Skip to content

Commit 25c80b9

Browse files
hodlenggerganovMrkvak
committed
Squash commits until Mixtral support for merging
llama : restore prefix space in llama tokenizer (ggml-org#4081) gguf : fix potential infinite loops while parsing (ggml-org#4100) Co-authored-by: Bernhard Gstrein <gstrein@cs.uni-freiburg.de> Respect tokenizer.ggml.add_bos_token value when tokenizing (ggml-org#4040) * gguf-py: gguf-dump: Respect --no-tensor flag in JSON mode. * Respect add_bos_token GGUF metadata value * gguf-py: Try to fix SpecialVocab giving up too easily for the Nth time llama : fix data units (ggml-org#4101) * llama : fix data units ggml-ci * Revert "llama : fix data units" This reverts commit f5feac8. * llama : disambiguate data units ggml-ci cuda : get_row_rounding F32 (ggml-org#4095) * Fix ggml-org#4017 * Update ggml-cuda.cu Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Update ggml-cuda.cu Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> finetune : zero the loraB initial vectors (ggml-org#4082) * finetune : zero the loraB initial vectors Without this, the first iteration is starting out far from the base model, instead of exactly on it. Zeroing loraB is what the paper recommends. loralib also zeroes at least one of the init vector pairs (though it departs from the paper in using a different distribution for the other vector, in some cases). * tabs to spaces * Use ggml_set_zero instead of adding a new function finetune : speed-up ggml_compute_forward_out_prod_f32 via BLAS (ggml-org#4079) * Remove logically superfluous assertions and order by dimension * Use cblas_sgemm() to implement ggml_compute_forward_out_prod() * Remove ggml_compute_forward_out_prod_use_blas(), fix compiling errors on cmake/zig, remove trailing whitespace * Add openBLAS support for sgemm() in compute_forward_out_prod() llama : add functions to get the model's metadata (ggml-org#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation train : move number of gpu layers argument parsing to common/train.cpp (ggml-org#4074) - introduces help entry for the argument - cuts '--gpu-layers' form in order to simplify usage and documentation. Signed-off-by: Jiri Podivin <jpodivin@gmail.com> Co-authored-by: Jiri Podivin <jpodivin@redhat.com> py : remove superfluous import statements (ggml-org#4076) Signed-off-by: Jiri Podivin <jpodivin@gmail.com> Co-authored-by: Jiri Podivin <jpodivin@redhat.com> llava : fix compilation warning that fread return value is not used (ggml-org#4069) common : improve yaml log escaping (ggml-org#4080) * logging: improve escaping in yaml output * logging: include review feedback py : Falcon HF compatibility (ggml-org#4104) Falcon HF compatibility convert : use 'model' value if it exists. This allows karpathy/tinyllamas to load (ggml-org#4089) Co-authored-by: Don Mahurin <@> examples : add tokenize (ggml-org#4039) tokenize : fix trailing whitespace build : support ppc64le build for make and CMake (ggml-org#3963) * build: support ppc64le build for make and CMake * build: keep __POWER9_VECTOR__ ifdef and extend with __powerpc64__ Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> llama : increase max nodes (ggml-org#4115) Clean up ggml-cuda.cu warnings when compiling with clang (for ROCM) (ggml-org#4124) * ggml-cuda.cu: Clean up warnings when compiling with clang * ggml-cuda.cu: Move static items into anonymous namespace * ggml-cuda.cu: Fix use of namespace start macro * Revert "ggml-cuda.cu: Fix use of namespace start macro" This reverts commit 26c1149. * Revert "ggml-cuda.cu: Move static items into anonymous namespace" This reverts commit e29757e. scripts : Remove missed baichuan convert script (ggml-org#4127) tokenize example: Respect normal add BOS token behavior (ggml-org#4126) Allow building with Makefile gguf-py : export chat templates (ggml-org#4125) * gguf-py : export chat templates * llama.cpp : escape new lines in gguf kv info prints * gguf-py : bump version * gguf-py : check chat_template type * gguf-py : initialize chat_template gitignore : tokenize common : comma should be semicolon (ggml-org#4137) server : relay error messages (ggml-org#4131) finetune : add --n-gpu-layers flag info to --help (ggml-org#4128) Revert "finetune : add --n-gpu-layers flag info to --help (ggml-org#4128)" This reverts commit 05e8301. speculative : fix prompt tokenization in speculative example (ggml-org#4025) * Support special tokens and not adding BOS to