Repository navigation
Merge latest master branch - #2
Merged
Merged
Conversation
* llama : the WPM vocabs use the CLS token as BOS ggml-ci * llama : add comment
* llama_server_response_fields * llama_server_response_fields_fix_issues * params fixes * fix * clarify docs * change to "response_fields" --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
* more perfo with llamafile tinyblas on x86_64. - add bf16 suport - change dispache strategie (thanks: ikawrakow/ik_llama.cpp#71 ) - reduce memory bandwidth simple tinyblas dispache and more cache freindly * tinyblas dynamic dispaching * sgemm: add M blocs. * - git 2.47 use short id of len 9. - show-progress is not part of GNU Wget2 * remove not stable test
…ngs endpoints (ggml-org#10967) * add support for base64 * fix base64 test * improve test --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
Warning types fixed (observed under MSYS2 GCC 14.2.0): * format '%ld' expects argument of type 'long int', but argument has type 'size_t' * llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp:81:46: warning: missing initializer for member '_STARTUPINFOA::lpDesktop' [-Wmissing-field-initializers] (emitted for all struct field except first)
* multi row k quant shaders! * better row selection * more row choices * readjust row selection * rm_kq=2 by default
…g#10942) * tests: Add im2col perf tests * vulkan: optimize im2col, more elements per thread * vulkan: increase small tile size for NV_coopmat2 * vulkan: change im2col to 512 elements per workgroup
Make the mul_mat_vec shaders support N>1 (as a spec constant, NUM_COLS) where the batch_strides are overloaded to hold the row strides. Put the loads from the B matrix in the innermost loop because it should cache better. Share some code for reducing the result values to memory in mul_mat_vec_base.
…hen building with LLAMA_CURL=ON and GGML_OPENCL=ON (ggml-org#11013) In common/common.cpp: * Convert usage of stat() function call to check if file exists to standard library function std::filesystem::exists (error unable to match to correct function signature) * Additional conditions to check if PATH_MAX is already defined in WIN32 environment (warning it is already defined in MSYS2) In examples/run/run.cpp: * Add io.h header inclusion (error cannot find function _get_osfhandle) * Change initialisers for OVERLAPPED to empty struct (warning about uninitialised members) * Add initialiser for hFile (warning it may be uninitialised) * Add cast for curl_off_t percentage value to long int in generate_progress_prefix function (warning that curl_off_t is long long int) In ggml/src/ggml-opencl/ggml-opencl.cpp: * Initialise certain declared cl_mem variables to nullptr for greater safety (warning about B_d variable possibly used unassigned)
* conflict resolution * move comments after bracket to its own line * DeciLMCausalModel now reads rope_theta from config.json properly
* server : add OAI compat for /v1/completions * add test * add docs * better docs
* server : clean up built-in template detection * fix compilation * add chat template test * fix condition
…g#11027) * Fixes for clang AVX VNNI * enable AVX VNNI and alder lake build for MSVC * Apply suggestions from code review --------- Co-authored-by: slaren <slarengh@gmail.com>
* list llama-swap under tools in README * readme: add llama-swap to Infrastructure
* slot.can_batch_with * lora per request * test: force disable cache prompt * move can_batch_with check * fix condition * add slow test with llama 8b * update docs * move lora change task to queue * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * lora_base * remove redundant check --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server/bench: - support openAI streaming standard output with [DONE]\n\n - export k6 raw results in csv - fix too many tcp idle connection in tcp_wait - add metric time to emit first token * server/bench: - fix when prometheus not started - wait for server to be ready before starting bench
* llama : scatter llama.cpp into multiple modules (wip) * llama : control-vector -> adapter * llama : arch * llama : mmap ggml-ci * ci : remove BUILD_SHARED_LIBS=OFF ggml-ci * llama : arch (cont) ggml-ci * llama : chat ggml-ci * llama : model ggml-ci * llama : hparams ggml-ci * llama : adapter ggml-ci * examples : fix ggml-ci * rebase ggml-ci * minor * llama : kv cache ggml-ci * llama : impl ggml-ci * llama : batch ggml-ci * cont ggml-ci * llama : context ggml-ci * minor * llama : context (cont) ggml-ci * llama : model loader ggml-ci * common : update lora ggml-ci * llama : quant ggml-ci * llama : quant (cont) ggml-ci * minor [no ci]
…ls (ggml-org#11053) * Disable KV cache shifting automatically for unsupported models instead of exiting directly Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * Update common/common.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Signed-off-by: Molly Sophia <mollysophia379@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit attempts to improve the log message for the inputs of the splits in the sched_print_assignments function. The motivation for this change is that currently even if there are no inputs a colon is displayed at the end of the line, which can make it a little confusing when reading the output as it could be interpreted as the line below are inputs when they are in fact nodes. With this change the colon will only be printed if there actually are inputs.
