From 096d98c3d9c0b61d3491b46c526b97b5ac6721b7 Mon Sep 17 00:00:00 2001 From: Ettore Di Giacinto Date: Thu, 27 Jul 2023 21:56:05 +0200 Subject: [PATCH] fix: add rope settings during model load, fix CUDA (#821) Signed-off-by: Ettore Di Giacinto --- Makefile | 3 +- README.md | 7 +- api/backend/options.go | 28 ++-- extra/grpc/huggingface/backend_pb2.py | 36 ++--- pkg/grpc/llm/llama/llama.go | 27 ++-- pkg/grpc/proto/backend.pb.go | 221 ++++++++++++++------------ pkg/grpc/proto/backend.proto | 2 + 7 files changed, 176 insertions(+), 148 deletions(-) diff --git a/Makefile b/Makefile index e7d6edf8..7d2bb4df 100644 --- a/Makefile +++ b/Makefile @@ -4,8 +4,7 @@ GOVET=$(GOCMD) vet BINARY_NAME=local-ai # llama.cpp versions -# Temporarly pinned to https://github.com/go-skynet/go-llama.cpp/pull/124 -GOLLAMA_VERSION?=562d2b5a71195627a63bb34f639e0fb0e2b2df3f +GOLLAMA_VERSION?=6ba16de8e965e5aa0f32d25ef9d6149bb6586565 # gpt4all version GPT4ALL_REPO?=https://github.com/nomic-ai/gpt4all diff --git a/README.md b/README.md index a8e02ac6..7668cd9e 100644 --- a/README.md +++ b/README.md @@ -13,15 +13,13 @@ **LocalAI** is a drop-in replacement REST API that's compatible with OpenAI API specifications for local inferencing. It allows you to run LLMs (and not only) locally or on-prem with consumer grade hardware, supporting multiple model families that are compatible with the ggml format. Does not require GPU. -For a list of the supported model families, please see [the model compatibility table](https://localai.io/model-compatibility/index.html#model-compatibility-table). - In a nutshell: - Local, OpenAI drop-in alternative REST API. You own your data. - NO GPU required. NO Internet access is required either - Optional, GPU Acceleration is available in `llama.cpp`-compatible LLMs. See also the [build section](https://localai.io/basics/build/index.html). - Supports multiple models: - - 📖 [Text generation with GPTs](https://localai.io/features/text-generation/) (`llama.cpp`, `gpt4all.cpp`, ... and more) + - 📖 [Text generation with GPTs](https://localai.io/features/text-generation/) (`llama.cpp`, `gpt4all.cpp`, ... [:book: and more](https://localai.io/model-compatibility/index.html#model-compatibility-table)) - 🗣 [Text to Audio](https://localai.io/features/text-to-audio/) - 🔈 [Audio to Text](https://localai.io/features/audio-to-text/) (Audio transcription with `whisper.cpp`) - 🎨 [Image generation with stable diffusion](https://localai.io/features/image-generation) @@ -31,6 +29,8 @@ In a nutshell: LocalAI was created by [Ettore Di Giacinto](https://github.com/mudler/) and is a community-driven project, focused on making the AI accessible to anyone. Any contribution, feedback and PR is welcome! +Note that this started just as a [fun weekend project](https://localai.io/#backstory) in order to try to create the necessary pieces for a full AI assistant like `ChatGPT`: the community is growing fast and we are working hard to make it better and more stable. If you want to help, please consider contributing (see below)! + See the [Getting started](https://localai.io/basics/getting_started/index.html) and [examples](https://github.com/go-skynet/LocalAI/tree/master/examples/) sections to learn how to use LocalAI. For a list of curated models check out the [model gallery](https://localai.io/models/). @@ -53,6 +53,7 @@ See the [Getting started](https://localai.io/basics/getting_started/index.html) - [ ] Upstream our golang bindings to llama.cpp (https://github.com/ggerganov/llama.cpp/issues/351) - [ ] Enable gallery management directly from the webui. - [x] 🔥 OpenAI functions: https://github.com/go-skynet/LocalAI/issues/588 +- [ ] 🔥 GPTQ support: https://github.com/go-skynet/LocalAI/issues/796 ## News diff --git a/api/backend/options.go b/api/backend/options.go index 71e18855..2623a780 100644 --- a/api/backend/options.go +++ b/api/backend/options.go @@ -15,19 +15,21 @@ func gRPCModelOpts(c config.Config) *pb.ModelOptions { b = c.Batch } return &pb.ModelOptions{ - ContextSize: int32(c.ContextSize), - Seed: int32(c.Seed), - NBatch: int32(b), - F16Memory: c.F16, - MLock: c.MMlock, - NUMA: c.NUMA, - Embeddings: c.Embeddings, - LowVRAM: c.LowVRAM, - NGPULayers: int32(c.NGPULayers), - MMap: c.MMap, - MainGPU: c.MainGPU, - Threads: int32(c.Threads), - TensorSplit: c.TensorSplit, + ContextSize: int32(c.ContextSize), + Seed: int32(c.Seed), + NBatch: int32(b), + F16Memory: c.F16, + MLock: c.MMlock, + RopeFreqBase: c.RopeFreqBase, + RopeFreqScale: c.RopeFreqScale, + NUMA: c.NUMA, + Embeddings: c.Embeddings, + LowVRAM: c.LowVRAM, + NGPULayers: int32(c.NGPULayers), + MMap: c.MMap, + MainGPU: c.MainGPU, + Threads: int32(c.Threads), + TensorSplit: c.TensorSplit, } } diff --git a/extra/grpc/huggingface/backend_pb2.py b/extra/grpc/huggingface/backend_pb2.py index 0b52b8f8..d9f4baa5 100644 --- a/extra/grpc/huggingface/backend_pb2.py +++ b/extra/grpc/huggingface/backend_pb2.py @@ -13,7 +13,7 @@ _sym_db = _symbol_database.Default() -DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x0f\n\rHealthMessage\"\x86\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18 \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xac\x02\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\x9e\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\"6\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t2\xeb\x03\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3') +DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x0f\n\rHealthMessage\"\x86\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18 \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xd9\x02\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\x9e\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\"6\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t2\xeb\x03\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3') _globals = globals() _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals) @@ -29,21 +29,21 @@ if _descriptor._USE_C_DESCRIPTORS == False: _globals['_REPLY']._serialized_start=820 _globals['_REPLY']._serialized_end=844 _globals['_MODELOPTIONS']._serialized_start=847 - _globals['_MODELOPTIONS']._serialized_end=1147 - _globals['_RESULT']._serialized_start=1149 - _globals['_RESULT']._serialized_end=1191 - _globals['_EMBEDDINGRESULT']._serialized_start=1193 - _globals['_EMBEDDINGRESULT']._serialized_end=1230 - _globals['_TRANSCRIPTREQUEST']._serialized_start=1232 - _globals['_TRANSCRIPTREQUEST']._serialized_end=1299 - _globals['_TRANSCRIPTRESULT']._serialized_start=1301 - _globals['_TRANSCRIPTRESULT']._serialized_end=1379 - _globals['_TRANSCRIPTSEGMENT']._serialized_start=1381 - _globals['_TRANSCRIPTSEGMENT']._serialized_end=1470 - _globals['_GENERATEIMAGEREQUEST']._serialized_start=1473 - _globals['_GENERATEIMAGEREQUEST']._serialized_end=1631 - _globals['_TTSREQUEST']._serialized_start=1633 - _globals['_TTSREQUEST']._serialized_end=1687 - _globals['_BACKEND']._serialized_start=1690 - _globals['_BACKEND']._serialized_end=2181 + _globals['_MODELOPTIONS']._serialized_end=1192 + _globals['_RESULT']._serialized_start=1194 + _globals['_RESULT']._serialized_end=1236 + _globals['_EMBEDDINGRESULT']._serialized_start=1238 + _globals['_EMBEDDINGRESULT']._serialized_end=1275 + _globals['_TRANSCRIPTREQUEST']._serialized_start=1277 + _globals['_TRANSCRIPTREQUEST']._serialized_end=1344 + _globals['_TRANSCRIPTRESULT']._serialized_start=1346 + _globals['_TRANSCRIPTRESULT']._serialized_end=1424 + _globals['_TRANSCRIPTSEGMENT']._serialized_start=1426 + _globals['_TRANSCRIPTSEGMENT']._serialized_end=1515 + _globals['_GENERATEIMAGEREQUEST']._serialized_start=1518 + _globals['_GENERATEIMAGEREQUEST']._serialized_end=1676 + _globals['_TTSREQUEST']._serialized_start=1678 + _globals['_TTSREQUEST']._serialized_end=1732 + _globals['_BACKEND']._serialized_start=1735 + _globals['_BACKEND']._serialized_end=2226 # @@protoc_insertion_point(module_scope) diff --git a/pkg/grpc/llm/llama/llama.go b/pkg/grpc/llm/llama/llama.go index 7ebe7816..7d867813 100644 --- a/pkg/grpc/llm/llama/llama.go +++ b/pkg/grpc/llm/llama/llama.go @@ -17,7 +17,10 @@ type LLM struct { } func (llm *LLM) Load(opts *pb.ModelOptions) error { - llamaOpts := []llama.ModelOption{} + llamaOpts := []llama.ModelOption{ + llama.WithRopeFreqBase(opts.RopeFreqBase), + llama.WithRopeFreqScale(opts.RopeFreqScale), + } if opts.ContextSize != 0 { llamaOpts = append(llamaOpts, llama.SetContext(int(opts.ContextSize))) @@ -56,15 +59,15 @@ func (llm *LLM) Load(opts *pb.ModelOptions) error { func buildPredictOptions(opts *pb.PredictOptions) []llama.PredictOption { predictOptions := []llama.PredictOption{ - llama.SetTemperature(float64(opts.Temperature)), - llama.SetTopP(float64(opts.TopP)), + llama.SetTemperature(opts.Temperature), + llama.SetTopP(opts.TopP), llama.SetTopK(int(opts.TopK)), llama.SetTokens(int(opts.Tokens)), llama.SetThreads(int(opts.Threads)), llama.WithGrammar(opts.Grammar), - llama.SetRopeFreqBase(float64(opts.RopeFreqBase)), - llama.SetRopeFreqScale(float64(opts.RopeFreqScale)), - llama.SetNegativePromptScale(float64(opts.NegativePromptScale)), + llama.SetRopeFreqBase(opts.RopeFreqBase), + llama.SetRopeFreqScale(opts.RopeFreqScale), + llama.SetNegativePromptScale(opts.NegativePromptScale), llama.SetNegativePrompt(opts.NegativePrompt), } @@ -86,11 +89,11 @@ func buildPredictOptions(opts *pb.PredictOptions) []llama.PredictOption { } if opts.MirostatETA != 0 { - predictOptions = append(predictOptions, llama.SetMirostatETA(float64(opts.MirostatETA))) + predictOptions = append(predictOptions, llama.SetMirostatETA(opts.MirostatETA)) } if opts.MirostatTAU != 0 { - predictOptions = append(predictOptions, llama.SetMirostatTAU(float64(opts.MirostatTAU))) + predictOptions = append(predictOptions, llama.SetMirostatTAU(opts.MirostatTAU)) } if opts.Debug { @@ -100,7 +103,7 @@ func buildPredictOptions(opts *pb.PredictOptions) []llama.PredictOption { predictOptions = append(predictOptions, llama.SetStopWords(opts.StopPrompts...)) if opts.PresencePenalty != 0 { - predictOptions = append(predictOptions, llama.SetPenalty(float64(opts.PresencePenalty))) + predictOptions = append(predictOptions, llama.SetPenalty(opts.PresencePenalty)) } if opts.NKeep != 0 { @@ -125,13 +128,13 @@ func buildPredictOptions(opts *pb.PredictOptions) []llama.PredictOption { //predictOptions = append(predictOptions, llama.SetLogitBias(c.Seed)) - predictOptions = append(predictOptions, llama.SetFrequencyPenalty(float64(opts.FrequencyPenalty))) + predictOptions = append(predictOptions, llama.SetFrequencyPenalty(opts.FrequencyPenalty)) predictOptions = append(predictOptions, llama.SetMlock(opts.MLock)) predictOptions = append(predictOptions, llama.SetMemoryMap(opts.MMap)) predictOptions = append(predictOptions, llama.SetPredictionMainGPU(opts.MainGPU)) predictOptions = append(predictOptions, llama.SetPredictionTensorSplit(opts.TensorSplit)) - predictOptions = append(predictOptions, llama.SetTailFreeSamplingZ(float64(opts.TailFreeSamplingZ))) - predictOptions = append(predictOptions, llama.SetTypicalP(float64(opts.TypicalP))) + predictOptions = append(predictOptions, llama.SetTailFreeSamplingZ(opts.TailFreeSamplingZ)) + predictOptions = append(predictOptions, llama.SetTypicalP(opts.TypicalP)) return predictOptions } diff --git a/pkg/grpc/proto/backend.pb.go b/pkg/grpc/proto/backend.pb.go index db993bd2..844d10bc 100644 --- a/pkg/grpc/proto/backend.pb.go +++ b/pkg/grpc/proto/backend.pb.go @@ -463,22 +463,24 @@ type ModelOptions struct { sizeCache protoimpl.SizeCache unknownFields protoimpl.UnknownFields - Model string `protobuf:"bytes,1,opt,name=Model,proto3" json:"Model,omitempty"` - ContextSize int32 `protobuf:"varint,2,opt,name=ContextSize,proto3" json:"ContextSize,omitempty"` - Seed int32 `protobuf:"varint,3,opt,name=Seed,proto3" json:"Seed,omitempty"` - NBatch int32 `protobuf:"varint,4,opt,name=NBatch,proto3" json:"NBatch,omitempty"` - F16Memory bool `protobuf:"varint,5,opt,name=F16Memory,proto3" json:"F16Memory,omitempty"` - MLock bool `protobuf:"varint,6,opt,name=MLock,proto3" json:"MLock,omitempty"` - MMap bool `protobuf:"varint,7,opt,name=MMap,proto3" json:"MMap,omitempty"` - VocabOnly bool `protobuf:"varint,8,opt,name=VocabOnly,proto3" json:"VocabOnly,omitempty"` - LowVRAM bool `protobuf:"varint,9,opt,name=LowVRAM,proto3" json:"LowVRAM,omitempty"` - Embeddings bool `protobuf:"varint,10,opt,name=Embeddings,proto3" json:"Embeddings,omitempty"` - NUMA bool `protobuf:"varint,11,opt,name=NUMA,proto3" json:"NUMA,omitempty"` - NGPULayers int32 `protobuf:"varint,12,opt,name=NGPULayers,proto3" json:"NGPULayers,omitempty"` - MainGPU string `protobuf:"bytes,13,opt,name=MainGPU,proto3" json:"MainGPU,omitempty"` - TensorSplit string `protobuf:"bytes,14,opt,name=TensorSplit,proto3" json:"TensorSplit,omitempty"` - Threads int32 `protobuf:"varint,15,opt,name=Threads,proto3" json:"Threads,omitempty"` - LibrarySearchPath string `protobuf:"bytes,16,opt,name=LibrarySearchPath,proto3" json:"LibrarySearchPath,omitempty"` + Model string `protobuf:"bytes,1,opt,name=Model,proto3" json:"Model,omitempty"` + ContextSize int32 `protobuf:"varint,2,opt,name=ContextSize,proto3" json:"ContextSize,omitempty"` + Seed int32 `protobuf:"varint,3,opt,name=Seed,proto3" json:"Seed,omitempty"` + NBatch int32 `protobuf:"varint,4,opt,name=NBatch,proto3" json:"NBatch,omitempty"` + F16Memory bool `protobuf:"varint,5,opt,name=F16Memory,proto3" json:"F16Memory,omitempty"` + MLock bool `protobuf:"varint,6,opt,name=MLock,proto3" json:"MLock,omitempty"` + MMap bool `protobuf:"varint,7,opt,name=MMap,proto3" json:"MMap,omitempty"` + VocabOnly bool `protobuf:"varint,8,opt,name=VocabOnly,proto3" json:"VocabOnly,omitempty"` + LowVRAM bool `protobuf:"varint,9,opt,name=LowVRAM,proto3" json:"LowVRAM,omitempty"` + Embeddings bool `protobuf:"varint,10,opt,name=Embeddings,proto3" json:"Embeddings,omitempty"` + NUMA bool `protobuf:"varint,11,opt,name=NUMA,proto3" json:"NUMA,omitempty"` + NGPULayers int32 `protobuf:"varint,12,opt,name=NGPULayers,proto3" json:"NGPULayers,omitempty"` + MainGPU string `protobuf:"bytes,13,opt,name=MainGPU,proto3" json:"MainGPU,omitempty"` + TensorSplit string `protobuf:"bytes,14,opt,name=TensorSplit,proto3" json:"TensorSplit,omitempty"` + Threads int32 `protobuf:"varint,15,opt,name=Threads,proto3" json:"Threads,omitempty"` + LibrarySearchPath string `protobuf:"bytes,16,opt,name=LibrarySearchPath,proto3" json:"LibrarySearchPath,omitempty"` + RopeFreqBase float32 `protobuf:"fixed32,17,opt,name=RopeFreqBase,proto3" json:"RopeFreqBase,omitempty"` + RopeFreqScale float32 `protobuf:"fixed32,18,opt,name=RopeFreqScale,proto3" json:"RopeFreqScale,omitempty"` } func (x *ModelOptions) Reset() { @@ -625,6 +627,20 @@ func (x *ModelOptions) GetLibrarySearchPath() string { return "" } +func (x *ModelOptions) GetRopeFreqBase() float32 { + if x != nil { + return x.RopeFreqBase + } + return 0 +} + +func (x *ModelOptions) GetRopeFreqScale() float32 { + if x != nil { + return x.RopeFreqScale + } + return 0 +} + type Result struct { state protoimpl.MessageState sizeCache protoimpl.SizeCache @@ -1175,7 +1191,7 @@ var file_pkg_grpc_proto_backend_proto_rawDesc = []byte{ 0x0e, 0x4e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x22, 0x21, 0x0a, 0x05, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x12, 0x18, 0x0a, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28, 0x0c, 0x52, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61, - 0x67, 0x65, 0x22, 0xca, 0x03, 0x0a, 0x0c, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69, + 0x67, 0x65, 0x22, 0x94, 0x04, 0x0a, 0x0c, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x12, 0x14, 0x0a, 0x05, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x05, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x20, 0x0a, 0x0b, 0x43, 0x6f, 0x6e, 0x74, 0x65, 0x78, 0x74, 0x53, 0x69, 0x7a, 0x65, 0x18, 0x02, 0x20, 0x01, 0x28, 0x05, 0x52, 0x0b, @@ -1203,90 +1219,95 @@ var file_pkg_grpc_proto_backend_proto_rawDesc = []byte{ 0x64, 0x73, 0x18, 0x0f, 0x20, 0x01, 0x28, 0x05, 0x52, 0x07, 0x54, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x12, 0x2c, 0x0a, 0x11, 0x4c, 0x69, 0x62, 0x72, 0x61, 0x72, 0x79, 0x53, 0x65, 0x61, 0x72, 0x63, 0x68, 0x50, 0x61, 0x74, 0x68, 0x18, 0x10, 0x20, 0x01, 0x28, 0x09, 0x52, 0x11, 0x4c, 0x69, - 0x62, 0x72, 0x61, 0x72, 0x79, 0x53, 0x65, 0x61, 0x72, 0x63, 0x68, 