prompt in speculative * Adapt to new should_add_bos function * Ensure tgt and dft have same add_bos setting ci : add flake8 to github actions (python linting) (ggml-org#4129) Disabled rules: * E203 Whitespace before ':' - disabled because we often use 'C' Style where values are aligned * E211 Whitespace before '(' (E211) - disabled because we often use 'C' Style where values are aligned * E221 Multiple spaces before operator - disabled because we often use 'C' Style where values are aligned * E225 Missing whitespace around operator - disabled because it's broken so often it seems like a standard * E231 Missing whitespace after ',', ';', or ':' - disabled because we often use 'C' Style where values are aligned * E241 Multiple spaces after ',' - disabled because we often use 'C' Style where values are aligned * E251 Unexpected spaces around keyword / parameter equals - disabled because it's broken so often it seems like a standard * E261 At least two spaces before inline comment - disabled because it's broken so often it seems like a standard * E266 Too many leading '#' for block comment - sometimes used as "section" separator * E501 Line too long - disabled because it's broken so often it seems like a standard * E701 Multiple statements on one line (colon) - broken only in convert.py when defining abstract methods (we can use# noqa instead) * E704 Multiple statements on one line - broken only in convert.py when defining abstract methods (we can use# noqa instead) main : Add ChatML functionality to main example (ggml-org#4046) Co-authored-by: Sebastian Cramond <sebby37@users.noreply.github.com> readme : update ROCm Windows instructions (ggml-org#4122) * Update README.md * Update README.md Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> finetune - update readme to mention llama support only (ggml-org#4148) stablelm : simplify + speedup generation (ggml-org#4153) docs : add llama-star arch idea examples : fix typo in parallel example doc comment (ggml-org#4181) Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com> readme : update hot topics llama : KV cache view API + better KV cache management (ggml-org#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (ggml-org#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com> Fix incorrect format strings and uninitialized variables. (ggml-org#4133) * Fix incorrect format strings and uninitialized variables. * Address comments * Add the missing include statement readme : use PATH for Windows ROCm (ggml-org#4195) * Update README.md to use PATH for Windows ROCm * Update README.md * Update README.md main.swift : fix eos checking (ggml-org#4197) llama_token_eos(const struct llama_model *) is currently getting struct llama_context type variable context as a parameter. convert : fix tensors using grad in some models (ggml-org#4173) ggml-cuda : support stablelm rope (ggml-org#4156) * ggml-cuda : support stablelm rope * remove unused freq_base kernel parameter * add n_dims parameter to llm_build_k_shift, default to n_rot via overload * llama : fix llm_build_k_shift args --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> llama : set metal log callback correctly (ggml-org#4204) server : OAI API compatibility (ggml-org#4198) * Add openai-compatible POST /v1/chat/completions API endpoint to server example * fix code style * Update server README.md * Improve server README.md * Fix server.cpp code style according to review * server : some style changes * server : indentation * server : enable special tokens during tokenization by default * server : minor code style * server : change random string generator * straightforward /v1/models endpoint --------- Co-authored-by: kir-gadjello <111190790+kir-gadjello@users.noreply.github.com> Co-authored-by: Tobi Lütke <tobi@Tobis-MacBook-Pro.local> readme : update hot topics Update docs for yarn_ext_factor <0.0 as unspecified instead of NaN (ggml-org#4189) llama : grammar `reserve` space in `decode_utf8` (ggml-org#4210) * reserve space for codepoints * improvement for the appended 0 scripts : Use mmap in torch load (ggml-org#4202) * Use mmap in torch load, prefer .bin files when loading * Revert .bin > .safetensors preference metal : fix yarn (ggml-org#4220) get the correct n_orig_ctx in metal lookahead : add example for lookahead decoding (ggml-org#4207) * lookahead : init * lookahead : generate and store n-grams * lookahead : use loop instead recursion to