…ggml-org#11047) * Added init tensor calling code * Added get_alloc_size forwarding * Cleaned up and improved type/error handling. * fix: remove trailing whitespaces. * Cleanup and use GGML error logging functions. * Handle potentially dangerous edge cases. * Apply suggestions from code review Co-authored-by: Diego Devesa <slarengh@gmail.com> --------- Co-authored-by: Diego Devesa <slarengh@gmail.com>
* convert : extend DEEPSEEK2 model architecture to support DeepseekV3ForCausalLM by adding EXPERT_WEIGHTS_NORM and EXPERT_GATING_FUNC model parameters and FFN_EXP_PROBS_B tensor type * vocab : add DeepSeek V3 pre-tokenizer regexes * unicode : handle ACCENT_MARK and SYMBOL categories in regex * llama : add DeepSeek V3 chat template, handle new model parameters and tensor types --------- Co-authored-by: Stanisław Szymczyk <sszymczy@gmail.com>
* GGUF: C++ refactor, backend support, misc fixes remove ggml_tensor.backend update CODEOWNERS [no ci] remove gguf_get_data from API revise GGUF API data types
* fix: Vulkan shader gen binary path when cross compiling
…gml-org#11117) * Disable GL_KHR_cooperative_matrix Vulkan extension if not available. * Perform Vulkan extensions checks in a more sensible order * Remove unnecessary #ifdef directive
This change upstreams llamafile's cpu matrix multiplication kernels for ppc64le using MMA builtins for quantised int8 datatype. This change results in 10% - 70% improvement in total speed(ie all tokens/total time), across various batch sizes. The patch is tested with Meta-Lllama-3-8B, Mistral-7B, Llama-2-7B-chat-hf models on a IBM POWER10 machine. Signed-off-by: Amrita H S <amritahs@linux.vnet.ibm.com>
* ci : pin dependency to specific version * will this fix ec?
2 of 4 tasks
* (wip) support mergekit-extracted lora * support mergekit-extract-lora * use lora->get_scale * correct comment * correct norm name & condition * add some hints
The main motivation for this change is it was not handing ctrl-c/ctrl-d correctly. Modify `read_user_input` to handle EOF, "/bye" command, and empty input cases. Introduce `get_user_input` function to manage user input loop and handle different return cases. Signed-off-by: Eric Curtin <ecurtin@redhat.com>
* Moved scripts dir and fixed pyproject.toml * updated readme * fixed README urls * bump pypi gguf to v0.14.0 * retrigger ci * empty commit - trigger ci
Signed-off-by: hydai <z54981220@gmail.com>
* model: support phimoe * python linter * doc: minor Co-authored-by: ThiloteE <73715071+ThiloteE@users.noreply.github.com> * doc: minor Co-authored-by: ThiloteE <73715071+ThiloteE@users.noreply.github.com> * doc: add phimoe as supported model ggml-ci --------- Co-authored-by: ThiloteE <73715071+ThiloteE@users.noreply.github.com>
* server : add tooltips to settings and themes btn This commit adds tooltips to the settings and themes buttons in the webui. The tooltip will be displayed below the actual buttons when hovered over. The motivation for this change is to clarify the purpose of the themes button. * squash! server : add tooltips to settings and themes btn This commit adds a tooltip to the '...' button when a chat has been started. The tooltip is "Chat options" which think could be a good description as the dropdown contains options to delete or download the current chat. * rm tooltip for 3 dots button --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
Since NVIDIA does not release CUDA for in-maintenance versions of Fedora, the process of setting up the CUDA toolkit on Fedora has become quite involved. This guide should help mere mortals install CUDA for development in a Fedora 39 toolbox environment, without affecting the host system.