0x50, 0x61, 0x74, 0x68, 0x22, - 0x3c, 0x0a, 0x06, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x12, 0x18, 0x0a, 0x07, 0x6d, 0x65, 0x73, - 0x73, 0x61, 0x67, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x07, 0x6d, 0x65, 0x73, 0x73, - 0x61, 0x67, 0x65, 0x12, 0x18, 0x0a, 0x07, 0x73, 0x75, 0x63, 0x63, 0x65, 0x73, 0x73, 0x18, 0x02, - 0x20, 0x01, 0x28, 0x08, 0x52, 0x07, 0x73, 0x75, 0x63, 0x63, 0x65, 0x73, 0x73, 0x22, 0x31, 0x0a, - 0x0f, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, - 0x12, 0x1e, 0x0a, 0x0a, 0x65, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73, 0x18, 0x01, - 0x20, 0x03, 0x28, 0x02, 0x52, 0x0a, 0x65, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73, - 0x22, 0x5b, 0x0a, 0x11, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, - 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x02, 0x20, 0x01, - 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, 0x74, 0x12, 0x1a, 0x0a, 0x08, 0x6c, 0x61, 0x6e, 0x67, 0x75, - 0x61, 0x67, 0x65, 0x18, 0x03, 0x20, 0x01, 0x28, 0x09, 0x52, 0x08, 0x6c, 0x61, 0x6e, 0x67, 0x75, - 0x61, 0x67, 0x65, 0x12, 0x18, 0x0a, 0x07, 0x74, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x18, 0x04, - 0x20, 0x01, 0x28, 0x0d, 0x52, 0x07, 0x74, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x22, 0x5e, 0x0a, - 0x10, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x73, 0x75, 0x6c, - 0x74, 0x12, 0x36, 0x0a, 0x08, 0x73, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x73, 0x18, 0x01, 0x20, - 0x03, 0x28, 0x0b, 0x32, 0x1a, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, - 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x52, - 0x08, 0x73, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x73, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78, - 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x22, 0x77, 0x0a, - 0x11, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x53, 0x65, 0x67, 0x6d, 0x65, - 0x6e, 0x74, 0x12, 0x0e, 0x0a, 0x02, 0x69, 0x64, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05, 0x52, 0x02, - 0x69, 0x64, 0x12, 0x14, 0x0a, 0x05, 0x73, 0x74, 0x61, 0x72, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, - 0x03, 0x52, 0x05, 0x73, 0x74, 0x61, 0x72, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x65, 0x6e, 0x64, 0x18, - 0x03, 0x20, 0x01, 0x28, 0x03, 0x52, 0x03, 0x65, 0x6e, 0x64, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, - 0x78, 0x74, 0x18, 0x04, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x12, 0x16, - 0x0a, 0x06, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73, 0x18, 0x05, 0x20, 0x03, 0x28, 0x05, 0x52, 0x06, - 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73, 0x22, 0xe4, 0x01, 0x0a, 0x14, 0x47, 0x65, 0x6e, 0x65, 0x72, - 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, - 0x16, 0x0a, 0x06, 0x68, 0x65, 0x69, 0x67, 0x68, 0x74, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05, 0x52, - 0x06, 0x68, 0x65, 0x69, 0x67, 0x68, 0x74, 0x12, 0x14, 0x0a, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, - 0x18, 0x02, 0x20, 0x01, 0x28, 0x05, 0x52, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, 0x12, 0x12, 0x0a, - 0x04, 0x6d, 0x6f, 0x64, 0x65, 0x18, 0x03, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x6d, 0x6f, 0x64, - 0x65, 0x12, 0x12, 0x0a, 0x04, 0x73, 0x74, 0x65, 0x70, 0x18, 0x04, 0x20, 0x01, 0x28, 0x05, 0x52, - 0x04, 0x73, 0x74, 0x65, 0x70, 0x12, 0x12, 0x0a, 0x04, 0x73, 0x65, 0x65, 0x64, 0x18, 0x05, 0x20, - 0x01, 0x28, 0x05, 0x52, 0x04, 0x73, 0x65, 0x65, 0x64, 0x12, 0x27, 0x0a, 0x0f, 0x70, 0x6f, 0x73, - 0x69, 0x74, 0x69, 0x76, 0x65, 0x5f, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x18, 0x06, 0x20, 0x01, - 0x28, 0x09, 0x52, 0x0e, 0x70, 0x6f, 0x73, 0x69, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, - 0x70, 0x74, 0x12, 0x27, 0x0a, 0x0f, 0x6e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x5f, 0x70, - 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x18, 0x07, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0e, 0x6e, 0x65, 0x67, - 0x61, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x64, - 0x73, 0x74, 0x18, 