generate n-grams * lookahead : initial working implementation * lookahead : filter repeating n-grams * lookahead : use deterministic init * lookahead : add to Makefile * lookahead : fix a bug in the seq_id of the lookahead tokens * lookahead : add comments --------- Co-authored-by: slaren <slarengh@gmail.com> readme : update hot topics lookahead : support `-n -1` infinite generation ggml : fix -Warray-bounds warning with gcc (ggml-org#4231) examples : iOS example with swift ui (ggml-org#4159) * copy to llama.cpp as subdir * attempt enabling metal, fails * ggml metal compiles! * Update README.md * initial conversion to new format, utf8 errors? * bug fixes, but now has an invalid memory access :( * added O3, now has insufficient memory access * begin sync with master * update to match latest code, new errors * fixed it! * fix for loop conditionals, increase result size * fix current workflow errors * attempt a llama.swiftui workflow * Update .github/workflows/build.yml Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> readme : add Amica to UI list (ggml-org#4230) cmake : fix issue with version info not getting baked into LlamaConfig.cmake (ggml-org#3970) * Split CPP generation from build-info query * Remove blank lines * Add BUILD_SHARED_LIBS option ggml : re-enable BLAS for CPU when src0 != F32 + remove redundant full offload checks in llama.cpp (ggml-org#4240) * ggml : use blas even if src0 is not F32 * llama : use n_threads_batch only when n_tokens >= 32 ggml-ci * llama : revert n_threads_batch logic ggml-ci ggml : restore abort() in GGML_ASSERT (ggml-org#4242) readme : add FreeChat (ggml-org#4248) examples : add readme files py : fix oai proxy (ggml-org#3972) * fix oai proxy fix generation not stoped while bot stop talking in chat mode fix possible `slot_id` not exist response for cors (and pre flight) * oai proxy: workaround for some client (such as Chatbox) * use stop as separator to replace hardcoded `\n` llama : fix typical sampling (ggml-org#4261) Typical sampling was broken because after copying new_candidates into canditates, the "sorted" bool is left at "true", but the new data is no longer sorted according to probability. Patch to set "sorted" to false. Test: Generating with temp=0.0001 (approx. argmax) should generate the same sequence at typical>=1.0 and typical=0.9999 (approx. disabled, but enters the typical sampling codepath). convert.py : fix llama/llama2 conversion due to vocab_size=-1 (ggml-org#4258) llama : fix alignment of general.name in print meta (ggml-org#4254) * llama: fix alignment of general.name in print meta This commit fixes the alignment of the general.name field in the llm_load_print_meta function. Currently the output looks like this: ```console llm_load_print_meta: model ftype = mostly Q4_0 llm_load_print_meta: model params = 13.02 B llm_load_print_meta: model size = 6.86 GiB (4.53 BPW) llm_load_print_meta: general.name = LLaMA v2 ``` And with this commit it looks like this: ```console llm_load_print_meta: model ftype = mostly Q4_0 llm_load_print_meta: model params = 13.02 B llm_load_print_meta: model size = 6.86 GiB (4.53 BPW) llm_load_print_meta: general.name = LLaMA v2 ``` Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com> * llama: fix alignment of special tokens Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com> --------- Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com> readme : fix typo (ggml-org#4253) llama.cpp uses GitHub Actions, not Gitlab Actions. cmake : fix the metal file foder path (ggml-org#4217) batched.swift : update README.md (ggml-org#4214) docs: update how to run docker : add finetune option (ggml-org#4211) readme : fix (ggml-org#4135) * fix: readme * chore: resolve comments * chore: resolve comments main : pass LOG_TEE callback to llama.cpp log (ggml-org#4033) * main : Call llama_log_set to use LOG_TEE * tabs to spaces llava : ShareGPT4V compatibility (vision encoder only loading) (ggml-org#4172) * ShareGPT4 compatibility (vision encoder only loading) Load only a CLIP vision encoder (as supplied by ShareGPT finetunes) Corrects the argument parsing for --img_mean and --img_std (which were previously not parsed but attempted to access) Defines defaults for img_mean and img_std which are equal to the llava 1.5 CLIP encoder, so you do not have to provide them * Update convert-image-encoder-to-gguf.py build : fix build info generation and cleanup Makefile (ggml-org#3920) * cmake : fix joining of REAL_GIT_DIR * fix includes with help from include-what-you-use * make : remove unneeded deps and add test-rope target * fix C includes in C++ source files * Revert "fix includes with help from include-what-you-use" This reverts commit 635e9fa. make : fix Apple clang determination bug (ggml-org#4272) Co-authored-by: Will Findley <findley@gmail.com> server : add single-client multi-prompt support (ggml-org#4232) * * add multiprompt support * * cleanup * * more cleanup * * remove atomicity of id_gen, and change lock_guard to unique_lock on completion requests * * remove all references to mutex_multitasks * Update examples/server/server.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Update examples/server/server.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Update examples/server/server.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Update examples/server/server.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * * change to set --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> server : add --log-disable to disable logging to file (ggml-org#4260) * * add --log-disable to disable logging to file in the server example * * typo fix ggml : add ggml_soft_max_ext (ggml-org#4256) * metal : implement soft_max_ext * cuda : implement soft_max_ext * ggml : implement soft_max_ext (CPU) * batched-bench : print threads ggml-ci * metal : simplify soft_max encoding ggml-ci * cuda : use 512 threads for soft_max instead of 32 * ggml : update soft max cpu * cuda : do warp-based block reduce * cuda : increase max block size to 1024 * cuda : fix warp reduction initialization of shared mem * metal : warp-based reduction for soft max kernel * metal : warp-based reduce for rms_norm * metal : simplify soft max kernel ggml-ci * alloc : fix build with debug py : add requirements file for convert-hf-to-gguf.py (ggml-org#4277) This commit adds a requirements file for the convert-hf-to-gguf.py script, and also add the torch and transformers packages to it. The motivation for this is that currently running convert-hf-to-gguf.py will produce the following error: ```console $ python3 -m venv venv $ source venv/bin/activate (venv) $ pip install -r requirements.txt Collecting numpy==1.24.4 Collecting sentencepiece==0.1.98 Collecting gguf>=0.1.0 Installing collected packages: sentencepiece, numpy, gguf Successfully installed gguf-0.5.1 numpy-1.24.4 sentencepiece-0.1.98 (venv) $ python convert-hf-to-gguf.py --help Traceback (most recent call last): File "llama.cpp/convert-hf-to-gguf.py", line 16, in <module> import torch ModuleNotFoundError: No module named 'torch' ``` With this commit, and using requirements-hf-to-gguf.txt instead of requirements.txt, the script can be run and shows the help output. Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com> llama : fix integer overflow during quantization (ggml-org#4284) happens with multi-threaded quantization of Qwen-72B ggml-ci llama : add Qwen support (ggml-org#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> llama : support attention bias on LLaMA architecture (ggml-org#4283) * Support attention_bias on LLaMA architecture QKVO bias, should fix InternLM (ggml-org#3133) and works for LLaMAfied Qwen models (ggml-org#3743 (comment)). * check existence of qkvo bias while loading llama models Tested on LLaMA2, CUDA and CPU. * Update llama.cpp build : enable libstdc++ assertions for debug builds (ggml-org#4275) swift : fix token_to_piece implementation (ggml-org#4278) * Fix token_to_piece implementation in Swift * Fix errors llama : support optional tensors (ggml-org#4283) llama : avoid using "optional" keyword (ggml-org#4283) llama : pad KV cache size (ggml-org#4280) * llama : pad KV cache size to 32 * metal : try to improve batched decoding py : add grammar to oai like api (ggml-org#4294) server : fix OpenAI API `stop` field to be optional (ggml-org#4299) (cherry picked from commit Mozilla-Ocho/llamafile@e8c92bc) ggml : fix soft max out-of-bounds access (ggml-org#4307) ggml-ci ggml : reuse ggml_get_n_tasks() in ggml_graph_plan() (ggml-org#4308) * ggml : fix soft max out-of-bounds access ggml-ci * ggml : reuse ggml_get_n_tasks() in ggml_graph_plan() ggml-ci grammar-parser : fix typo (ggml-org#4318) preceeding -> preceding swift : fix prompt tokenization logic (ggml-org#4321) swift : fix concatenation method to avoid invalid UTF8 stringfication (ggml-org#4325) simple : update error message for KV cache check (ggml-org#4324) This commit updates the error message that is printed when the KV cache is not big enough to hold all the prompt and generated tokens. Specifically it removes the reference to n_parallel and replaces it with n_len. Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com> swift : revert compiler checks for swift package (ggml-org#4332) sampling : custom samplers order (ggml-org#4285) * Samplers sequence order w parameter * Cleaned commented code * Fixed formatting * Rewrote with unordered_map * Revert and rewrite, too many problems and safeguards would be needed * Fixed code style * Code style fixes according to review * More readable samplers input string, fixed help * Style fix in sampler_queue * Formatting fixes * Fixing whitespaces llama : allow overriding GGUF metadata when loading model (ggml-org#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> grammar : pre-computed pieces + reserve mem + less string copies (ggml-org#4330) * reserve space for codepoints * improvement for the appended 0 * used precomputed token text for grammar sample * reserve canidates_decoded * reserve canidates_grammar * remove candidates_decoded * Revert "remove candidates_decoded" This reverts commit 3773328. * changed decode_utf8 to take src by ref speculative : support `--color` (ggml-org#4343) * speculative: add some colors * minor : add braces --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> common : fix compile warning server : recognize cache_prompt parameter in OAI API (ggml-org#4347) train : fix ggml-org#4227 (double free in examples/train-text-from-scratch/train-text-from-scratch.cpp) (ggml-org#4351) On commit b1108 (44c117f) xaedes added ggml_allocr * alloc = NULL; ... (many lines in between) if (alloc) { ggml_allocr_free(alloc); } Which is correct, but it's easy to lose context after many lines in between. On commit b1287 (0e76a899) xaedes made a big change. From here on, alloc is freed eagerly. alloc = ggml_allocr_new(...) ... (short lines of code) ggml_allocr_free(alloc) This happens a few times, but alloc is never set to NULL, and many lines below, we still have if (alloc) { ggml_allocr_free(alloc); } which causes a double-free. llama : per-layer KV cache + quantum K cache (ggml-org#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (ggml-org#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com> sync : ggml (new ops, tests, backend, etc.) (ggml-org#4359) * sync : ggml (part 1) * sync : ggml (part 2, CUDA) * sync : ggml (part 3, Metal) * ggml : build fixes ggml-ci * cuda : restore lost changes * cuda : restore lost changes (StableLM rope) * cmake : enable separable compilation for CUDA ggml-ci * ggml-cuda : remove device side dequantize * Revert "cmake : enable separable compilation for CUDA" This reverts commit 09e35d0. * cuda : remove assert for rope * tests : add test-backend-ops * ggml : fix bug in ggml_concat * ggml : restore `ggml_get_n_tasks()` logic in `ggml_graph_plan()` * ci : try to fix macOS * ggml-backend : remove backend self-registration * ci : disable Metal for macOS cmake build ggml-ci * metal : fix "supports family" call * metal : fix assert * metal : print resource path ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com> grammar : revert the replacement of llama_token_to_piece with id_to_token (ggml-org#4396) Update README.md (ggml-org#4388) Fix small typo. ggml : increased GGML_MAX_PARAMS to allow finetuning of 70b models (ggml-org#4424) server : fix local model name in server (ggml-org#4420) llama : document logits_all deprecation (ggml-org#4418) llama_context_params.logits_all is a parameter for controlling llama_eval. This documents that logits_all should not be used with llama_decode and llama_batch. build : target Windows 8 for standard mingw-w64 (ggml-org#4405) * build : target Windows 8 for standard mingw-w64 * make : fix missing console.o deps This was causing a link error with `make all` on Windows. english : use `typos` to fix comments and logs (ggml-org#4354) server : tweak default sampling parameters (ggml-org#4367) * Set a more typical Top P setting as the default * Update temp max llama : add Mixtral support (ggml-org#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (ggml-org#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: ggml-org/ggml@1914017 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
1 parent 1cf2850 commit 25c80b9