sparkleholic
pushed a commit
that referenced
this pull request
Oct 1, 2025
…gml-org#16038) Initalizing RESERVED_NAME in is_reserved_name() is not thread safe and leads to corrupted memory when used from multiple threads as can be seen in the asan trace below. This fixes the initialization to make it thread-safe. #0 0x000100abd018 in std::__1::pair<std::__1::__hash_iterator<std::__1::__hash_node<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, void*>*>, bool> std::__1::__hash_table<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, std::__1::hash<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>, std::__1::equal_to<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>, std::__1::allocator<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>>::__emplace_unique_key_args<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&>(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&) __hash_table:1565 #1 0x000100ab0320 in SchemaConverter::visit(nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&) json-schema-to-grammar.cpp:802 #2 0x000100aafc48 in std::__1::__function::__func<build_grammar(std::__1::function<void (common_grammar_builder const&)> const&, common_grammar_options const&)::$_2, std::__1::allocator<build_grammar(std::__1::function<void (common_grammar_builder const&)> const&, common_grammar_options const&)::$_2>, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> (std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&, nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&)>::operator()(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&, nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&) function.h:319 #3 0x000100a2c938 in std::__1::__function::__func<common_chat_params_init_llama_3_x(minja::chat_template const&, templates_params const&, bool)::$_0::operator()(common_grammar_builder const&) const::'lambda'(nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&), std::__1::allocator<common_chat_params_init_llama_3_x(minja::chat_template const&, templates_params const&, bool)::$_0::operator()(common_grammar_builder const&) const::'lambda'(nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&)>, void (nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&)>::operator()(nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&) function.h:319 ggml-org#4 0x000100a139f8 in foreach_function(nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&, std::__1::function<void (nlohmann::json_abi_v3_12_0::basic_json<nlohmann::json_abi_v3_12_0::ordered_map, std::__1::vector, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, bool, long long, unsigned long long, double, std::__1::allocator, nlohmann::json_abi_v3_12_0::adl_serializer, std::__1::vector<unsigned char, std::__1::allocator<unsigned char>>, void> const&)> const&) chat.cpp:762 ggml-org#5 0x000100a2a7f4 in std::__1::__function::__func<common_chat_params_init_llama_3_x(minja::chat_template const&, templates_params const&, bool)::$_0, std::__1::allocator<common_chat_params_init_llama_3_x(minja::chat_template const&, templates_params const&, bool)::$_0>, void (common_grammar_builder const&)>::operator()(common_grammar_builder const&) function.h:319 ggml-org#6 0x000100aa98f4 in build_grammar(std::__1::function<void (common_grammar_builder const&)> const&, common_grammar_options const&) json-schema-to-grammar.cpp:982 ggml-org#7 0x0001009c9314 in common_chat_params_init_llama_3_x(minja::chat_template const&, templates_params const&, bool) chat.cpp:1110 ggml-org#8 0x0001009b8afc in common_chat_templates_apply_jinja(common_chat_templates const*, common_chat_templates_inputs const&) chat.cpp:1992 ggml-org#9 0x0001009b533c in common_chat_templates_apply(common_chat_templates const*, common_chat_templates_inputs const&) chat.cpp:2074 ggml-org#10 0x000100810120 in llamacpp_apply_chat_template+0x724 (predict_oai-98384e17fb94e863:arm64+0x100090120) ... ==45482==Register values: x[0] = 0x00006020004147f8 x[1] = 0x00006080000013c8 x[2] = 0x0000000000000000 x[3] = 0x0000604006289738 x[4] = 0x0000000000000002 x[5] = 0x0000000000000001 x[6] = 0x04034000004b4000 