0x08, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, 0x74, 0x22, 0x48, 0x0a, - 0x0a, 0x54, 0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x12, 0x0a, 0x04, 0x74, - 0x65, 0x78, 0x74, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x12, - 0x14, 0x0a, 0x05, 0x6d, 0x6f, 0x64, 0x65, 0x6c, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x05, - 0x6d, 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x03, 0x20, 0x01, - 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, 0x74, 0x32, 0xeb, 0x03, 0x0a, 0x07, 0x42, 0x61, 0x63, 0x6b, - 0x65, 0x6e, 0x64, 0x12, 0x32, 0x0a, 0x06, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x12, 0x16, 0x2e, - 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65, - 0x73, 0x73, 0x61, 0x67, 0x65, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, - 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x34, 0x0a, 0x07, 0x50, 0x72, 0x65, 0x64, 0x69, - 0x63, 0x74, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, - 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62, 0x61, - 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x35, 0x0a, - 0x09, 0x4c, 0x6f, 0x61, 0x64, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x15, 0x2e, 0x62, 0x61, 0x63, - 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, - 0x73, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, - 0x6c, 0x74, 0x22, 0x00, 0x12, 0x3c, 0x0a, 0x0d, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x53, - 0x74, 0x72, 0x65, 0x61, 0x6d, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, - 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, - 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, - 0x30, 0x01, 0x12, 0x40, 0x0a, 0x09, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x12, - 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, - 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, - 0x6e, 0x64, 0x2e, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, - 0x6c, 0x74, 0x22, 0x00, 0x12, 0x41, 0x0a, 0x0d, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, - 0x49, 0x6d, 0x61, 0x67, 0x65, 0x12, 0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, - 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x52, 0x65, 0x71, - 0x75, 0x65, 0x73, 0x74, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, - 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x4d, 0x0a, 0x12, 0x41, 0x75, 0x64, 0x69, 0x6f, - 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x12, 0x1a, 0x2e, + 0x62, 0x72, 0x61, 0x72, 0x79, 0x53, 0x65, 0x61, 0x72, 0x63, 0x68, 0x50, 0x61, 0x74, 0x68, 0x12, + 0x22, 0x0a, 0x0c, 0x52, 0x6f, 0x70, 0x65, 0x46, 0x72, 0x65, 0x71, 0x42, 0x61, 0x73, 0x65, 0x18, + 0x11, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0c, 0x52, 0x6f, 0x70, 0x65, 0x46, 0x72, 0x65, 0x71, 0x42, + 0x61, 0x73, 0x65, 0x12, 0x24, 0x0a, 0x0d, 0x52, 0x6f, 0x70, 0x65, 0x46, 0x72, 0x65, 0x71, 0x53, + 0x63, 0x61, 0x6c, 0x65, 0x18, 0x12, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0d, 0x52, 0x6f, 0x70, 0x65, + 0x46, 0x72, 0x65, 0x71, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x22, 0x3c, 0x0a, 0x06, 0x52, 0x65, 0x73, + 0x75, 0x6c, 0x74, 0x12, 0x18, 0x0a, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x18, 0x01, + 0x20, 0x01, 0x28, 0x09, 0x52, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x12, 0x18, 0x0a, + 0x07, 0x73, 0x75, 0x63, 0x63, 0x65, 0x73, 0x73, 0x18, 0x02, 0x20, 0x01, 0x28, 0x08, 0x52, 0x07, + 0x73, 0x75, 0x63, 0x63, 0x65, 0x73, 0x73, 0x22, 0x31, 0x0a, 0x0f, 0x45, 0x6d, 0x62, 0x65, 0x64, + 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x12, 0x1e, 0x0a, 0x0a, 0x65, 0x6d, + 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73, 0x18, 0x01, 0x20, 0x03, 0x28, 0x02, 0x52, 0x0a, + 0x65, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73, 0x22, 0x5b, 0x0a, 0x11, 0x54, 0x72, + 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, + 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, + 0x74, 0x12, 0x1a, 0x0a, 0x08, 0x6c, 0x61, 0x6e, 0x67, 0x75, 0x61, 0x67, 0x65, 0x18, 0x03, 0x20, + 0x01, 0x28, 0x09, 0x52, 0x08, 0x6c, 0x61, 0x6e, 0x67, 0x75, 0x61, 