File tree

105 files changed

+12091
-2787
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

105 files changed

+12091
-2787
lines changed

‎.devops/tools.sh

+4
Original file line numberDiff line numberDiff line change
@@ -13,6 +13,8 @@ elif [[ "$arg1" == '--quantize' || "$arg1" == '-q' ]]; then
1313
./quantize "$@"
1414
elif [[ "$arg1" == '--run' || "$arg1" == '-r' ]]; then
1515
./main "$@"
16+
elif [[ "$arg1" == '--finetune' || "$arg1" == '-f' ]]; then
17+
./finetune "$@"
1618
elif [[ "$arg1" == '--all-in-one' || "$arg1" == '-a' ]]; then
1719
echo "Converting PTH to GGML..."
1820
for i in `ls $1/$2/ggml-model-f16.bin*`; do
@@ -34,6 +36,8 @@ else
3436
echo " ex: --outtype f16 \"/models/7B/\" "
3537
echo " --quantize (-q): Optimize with quantization process ggml"
3638
echo " ex: \"/models/7B/ggml-model-f16.bin\" \"/models/7B/ggml-model-q4_0.bin\" 2"
39+
echo " --finetune (-f): Run finetune command to create a lora finetune of the model"
40+
echo " See documentation for finetune for command-line parameters"
3741
echo " --all-in-one (-a): Execute --convert & --quantize"
3842
echo " ex: \"/models/\" 7B"
3943
echo " --server (-s): Run a model on the server"

‎.github/workflows/build.yml

+22-4
Original file line numberDiff line numberDiff line change
@@ -143,6 +143,9 @@ jobs:
143143
cd build
144144
ctest --verbose
145145
146+
# TODO: build with LLAMA_NO_METAL because test-backend-ops fail on "Apple Paravirtual device" and I don't know
147+
# how to debug it.
148+
# ref: https://github.com/ggerganov/llama.cpp/actions/runs/7131777249/job/19420981052#step:5:1124
146149
macOS-latest-make:
147150
runs-on: macos-latest
148151

@@ -160,14 +163,18 @@ jobs:
160163
- name: Build
161164
id: make_build
162165
run: |
163-
make -j $(sysctl -n hw.logicalcpu)
166+
LLAMA_NO_METAL=1 make -j $(sysctl -n hw.logicalcpu)
164167
165168
- name: Test
166169
id: make_test
167170
run: |
168-
make tests -j $(sysctl -n hw.logicalcpu)
169-
make test -j $(sysctl -n hw.logicalcpu)
171+
LLAMA_NO_METAL=1 make tests -j $(sysctl -n hw.logicalcpu)
172+
LLAMA_NO_METAL=1 make test -j $(sysctl -n hw.logicalcpu)
170173
174+
# TODO: build with LLAMA_METAL=OFF because test-backend-ops fail on "Apple Paravirtual device" and I don't know
175+
# how to debug it.
176+
# ref: https://github.com/ggerganov/llama.cpp/actions/runs/7132125951/job/19422043567?pr=4359#step:5:6584
177+
# would be great if we fix these
171178
macOS-latest-cmake:
172179
runs-on: macos-latest
173180