x[7] = 0x0000000000000001 x[8] = 0xbebebebebebebebe x[9] = 0x17d7d7d7d7d7d7d7 x[10] = 0x00000c04000828ff x[11] = 0x0000000000000001 x[12] = 0x000000002018d383 x[13] = 0x0000000000000000 x[14] = 0xfa0000000000fafa x[15] = 0x000010700001ffff x[16] = 0x000000019dc012c0 x[17] = 0x00000001021284f8 x[18] = 0x0000000000000000 x[19] = 0x00000001700acdc0 x[20] = 0x0000000000000002 x[21] = 0x000000002018d384 x[22] = 0x16dd16fd2e731151 x[23] = 0x0000007000020000 x[24] = 0x0000000100c69c08 x[25] = 0x0000000100c69c20 x[26] = 0x00006080000013c7 x[27] = 0x0000000100c69c00 x[28] = 0x00000001700acd60 fp = 0x00000001700aceb0 lr = 0x0000000100abce30 sp = 0x00000001700acd60 AddressSanitizer can not provide additional info. SUMMARY: AddressSanitizer: SEGV __hash_table:1565 in std::__1::pair<std::__1::__hash_iterator<std::__1::__hash_node<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, void*>*>, bool> std::__1::__hash_table<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, std::__1::hash<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>, std::__1::equal_to<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>, std::__1::allocator<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>>>::__emplace_unique_key_args<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>>, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&>(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char>> const&) Thread T5 created by T0 here: #0 0x0001020b99d4 in pthread_create+0x5c (libclang_rt.asan_osx_dynamic.dylib:arm64e+0x359d4) #1 0x000100873910 in std::sys::pal::unix::thread::Thread::new::h77254fdd87a28e05+0x118 (predict_oai-98384e17fb94e863:arm64+0x1000f3910) #2 0x0001007c7a1c in test::run_test::haeb3c2bcd5ed6cf6+0x76c (predict_oai-98384e17fb94e863:arm64+0x100047a1c) #3 0x0001007aedb0 in test::console::run_tests_console::he9d142d704f3a986+0x149c (predict_oai-98384e17fb94e863:arm64+0x10002edb0) ggml-org#4 0x0001007c5758 in test::test_main::hf86a5e20735245b9+0x118 (predict_oai-98384e17fb94e863:arm64+0x100045758) ggml-org#5 0x0001007c5da0 in test::test_main_static::h61ee9c8fd30abca0+0x54 (predict_oai-98384e17fb94e863:arm64+0x100045da0) ... ==45482==ABORTING
sparkleholic
pushed a commit
that referenced
this pull request
Aug 20, 2026
…back (ggml-org#26952) * OpenVINO backend: 1) enable gpt-oss moe on OV bk; 2) enable mxfp4 support * OpenVINO backend: disable TOPK_MOE op test * OpenVINO Backend: Add op FILL support * OpenVINO backend: enable set rows with multi dims * fix the name missmatch in setrow + view * OpenVINO backend: enable op GGML_UNARY_OP_SIGMOID * OpenVINO Backend: enable SQR & SQRT * OpenVINO backend: 1) ensure unique node names for OpenVINO; 2) add org_src to recorde the src ggml tensor for OpenVINO dynamic shape infer * OpenVINO backend: enable fallback for openVINO to CPU backend * OpenVINO backend: fix accurace issue in gemma3n arch test * fix mpt failed case * OpenVINO backend: clean nodeinfo * OpenVINO Backend: enable zero-size copy for view * add concat ssm_conv in compute_dynamic_dim enable qwen35 Fix after rebase remove logging * OpenVINO backend: disable EXP with FP32, which failed in op test. Root reason: the backend test initializes unary op inputs over a wide range, [-150, 150]. For FP32, exp(x) overflows around x ~= 88.7, so this test can randomly generate values right in or beyond the overflow region * OpenVINO backend: fix CPY op test failed issue * OpenVINO backend: fix GATED_DELTA_NET op test failed issue * handle in-place op, handle qwen35 dynamic clearing of cache in cgraph * handle qwen35 dynamic clearing of cache correctly * Enable qwen35 dense multi seq * Fix qwen35 9b gqa * Fix after rebase * Disable SOLVE_TRI * openvino: fix NEOX RoPE accuracy on GPU stateful (mixed-rank Multiply) In stateful mode the NEOX RoPE branch fed rank-3 data ([S, n_heads, head_size]) into the Multiply against the rank-4 cos/sin tables ([1, S, 1, n_dims/2]). That mixed-rank broadcast is miscomputed by the OpenVINO GPU plugin, corrupting the rotated Q/K and producing garbage output (e.g. Phi-3-mini). Lift the data to rank-4 before the split/ Multiply so the operands are equal-rank, matching what the TYPE_NORMAL branch already does. CPU and stateless paths are unaffected. Phi-3-mini-Q4_K_M, wiki.test perplexity, GPU stateful: before: PPL = 27120.43 after: PPL = 6.2263 (CPU reference: 6.2251) * OpenVINO backend: 1) remove the unique name in llama.cpp; 2) add new ov name in ov bk; 3) fix issue in arch test & op test with latest code update * OpenVINO Backenb: remove changes in llama.cpp * Doc change (use x64 Native Tools Command Prompt for VS) * Cleaner sentence Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * OpenVINO Backend: cache key upgrade includes all src name * OpenVINO Backend: enable llama arch test on ci * OpenVINO Backend: move parameter node creating from decoder into translate * OpenVINO Backend: create extra input ov node move from decoder to translate * fix for op regression due to is_model_splitted * openvino: fix CPY writeback for recurrent state rollback Detect the rollback conv/gdn state writeback CPY nodes structurally instead of by tensor name, since the rollback path in build_conv_state does not call cb() and left the nodes unnamed. Add per-node runtime offsets (rs_slot_begin_*, rs_src_begin_*) so the cached IR handles any kv head, sequence count and snapshot slot for both the conv state and the GDN state writeback. Assisted-by: GitHub Copilot * qwen35 moe * optimize MoE expert aggregation with ReduceSum * Skip GET_ROWS inaccurate test * openvino: fallback dynamic MUL_MAT_ID shapes * OpenVINO Backend: fix error in arch test model mpt * fix error caused by cpy in arch test model kimi-linear * OpenVINO Backend: fix error in arch test model minimax-m3 * openvino: fix GPU mul_mat_id op tests * ggml-openvino: add GGML_OPENVINO_RELEASE_WEIGHTS to reclaim host weight RSS on GPU The OpenVINO weight Constants are zero-copy views into host buffers allocated by the backend (ggml_aligned_malloc, anonymous memory). On GPU the plugin holds its own device copy after compile_model, so these host pages are dead weight for inference. For a 1B Q4_K_M model this leaves ~850 MB of host RSS resident that the GPU path never reads again. Add an opt-in GGML_OPENVINO_RELEASE_WEIGHTS mode that madvise(MADV_DONTNEED)s the registered host weight buffers once the model is compiled, dropping their resident pages while keeping the mappings valid (ggml still owns the lifetime; tensors still point in). Measured steady-state RSS drops from ~1555 MB to ~710 MB on Llama-3.2-1B-Q4_K_M (Arc iGPU) with unchanged throughput and correct output. The GPU backend uses a single dynamic-shape model for both prefill and decode, so a graph is compiled once and reused; the only event that forces a recompile is clear_caches() on backend teardown. The change therefore: - releases on the first cache-hit (model compiled, plugin has its copy); - pins the compiled-model cache across backend teardown so a later context reuses it instead of recompiling against the dropped pages; - fails loud (GGML_ABORT) on a cache-miss recompile or on a second model load, both of which would otherwise read zeroed weights or silently reuse the wrong compiled graph. Scope/limitations (all fail loud, never silently wrong): GPU only (the CPU plugin reads the host Constants at inference time), one model per process, and stable graph shapes. This reduces steady-state RSS, not the transient compile-time peak. All changes are confined to the OpenVINO backend. * ggml-openvino: stream weight requantization to cut the compile-time RSS peak requantize_to_buffers() dequantized the entire tensor to a temporary std::vector<float> of n_elements before requantizing. For token_embd.weight (128256 x 2048) that transient is ~1 GB (1B model) / ~2 GB (8B), and it is the single largest contributor to the OpenVINO compile-time memory peak -- it also fires twice for token_embd (once at load, once at graph build, because token_embd is loaded via a CPU/mmap buffer and not cached as an OV weight extra). Stream the dequant instead: process a fixed window of complete rows (CHUNK_ROWS=256) into a small scratch buffer and quantize/convert each chunk