0x67, 0x65, 0x12, 0x18, 0x0a, + 0x07, 0x74, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x18, 0x04, 0x20, 0x01, 0x28, 0x0d, 0x52, 0x07, + 0x74, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x22, 0x5e, 0x0a, 0x10, 0x54, 0x72, 0x61, 0x6e, 0x73, + 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x12, 0x36, 0x0a, 0x08, 0x73, + 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x73, 0x18, 0x01, 0x20, 0x03, 0x28, 0x0b, 0x32, 0x1a, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, - 0x70, 0x74, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x19, 0x2e, 0x62, 0x61, 0x63, 0x6b, - 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, - 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x2d, 0x0a, 0x03, 0x54, 0x54, 0x53, 0x12, 0x13, 0x2e, - 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65, - 0x73, 0x74, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, - 0x75, 0x6c, 0x74, 0x22, 0x00, 0x42, 0x5a, 0x0a, 0x19, 0x69, 0x6f, 0x2e, 0x73, 0x6b, 0x79, 0x6e, - 0x65, 0x74, 0x2e, 0x6c, 0x6f, 0x63, 0x61, 0x6c, 0x61, 0x69, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, - 0x6e, 0x64, 0x42, 0x0e, 0x4c, 0x6f, 0x63, 0x61, 0x6c, 0x41, 0x49, 0x42, 0x61, 0x63, 0x6b, 0x65, - 0x6e, 0x64, 0x50, 0x01, 0x5a, 0x2b, 0x67, 0x69, 0x74, 0x68, 0x75, 0x62, 0x2e, 0x63, 0x6f, 0x6d, - 0x2f, 0x67, 0x6f, 0x2d, 0x73, 0x6b, 0x79, 0x6e, 0x65, 0x74, 0x2f, 0x4c, 0x6f, 0x63, 0x61, 0x6c, - 0x41, 0x49, 0x2f, 0x70, 0x6b, 0x67, 0x2f, 0x67, 0x72, 0x70, 0x63, 0x2f, 0x70, 0x72, 0x6f, 0x74, - 0x6f, 0x62, 0x06, 0x70, 0x72, 0x6f, 0x74, 0x6f, 0x33, + 0x70, 0x74, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x52, 0x08, 0x73, 0x65, 0x67, 0x6d, 0x65, + 0x6e, 0x74, 0x73, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, + 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x22, 0x77, 0x0a, 0x11, 0x54, 0x72, 0x61, 0x6e, 0x73, + 0x63, 0x72, 0x69, 0x70, 0x74, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x12, 0x0e, 0x0a, 0x02, + 0x69, 0x64, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05, 0x52, 0x02, 0x69, 0x64, 0x12, 0x14, 0x0a, 0x05, + 0x73, 0x74, 0x61, 0x72, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, 0x03, 0x52, 0x05, 0x73, 0x74, 0x61, + 0x72, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x65, 0x6e, 0x64, 0x18, 0x03, 0x20, 0x01, 0x28, 0x03, 0x52, + 0x03, 0x65, 0x6e, 0x64, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78, 0x74, 0x18, 0x04, 0x20, 0x01, + 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x12, 0x16, 0x0a, 0x06, 0x74, 0x6f, 0x6b, 0x65, + 0x6e, 0x73, 0x18, 0x05, 0x20, 0x03, 0x28, 0x05, 0x52, 0x06, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73, + 0x22, 0xe4, 0x01, 0x0a, 0x14, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61, + 0x67, 0x65, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x16, 0x0a, 0x06, 0x68, 0x65, 0x69, + 0x67, 0x68, 0x74, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05, 0x52, 0x06, 0x68, 0x65, 0x69, 0x67, 0x68, + 0x74, 0x12, 0x14, 0x0a, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, 0x18, 0x02, 0x20, 0x01, 0x28, 0x05, + 0x52, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, 0x12, 0x12, 0x0a, 0x04, 0x6d, 0x6f, 0x64, 0x65, 0x18, + 0x03, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x6d, 0x6f, 0x64, 0x65, 0x12, 0x12, 0x0a, 0x04, 0x73, + 0x74, 0x65, 0x70, 0x18, 0x04, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x73, 0x74, 0x65, 0x70, 0x12, + 0x12, 0x0a, 0x04, 0x73, 0x65, 0x65, 0x64, 0x18, 0x05, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x73, + 0x65, 0x65, 0x64, 0x12, 0x27, 0x0a, 0x0f, 0x70, 0x6f, 0x73, 0x69, 0x74, 0x69, 0x76, 0x65, 0x5f, + 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x18, 0x06, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0e, 0x70, 0x6f, + 0x73, 0x69, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x12, 0x27, 0x0a, 0x0f, + 0x6e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x5f, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x18, + 0x07, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0e, 0x6e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x50, + 