@@ -188,7 +195,7 @@ jobs:
188195
sysctl -a
189196
mkdir build
190197
cd build
191-
cmake ..
198+
cmake -DLLAMA_METAL=OFF ..
192199
cmake --build . --config Release -j $(sysctl -n hw.logicalcpu)
193200
194201
- name: Test
@@ -498,6 +505,17 @@ jobs:
498505
path: |
499506
cudart-llama-bin-win-cu${{ matrix.cuda }}-x64.zip
500507
508+
ios-xcode-build:
509+
runs-on: macos-latest
510+
511+
steps:
512+
- name: Checkout code
513+
uses: actions/checkout@v3
514+
515+
- name: Build Xcode project
516+
run: xcodebuild -project examples/llama.swiftui/llama.swiftui.xcodeproj -scheme llama.swiftui -sdk iphoneos CODE_SIGNING_REQUIRED=NO CODE_SIGN_IDENTITY= -destination 'generic/platform=iOS' build
517+
518+
501519
# freeBSD-latest:
502520
# runs-on: macos-12
503521
# steps:

‎.github/workflows/python-lint.yml

+20
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,20 @@
1+
name: flake8 Lint
2+
3+
on: [push, pull_request]
4+
5+
jobs:
6+
flake8-lint:
7+
runs-on: ubuntu-latest
8+
name: Lint
9+
steps:
10+
- name: Check out source repository
11+
uses: actions/checkout@v3
12+
- name: Set up Python environment
13+
uses: actions/setup-python@v4
14+
with:
15+
python-version: "3.11"
16+
- name: flake8 Lint
17+
uses: py-actions/flake8@v2
18+
with:
19+
ignore: "E203,E211,E221,E225,E231,E241,E251,E261,E266,E501,E701,E704"
20+
exclude: "examples/*,examples/*/**,*/**/__init__.py"

‎.gitignore

+16-12
Original file line numberDiff line numberDiff line change
@@ -47,6 +47,7 @@ models-mnt
4747
/libllama.so
4848
/llama-bench
4949
/llava-cli
50+
/lookahead
5051
/main
5152
/metal
5253
/perplexity
@@ -64,6 +65,7 @@ models-mnt
6465
/speculative
6566
/parallel
6667
/train-text-from-scratch
68+
/tokenize
6769
/vdot
6870
/common/build-info.cpp
6971
arm_neon.h
@@ -86,15 +88,17 @@ poetry.lock
8688
poetry.toml
8789

8890
# Test binaries
89-
tests/test-grammar-parser
90-
tests/test-llama-grammar
91-
tests/test-double-float
92-
tests/test-grad0
93-
tests/test-opt
94-
tests/test-quantize-fns
95-
tests/test-quantize-perf
96-
tests/test-sampling
97-
tests/test-tokenizer-0-llama
98-
tests/test-tokenizer-0-falcon
99-
tests/test-tokenizer-1-llama
100-
tests/test-tokenizer-1-bpe
91+
/tests/test-grammar-parser
92+
/tests/test-llama-grammar
93+
/tests/test-double-float
94+
/tests/test-grad0
95+
/tests/test-opt
96+
/tests/test-quantize-fns
97+
/tests/test-quantize-perf
98+
/tests/test-sampling
99+
/tests/test-tokenizer-0-llama
100+
/tests/test-tokenizer-0-falcon
101+
/tests/test-tokenizer-1-llama
102+
/tests/test-tokenizer-1-bpe
103+
/tests/test-rope
104+
/tests/test-backend-ops

‎CMakeLists.txt

+28-10
Original file line numberDiff line numberDiff line change
@@ -43,6 +43,7 @@ else()
4343
endif()
4444

4545
# general
46+
option(BUILD_SHARED_LIBS "build shared libraries" OFF)
4647
option(LLAMA_STATIC "llama: static link libraries" OFF)
4748
option(LLAMA_NATIVE "llama: enable -march=native flag" ON)
4849
option(LLAMA_LTO "llama: enable link time optimization" OFF)
@@ -96,9 +97,12 @@ option(LLAMA_METAL_NDEBUG "llama: disable Metal debugging"
9697
option(LLAMA_MPI "llama: use MPI" OFF)
9798
option(LLAMA_QKK_64 "llama: use super-block size of 64 for k-quants" OFF)
9899