straight into the output buffers. The transient F32 footprint is now CHUNK_ROWS*ne0 floats regardless of tensor size. quantize_q8_0/q8_1 gain an optional block_offset arg (default 0) so a chunk writes its weights/scales/zp at the correct block. Streaming is applied to the Q8_0_C / Q8_1_C / F16 targets (the large requant cases); the u4 (Q4_0) path keeps the whole-array call because it packs two weights per byte with running zp ORs, and a fallback handles any future target whose block size does not divide a row. Measured peak RSS (cold compile, GPU): 1B 2868 -> 1809 MB (-1.06 GB); 8B 11618 -> 9608 MB (-2.0 GB). Output verified unchanged ("capital of France is Paris"); throughput unchanged. Unlike GGML_OPENVINO_RELEASE_WEIGHTS this reduces the transient peak, not just steady-state, and needs no env flag. All changes confined to the OpenVINO backend. * ggml-openvino: avoid redundant token_embd requantization at compile token_embd.weight is referenced twice in the graph path: as the GET_ROWS embedding (a CPU/mmap-buffer tensor) it was re-extracted/re-requantized on every weight-node build, and is_model_splitted() built a full (naive) set of weight nodes just to test name membership — each requant is a ~1-2 GB F32 dequant of the 262M-element embedding. Two changes: - Add collect_weight_names(): a name-only collector for topology checks. is_model_splitted() now uses it instead of create_weight_nodes(cgraph, true), so the splitted-check no longer triggers any weight extraction. - Memoize weight nodes built from non-OpenVINO buffers in a process-lifetime cache keyed by tensor->data. These tensors have no OV buffer context to own a cached extra, so without this they were rebuilt on every (re)compile; prefill and decode graphs now share one build (verified: 2nd graph hits the cache instead of re-requantizing). Peak RSS is unchanged (the streaming-requant commit already removed the F32 transient); this removes redundant compile-time work. Output verified unchanged ("capital of France is Paris"). Confined to the OpenVINO backend. * ggml-openvino: gate compile-memory optimizations behind GGML_OPENVINO_REDUCE_COMPILE_MEM The streaming requantization and the non-OpenVINO-buffer weight-node cache (plus the name-only is_model_splitted path that pairs with it) are now opt-in via GGML_OPENVINO_REDUCE_COMPILE_MEM. When unset, requantize_to_buffers() fully materializes the F32 buffer and weights are rebuilt per compile exactly as before; when set, the streaming path and the cross-compile weight cache are used. Default off keeps behavior identical to upstream unless explicitly enabled. Verified: flag off -> peak RSS 2800 MB (original), flag on -> 1810 MB; output "capital of France is Paris" in both modes. (GGML_OPENVINO_RELEASE_WEIGHTS, added earlier, remains a separate opt-in for the steady-state release.) * ggml-openvino: add frontend model cache (GGML_OPENVINO_MODEL_CACHE_DIR) The plugin-level ov::cache_dir caches the compiled blob keyed by the OV model, but producing that model still runs the full frontend every time: weight requantization (incl. the large token_embd F32 transient) and the ggml->OV graph conversion. This adds an opt-in frontend cache keyed off a fingerprint computed directly from the ggml cgraph, so a hit imports a previously exported CompiledModel and skips requant + convert + compile entirely. Key (model-cache.{h,cpp}) = 64-bit FNV-1a of: graph topology (n_nodes + per node op/name), a sampled per-weight fingerprint (name/shape/type + bounded head+tail byte sample), and blob-affecting config (device, flash-attn, rope params, REDUCE_COMPILE_MEM/stateful flags, OpenVINO version). A sidecar manifest stores every weight's fingerprint and is re-verified on load, so a sampled-hash collision cannot cause a wrong-model hit (verified: two different quantizations of the same model produce distinct cache entries). Flow (dynamic single-model path only; split models defer to ov::cache_dir): on a verified hit, core.import_model() restores the CompiledModel and a lightweight decoder is built with a names-only weight map (membership is all the decoder