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x08, 0x20, 0x01, + 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, 0x74, 0x22, 0x48, 0x0a, 0x0a, 0x54, 0x54, 0x53, 0x52, 0x65, + 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78, 0x74, 0x18, 0x01, 0x20, + 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x12, 0x14, 0x0a, 0x05, 0x6d, 0x6f, 0x64, + 0x65, 0x6c, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x05, 0x6d, 0x6f, 0x64, 0x65, 0x6c, 0x12, + 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x03, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, + 0x74, 0x32, 0xeb, 0x03, 0x0a, 0x07, 0x42, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x12, 0x32, 0x0a, + 0x06, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x12, 0x16, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, + 0x64, 0x2e, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x1a, + 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, + 0x00, 0x12, 0x34, 0x0a, 0x07, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x12, 0x17, 0x2e, 0x62, + 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, + 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, + 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x35, 0x0a, 0x09, 0x4c, 0x6f, 0x61, 0x64, 0x4d, + 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x15, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x4d, + 0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0f, 0x2e, 0x62, 0x61, + 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x3c, + 0x0a, 0x0d, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x53, 0x74, 0x72, 0x65, 0x61, 0x6d, 0x12, + 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, + 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, + 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x30, 0x01, 0x12, 0x40, 0x0a, 0x09, + 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, + 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, + 0x6e, 0x73, 0x1a, 0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x45, 0x6d, 0x62, + 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x41, + 0x0a, 0x0d, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x12, + 0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, + 0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x0f, + 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, + 0x00, 0x12, 0x4d, 0x0a, 0x12, 0x41, 0x75, 0x64, 0x69, 0x6f, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, + 0x72, 0x69, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x12, 0x1a, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, + 0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x71, 0x75, + 0x65, 0x73, 0x74, 0x1a, 0x19, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, + 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, + 0x12, 0x2d, 0x0a, 0x03, 0x54, 0x54, 0x53, 0x12, 0x13, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, + 0x64, 0x2e, 0x54, 0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x0f, 0x2e, 0x62, + 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x42, + 0x5a, 0x0a, 0x19, 0x69, 0x6f, 0x2e, 0x73, 0x6b, 0x79, 0x6e, 0x65, 0x74, 0x2e, 0x6c, 0x6f, 0x63, + 0x61, 0x6c, 0x61, 0x69, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x42, 0x0e, 0x4c, 0x6f, + 0x63, 0x61, 0x6c, 0x41, 0x49, 0x42, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x50, 0x01, 0x5a, 0x2b, + 0x67, 0x69, 0x74, 0x68, 0x75, 0x62, 0x2e, 0x63, 0x6f, 0x6d, 0x2f, 0x67, 0x6f, 0x2d, 0x73, 0x6b, + 0x79, 0x6e, 0x65, 0x74, 0x2f, 0x4c, 0x6f, 0x63, 0x61, 0x6c, 0x41, 0x49, 0x2f, 0x70, 0x6b, 0x67, + 0x2f, 0x67, 0x72, 0x70, 0x63, 0x2f, 0x70, 0x72, 0x6f, 0x74, 0x6f, 0x62, 0x06, 0x70, 0x72, 0x6f, + 0x74, 0x6f, 0x33, } var ( diff --git a/pkg/grpc/proto/backend.proto b/pkg/grpc/proto/backend.proto index cdad7891..386662ab 100644 --- a/pkg/grpc/proto/backend.proto +++ b/pkg/grpc/proto/backend.proto @@ -85,6 +85,8 @@ message ModelOptions { string TensorSplit = 14; int32 Threads = 15; string LibrarySearchPath = 16; + float RopeFreqBase = 17; + float RopeFreqScale = 18; } message Result {