99-
option(LLAMA_BUILD_TESTS "llama: build tests" ${LLAMA_STANDALONE})
100-
option(LLAMA_BUILD_EXAMPLES "llama: build examples" ${LLAMA_STANDALONE})
101-
option(LLAMA_BUILD_SERVER "llama: build server example" ON)
100+
option(LLAMA_BUILD_TESTS "llama: build tests" ${LLAMA_STANDALONE})
101+
option(LLAMA_BUILD_EXAMPLES "llama: build examples" ${LLAMA_STANDALONE})
102+
option(LLAMA_BUILD_SERVER "llama: build server example" ON)
103+
104+
# Required for relocatable CMake package
105+
include(${CMAKE_CURRENT_SOURCE_DIR}/scripts/build-info.cmake)
102106

103107
#
104108
# Compile flags
@@ -112,6 +116,11 @@ set(THREADS_PREFER_PTHREAD_FLAG ON)
112116
find_package(Threads REQUIRED)
113117
include(CheckCXXCompilerFlag)
114118

119+
# enable libstdc++ assertions for debug builds
120+
if (CMAKE_SYSTEM_NAME MATCHES "Linux")
121+
add_compile_definitions($<$<CONFIG:Debug>:_GLIBCXX_ASSERTIONS>)
122+
endif()
123+
115124
if (NOT MSVC)
116125
if (LLAMA_SANITIZE_THREAD)
117126
add_compile_options(-fsanitize=thread)
@@ -161,7 +170,7 @@ if (LLAMA_METAL)
161170
#add_compile_definitions(GGML_METAL_DIR_KERNELS="${CMAKE_CURRENT_SOURCE_DIR}/")
162171

163172
# copy ggml-metal.metal to bin directory
164-
configure_file(ggml-metal.metal bin/ggml-metal.metal COPYONLY)
173+
configure_file(ggml-metal.metal ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/ggml-metal.metal COPYONLY)
165174

166175
set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS}
167176
${FOUNDATION_LIBRARY}
@@ -574,12 +583,21 @@ elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^(x86_64|i686|AMD64)$" OR "${CMAKE_GE
574583
endif()
575584
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64")
576585
message(STATUS "PowerPC detected")
577-
add_compile_options(-mcpu=native -mtune=native)
578-
#TODO: Add targets for Power8/Power9 (Altivec/VSX) and Power10(MMA) and query for big endian systems (ppc64/le/be)
586+
if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64le")
587+
add_compile_options(-mcpu=powerpc64le)
588+
else()
589+
add_compile_options(-mcpu=native -mtune=native)
590+
#TODO: Add targets for Power8/Power9 (Altivec/VSX) and Power10(MMA) and query for big endian systems (ppc64/le/be)
591+
endif()
579592
else()
580593
message(STATUS "Unknown architecture")
581594
endif()
582595

596+
if (MINGW)
597+
# Target Windows 8 for PrefetchVirtualMemory
598+
add_compile_definitions(_WIN32_WINNT=0x602)
599+
endif()
600+
583601
#
584602
# POSIX conformance
585603
#
@@ -649,11 +667,11 @@ add_library(ggml OBJECT
649667
ggml-backend.h
650668
ggml-quants.c
651669
ggml-quants.h
652-
${GGML_SOURCES_CUDA} ${GGML_HEADERS_CUDA}
670+
${GGML_SOURCES_CUDA} ${GGML_HEADERS_CUDA}
653671
${GGML_SOURCES_OPENCL} ${GGML_HEADERS_OPENCL}
654-
${GGML_SOURCES_METAL} ${GGML_HEADERS_METAL}
655-
${GGML_SOURCES_MPI} ${GGML_HEADERS_MPI}
656-
${GGML_SOURCES_EXTRA} ${GGML_HEADERS_EXTRA}
672+
${GGML_SOURCES_METAL} ${GGML_HEADERS_METAL}
673+
${GGML_SOURCES_MPI} ${GGML_HEADERS_MPI}
674+
${GGML_SOURCES_EXTRA} ${GGML_HEADERS_EXTRA}
657675
)
658676

659677
target_include_directories(ggml PUBLIC . ${LLAMA_EXTRA_INCLUDES})

0 commit comments

Comments
 (0)