needs for I/O mapping; weights live in the imported model). On a miss, compile as usual then export the blob (atomic temp+rename, manifest written first). The frontend cache supersedes ov::cache_dir, so CACHE_DIR/ CACHE_MODE are stripped from the config used for the cached compile and the import — a blob compiled with cache_dir set cannot be re-imported. Measured 8B Q4_K_M (GPU): full requant+convert+compile 15.3s -> import 6.3s (~2.4x faster compile phase). Output verified unchanged on cold and warm, standalone and combined with REDUCE_COMPILE_MEM + RELEASE_WEIGHTS. Default off; confined to the OpenVINO backend. * ggml-openvino: harden frontend model cache correctness The frontend model cache imports a previously exported CompiledModel keyed by a fingerprint of the ggml graph, weights, and blob-affecting config. The original key covered device, stateful execution, REDUCE_COMPILE_MEM, RoPE params, OpenVINO version, topology, and sampled weights, but missed runtime/frontend toggles that can change the lowered graph or the I/O binding contract. That made it possible to reuse a blob produced under a different OpenVINO backend configuration. Add a small extra-config helper for the dynamic model-cache path and fold in the effective values of GGML_OPENVINO_DISABLE_KV_SLICE and GGML_OPENVINO_MANUAL_GQA_ATTN. MANUAL_GQA_ATTN is keyed by the behavior that actually takes effect: an explicit env value wins, otherwise GPU defaults to enabled and other devices default to disabled. This matches flash_attn_ext lowering and avoids unnecessary cache splits for equivalent configurations while separating genuinely different attention graphs. DISABLE_KV_SLICE is also included because it changes the KV-cache tensor shape/output binding strategy used around imported models. Even when weights and graph topology are identical, switching this flag should not inherit a CompiledModel cache entry created for a different binding mode. Also make cache artifact publication cleaner: write manifest.tmp and blob.tmp, publish the blob first, and publish the manifest last. Cache hits already require both blob and a verified manifest, so making the manifest the final visible artifact avoids leaving an apparently complete manifest for a failed or interrupted blob export. Temporary files are removed on the handled failure paths. While touching this path, fix the indentation of the non-imported compile branch so the cache miss flow is easier to review. Behavior is otherwise unchanged: verified hits still import, misses still create weights, convert, compile, export, and create the infer request normally. * ggml-openvino: add memory optimization umbrella switch Add GGML_OPENVINO_MEMORY_OPTIMIZE as a single opt-in switch for the OpenVINO backend memory-saving paths. The existing fine-grained GGML_OPENVINO_REDUCE_COMPILE_MEM and GGML_OPENVINO_RELEASE_WEIGHTS variables remain supported and explicitly override the umbrella switch when set, so users can still bisect or disable one side of the optimization independently. Centralize the policy in ggml_openvino_reduce_compile_mem_enabled() and ggml_openvino_release_weights_enabled(device). The umbrella switch enables compile-memory reductions everywhere REDUCE_COMPILE_MEM is used today: streaming requantization, non-OV weight-node caching, split-model weight-name collection, and the frontend model-cache fingerprint. On GPU it also enables host weight-buffer release unless GGML_OPENVINO_RELEASE_WEIGHTS is explicitly set. Keep host weight release GPU-only because it relies on the plugin holding its own device copy after compile_model. Update the fail-fast diagnostic and comments to mention GGML_OPENVINO_MEMORY_OPTIMIZE, so users who enable the umbrella switch get accurate guidance if a later cache-miss recompile would read released host weight pages. * ggml-openvino: rename compiled model cache env Rename the frontend export/import cache environment variable from GGML_OPENVINO_MODEL_CACHE_DIR to GGML_OPENVINO_COMPILED_MODEL_CACHE_DIR. The cache stores blobs produced by ov::CompiledModel::export_model() and restores them with