "title": "Quantize Data (Q1_0)", "concept": [ "quantization" ], "topics": [ "quantize_q1_0" ] }, { "page": "quantize_q2_K", "title": "Quantize Data (K-quants)", "concept": [ "quantization" ], "topics": [ "quantize_q2_K", "quantize_q3_K", "quantize_q4_K", "quantize_q5_K", "quantize_q6_K" ] }, { "page": "quantize_q4_0", "title": "Quantize Data (Q4_0)", "concept": [ "quantization" ], "topics": [ "quantize_q4_0", "quantize_q4_1", "quantize_q5_0", "quantize_q5_1", "quantize_q8_0" ] }, { "page": "quantize_row_iq3_xxs_ref", "title": "Quantize Row Reference (IQ)", "concept": [ "quantization" ], "topics": [ "quantize_row_iq2_s_ref", "quantize_row_iq3_s_ref", "quantize_row_iq3_xxs_ref", "quantize_row_iq4_nl_ref", "quantize_row_iq4_xs_ref" ] }, { "page": "quantize_row_mxfp4_ref", "title": "Quantize Row Reference (MXFP4)", "concept": [ "quantization" ], "topics": [ "quantize_row_mxfp4_ref" ] }, { "page": "quantize_row_nvfp4_ref", "title": "Quantize Row to NVFP4 (Reference)", "concept": [ "quantization" ], "topics": [ "quantize_row_nvfp4_ref" ] }, { "page": "quantize_row_q1_0_ref", "title": "Quantize Row to Q1_0 (Reference)", "concept": [ "quantization" ], "topics": [ "quantize_row_q1_0_ref" ] }, { "page": "quantize_row_q2_K_ref", "title": "Quantize Row Reference (K-quants)", "concept": [ "quantization" ], "topics": [ "quantize_row_q2_K_ref", "quantize_row_q3_K_ref", "quantize_row_q4_K_ref", "quantize_row_q5_K_ref", "quantize_row_q6_K_ref", "quantize_row_q8_K_ref" ] }, { "page": "quantize_row_q4_0_ref", "title": "Quantize Row Reference (Basic)", "concept": [ "quantization" ], "topics": [ "quantize_row_q4_0_ref", "quantize_row_q4_1_ref", "quantize_row_q5_0_ref", "quantize_row_q5_1_ref", "quantize_row_q8_0_ref", "quantize_row_q8_1_ref" ] }, { "page": "quantize_row_tq1_0_ref", "title": "Quantize Row Reference (Ternary)", "concept": [ "quantization" ], "topics": [ "quantize_row_tq1_0_ref", "quantize_row_tq2_0_ref" ] }, { "page": "quantize_tq1_0", "title": "Quantize Data (Ternary)", "concept": [ "quantization" ], "topics": [ "quantize_tq1_0", "quantize_tq2_0" ] }, { "page": "rope_types", "title": "RoPE Mode Constants", "topics": [ "GGML_ROPE_TYPE_MROPE", "GGML_ROPE_TYPE_NEOX", "GGML_ROPE_TYPE_NORM", "GGML_ROPE_TYPE_VISION", "rope_types" ] }, { "page": "RunGGML", "title": "Run a GGML GPU operation on a Seurat object", "topics": [ "RunGGML", "RunGGML.default", "RunGGML.Seurat", "RunGGML.SingleCellExperiment" ] }, { "page": "summary.ggml_sequential_model", "title": "Summary method for ggml_sequential_model", "topics": [ "summary.ggml_sequential_model" ] }, { "page": "tidy.ggmlr_parsnip_model", "title": "Tidy a fitted ggml parsnip model into a per-layer table", "topics": [ "tidy.ggmlr_parsnip_model" ] }, { "page": "with_grad_tape", "title": "Run code with gradient tape enabled", "topics": [ "with_grad_tape" ] } ], "_readme": "https://github.com/zabis13/ggmlr/raw/HEAD/README.md", "_rundeps": [ "generics", "R6" ], "_sysdeps": [ { "shlib": "libstdc++", "package": "libstdc++6", "source": "gcc", "version": "16-20260322-1ubuntu1", "name": "c++", "homepage": "http://gcc.gnu.org/", "description": "GNU Standard C++ Library v3" }, { "shlib": "libgomp", "package": "libgomp1", "source": "gcc", "version": "16-20260322-1ubuntu1", "name": "openmp", "homepage": "http://gcc.gnu.org/", "description": "GCC OpenMP (GOMP) support library" } ], "_vignettes": [ { "source": "quickstart.Rmd", "filename": "quickstart.html", "title": "Quickstart: from data to prediction in ~10 lines", "engine": "knitr::rmarkdown", "headings": [ "Next steps" ], "created": "2026-06-08 21:38:10", "modified": "2026-06-08 21:38:10", "commits": 1 }, { "source": "keras-like-api.Rmd", "filename": "keras-like-api.html", "title": "Keras-like API in ggmlR", "engine": "knitr::rmarkdown", "headings": [ "Installation", "1. Sequential API", "1.1 Prepare data — iris (3-class classification)", "1.2 Build and compile", "1.3 Train", "1.4 Evaluate and predict", "1.5 Save and reload", "2. Functional API", "2.1 Single-input DAG — residual-style network on iris", "2.2 Multi-input model — mtcars regression", "2.3 Multi-output model — one trunk, two heads", "2.4 Attention and a transformer block", "3. Callbacks", "3.1 Early stopping", "3.2 Learning-rate schedulers", "4. Autograd Engine (ag_*)", "4.1 Prepare data (iris, col-major)", "4.2 Build a model with ag_sequential", "4.3 Training loop", "4.4 Inference and accuracy", "4.5 LR scheduler + gradient clipping", "4.6 Manual model from raw ag_param", "5. GPU / device selection", "API summary" ], "created": "2026-04-13 21:53:45", "modified": "2026-09-19 05:39:28", "commits": 4 }, { "source": "autograd-engine.Rmd", "filename": "autograd-engine.html", "title": "Autograd Engine", "engine": "knitr::rmarkdown", "headings": [ "1. Core primitives", "ag_tensor and ag_param", "Forward pass and gradient tape", "2. Built-in operations", "3. ag_sequential — module API", "4. Optimizers", "Training loop", "5. LR schedulers", "6. Gradient clipping", "7. Dataloader", "8. Eval mode and inference", "9. Raw ag_param — full manual control", "10. Mixed precision", "11. GPU residency", "Two rules this imposes", "Switching it off", "Device buffers", "12. Data-parallel training" ], "created": "2026-04-13 21:53:45", "modified": "2026-10-09 21:03:10", "commits": 6 }, { "source": "tidymodels-integration.Rmd", "filename": "tidymodels-integration.html", "title": "tidymodels / parsnip Integration", "engine": "knitr::rmarkdown", "headings": [ "1. Classification", "2. Regression", "3. Engine parameters", "4. Resampling with rsample", "5. Recipes for preprocessing", "6. Hyperparameter tuning", "7. Comparison with other engines", "8. Extracting the fitted engine and fit time", "9. Limitations of the engine object", "Summary" ], "created": "2026-04-13 21:53:45", "modified": "2026-06-08 21:38:10", "commits": 2 }, { "source": "mlr3-integration.Rmd", "filename": "mlr3-integration.html", "title": "mlr3 Integration", "engine": "knitr::rmarkdown", "headings": [ "Quick start", "Classification", "Regression", "Learner parameters", "GPU acceleration", "Custom model architecture", "Resampling and benchmarking", "Hyperparameter tuning", "Callbacks", "Observation weights", "Marshal and parallel execution", "Summary" ], "created": "2026-04-13 21:53:45", "modified": "2026-06-08 21:38:10", "commits": 3 }, { "source": "gpu-vulkan.Rmd", "filename": "gpu-vulkan.html", "title": "GPU / Vulkan Backend", "engine": "knitr::rmarkdown", "headings": [ "1. Check availability", "2. Device enumeration", "3. Autograd device selection", "4. Mixed precision (f16 / bf16)", "5. Vulkan memory info", "6. Multi-GPU", "7. Sequential / functional API on GPU", "8. ONNX inference on GPU", "9. Build configuration", "Restoring the default device" ], "created": "2026-04-13 21:53:45", "modified": "2026-09-19 05:39:28", "commits": 6 }, { "source": "quantization.Rmd", "filename": "quantization.html", "title": "Quantization", "engine": "knitr::rmarkdown", "headings": [ "1. Quantization formats", "2. Quantize and dequantize", "3. K-quants (better quality)", "4. IQ quants — importance matrix", "5. Comparing formats", "6. Reference (row-level) functions" ], "created": "2026-02-06 08:50:59", "modified": "2026-06-08 21:38:10", "commits": 4 }, { "source": "data-parallel-training.Rmd", "filename": "data-parallel-training.html", "title": "Data-Parallel Training", "engine": "knitr::rmarkdown", "headings": [ "1. Concept", "2. Minimal example", "3. Multi-GPU", "4. Gradient clipping", "5. ag_dataloader — batched training loop", "6. Full example" ], "created": "2026-04-13 21:53:45", "modified": "2026-09-19 05:39:28", "commits": 3 }, { "source": "onnx-import.Rmd", "filename": "onnx-import.html", "title": "ONNX Model Import", "engine": "knitr::rmarkdown", "headings": [ "1. Load and inspect a model", "2. Run inference", "3. GPU inference", "4. Dynamic input shapes", "5. FP16 inference", "6. Supported operators", "7. Examples", "8. Debugging tips" ], "created": "2026-04-13 21:53:45", "modified": "2026-06-08 21:38:10", "commits": 2 }, { "source": "embedding-ggmlR.Rmd", "filename": "embedding-ggmlR.html", "title": "Using ggmlR as a Backend in Your Package", "engine": "knitr::rmarkdown", "headings": [ "Part A — Using the ggmlR R API from your package", "A1. Depend on ggmlR", "A2. Everything the examples use is exported", "A3. tidymodels / mlr3 engines", "Part B — Linking ggml from your package's C/C++ code", "1. What ggmlR exports", "2. DESCRIPTION", "3. src/Makevars", "4. configure — detect ggmlR at build time", "5. C code — minimal example", "6. R wrapper", "7. Thread count", "8. Vulkan in downstream packages", "9. Real-world references" ], "created": "2026-04-13 21:53:45", "modified": "2026-06-08 21:38:10", "commits": 2 }, { "source": "single-cell-seurat.Rmd", "filename": "single-cell-seurat.html", "title": "Single-cell GPU Acceleration with Seurat", "engine": "knitr::rmarkdown", "headings": [ "The one call: RunGGML()", "A worked example", "Normalize and scale on the GPU", "PCA (embed)", "UMAP (umap)", "Neighbour graphs (neighbors)", "Provenance", "The layers underneath", "Bioconductor: SingleCellExperiment", "A real dataset, end to end", "Speed-up", "Agreement", "What is and isn't accelerated" ], "created": "2026-06-21 04:28:49", "modified": "2026-07-12 13:14:31", "commits": 2 }, { "source": "multi-gpu.Rmd", "filename": "multi-gpu.html", "title": "Multi-GPU: Tensor, Pipeline and Hybrid Parallelism", "engine": "knitr::rmarkdown", "headings": [ "1. Which mode do you need?", "Are you sure you want this layer?", "2. The transport, and why it constrains everything", "3. Tensor parallelism", "4. Pipeline parallelism", "5. Hybrid: DP over TP or PP", "6. Clean shutdown (and the --enable-hard-exit flag)", "7. Where to look next", "8. Summary" ], "created": "2026-02-06 08:50:59", "modified": "2026-07-12 13:14:31", "commits": 3 } ], "_score": 9.541279627182782, "_indexed": true, "_nocasepkg": "ggmlr", "_universes": [ "zabis13" ], "_previous": "0.8.7", "_binaries": [ { "r": "4.7.0", "os": "linux", "version": "0.8.8", "date": "2026-10-18T11:38:34.000Z", "distro": "resolute", "arch": "aarch64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/6834db6f25a29741d5b8993c68c6b8ec3d139169c411a0480240e6a8b8377ca6", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.7.0", "os": "linux", "version": "0.8.8", "date": "2026-10-18T11:39:08.000Z", "distro": "resolute", "arch": "x86_64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/363554af6b329a1f78cb0d76982d83002b89ba3e530edcff44a156c7123651aa", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.6.1", "os": "linux", "version": "0.8.8", "date": "2026-10-18T11:38:37.000Z", "distro": "resolute", "arch": "aarch64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/5e8824e054c390f0ab0c4a80c1ac4dab4f9a045d9f5ca80e1eadd8665a38ef71", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.6.1", "os": "linux", "version": "0.8.8", "date": "2026-10-18T11:38:46.000Z", "distro": "resolute", "arch": "x86_64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/efa2c29b68ad32728e813b399f8c7eca5fd4c33c023961aaaa81d4455e4df36f", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.5.3", "os": "mac", "version": "0.8.8", "date": "2026-10-18T11:45:18.000Z", "arch": "aarch64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/046e7cab4f8e85e0b61ab026d5cee568dfad1191049e045171c55668a5a247be", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.5.3", "os": "mac", "version": "0.8.8", "date": "2026-10-18T11:52:26.000Z", "arch": "x86_64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/fbbb36db89c5ed71e8dbd2b92a77a5c0e6db6a9e55364f56cdbb90c618646e00", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.6.1", "os": "mac", "version": "0.8.8", "date": "2026-10-18T11:54:27.000Z", "arch": "aarch64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/e72d2ce2a3e7140836b8d0cb4038b1844e71b00b59aaecfddae1b9905be5d470", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.6.1", "os": "mac", "version": "0.8.8", "date": "2026-10-18T12:00:44.000Z", "arch": "x86_64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/d24c6cc97a9e01cefdd68161d1551918a04cd604f34e26a24bf4b48b6774a9d1", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.6.0", "os": "wasm", "version": "0.8.8", "date": "2026-10-18T11:38:59.000Z", "arch": "emscripten", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/2e8e888dff27c64abddfe40b029d8a7e6ce83e3dc51b5902e4c75aaa789c8647", "status": "success", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.7.0", "os": "win", "version": "0.8.8", "date": "2026-10-18T11:39:37.000Z", "arch": "aarch64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/2e2bfccf2dfdf9db94fc5065524f19d472894bde8effab103906edb49f88957c", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.7.0", "os": "win", "version": "0.8.8", "date": "2026-10-18T11:38:15.000Z", "arch": "x86_64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/cb2a83cfe22b1f7bd535a59870ae0968b530d070750d4d99112b6c13f02aac21", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.5.3", "os": "win", "version": "0.8.8", "date": "2026-10-18T11:37:54.000Z", "arch": "x86_64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/a6e68bf2a52ba645fdae8ed15648445696d6c91111e66d0458814e985418a4b4", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.6.1", "os": "win", "version": "0.8.8", "date": "2026-10-18T11:39:40.000Z", "arch": "aarch64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/565ec5625a97656c6f2e051f4425bea3cdb9e5b8f3a10498b27f3c13b743c840", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" }, { "r": "4.6.1", "os": "win", "version": "0.8.8", "date": "2026-10-18T11:37:52.000Z", "arch": "x86_64", "commit": "61c835098454641a78c8bec83a07ff06fbdc8a79", "fileid": "https://r2.ropensci.org/5c3c03a2310261b8201a2e01a019a265cc68da61b5753657cd13cb66057c1530", "status": "success", "check": "NOTE", "buildurl": "https://github.com/r-universe/zabis13/actions/runs/39123392047" } ] }