core.import_model(), so the new name distinguishes it from GGML_OPENVINO_CACHE_DIR, which configures OpenVINO plugin-level ov::cache_dir. Update the registered env var, the cache-directory lookup, and comments around the frontend compiled-model cache. The old GGML_OPENVINO_MODEL_CACHE_DIR name is removed rather than kept as a fallback so there is a single spelling for the new option. * docs: document OpenVINO memory optimization env vars Add runtime configuration entries for the newly recognized OpenVINO environment variables. Document GGML_OPENVINO_COMPILED_MODEL_CACHE_DIR as the frontend compiled-model cache used to export and import compiled blobs for matching single-graph models. Document GGML_OPENVINO_MEMORY_OPTIMIZE as the umbrella switch, including how GGML_OPENVINO_REDUCE_COMPILE_MEM and the GPU-only GGML_OPENVINO_RELEASE_WEIGHTS override or inherit from it. * ggml-openvino: fix Qwen3VL crash and deepstack correctness bug 1. GGML_OP_PAD was missing from compute_node_dynamic_dims(), causing a crash on decode for models that pad the token embedding (n_embd -> n_embd_inp). PAD never reorders/merges dims, so it keeps the same dynamic dim index as its source. 2. process_view_input_new() chained VIEW inputs through src[0] (the immediate op-graph parent) using offsets treated as relative to that parent. But ggml_tensor::view_offs is always absolute from the true root allocation (ggml collapses VIEW-of-VIEW chains internally). For the per-layer deepstack view ("embd (view)", whose src[0] is "embd" - itself an already-narrowed, zero-offset VIEW of the padded root, with the SAME ggml shape as the deepstack view but a different absolute offset), this caused an out-of-bounds re-slice that silently fell back to returning the wrong (already-resolved sibling) tensor. In practice every deepstack ADD ended up adding the real base token embedding into the residual stream instead of zero, corrupting generation ("Hello my name is 1000000..." instead of coherent text). Fixed by detecting this pattern (same shape as the immediate src, different absolute offset) and re-slicing directly from the untouched root tensor using the innermost view's absolute offset. Also adds a GGML_OPENVINO_DEBUG_NODE=<name1>,<name2>,... env var that attaches extra debug Result nodes for arbitrary intermediate tensors, without binding them to any ggml buffer (avoiding the risk of reading a ggml buffer that has since been overwritten by a later in-place op). This was instrumental in diagnosing bug #2 above and is left in as a general-purpose debugging aid. * ggml-openvino: fix IMROPE inp_pos padding for NPU static shapes IMROPE's inp_pos tensor packs 4 stacked t/h/w/e position planes into ne[0] = 4*n_tokens instead of one value per token. On NPU's static-shape path, inp_pos was padded/shaped as if it held a single plane, which interleaved padding across the 4 planes and desynced later reshapes from the rest of the (chunk_size-wide) graph. - add GgmlOvDecoder::get_inp_pos_n_planes() to detect IMROPE's 4-plane layout - get_graph_input_shape(): size inp_pos as n_planes * chunk_size (prefill) or n_planes (decode) instead of assuming 1 value per token - get_ov_input_tensor_static_prefill(): pad each plane to chunk_size independently instead of one flat block - get_ov_input_tensor_static_decode(): copy n_planes contiguous values instead of asserting/copying a single scalar * disable test-llama-archs tests. * openvino: gate fallback with env var * Revert changes in test-llama-archs * Apply editor config * reject CPY with quantized destination as unsupported --------- Co-authored-by: Xuejun <Xuejun.Zhai@intel.com> Co-authored-by: Mustafa Cavus <mustafa.cavus@intel.com> Co-authored-by: virajwad <84867530+virajwad@users.noreply.github.com> Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> Co-authored-by: suryasidd <surya.siddharth.pemmaraju@intel.com> Co-authored-by: Mustafa Cavus <mustafacavus@intel.com> Co-authored-by: Ravi Panchumarthy <ravi.panchumarthy@intel.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Make sure to read the contributing guidelines before submitting a PR