hf_text-generation-inference/backends/trtllm/lib/backend.cpp

135 lines
5.0 KiB
C++
Raw Normal View History

#include <fmt/std.h>
#include <nvml.h>
2024-07-08 16:08:49 -06:00
#include <spdlog/spdlog.h>
#include "backend.h"
2024-07-08 16:08:49 -06:00
void huggingface::tgi::backends::InitializeBackend() {
SPDLOG_INFO("Initializing Backend...");
nvmlInit_v2();
2024-07-08 16:08:49 -06:00
initTrtLlmPlugins();
}
[[nodiscard]]
2024-07-08 16:08:49 -06:00
tle::ExecutorConfig huggingface::tgi::backends::GetExecutorConfig(const json &config, const std::string &workerPath) {
tle::ExecutorConfig execConfig(1);
// Get the compute capabilities of the current hardware
nvmlDevice_t device;
int32_t cudaComputeCapabilitiesMajor = 0, cudaComputeCapabilitiesMinor = 0;
if(nvmlDeviceGetHandleByIndex_v2(0, &device) == NVML_SUCCESS) {
SPDLOG_DEBUG("Successfully acquired nvmlDevice_t = 0");
if(nvmlDeviceGetCudaComputeCapability(device, &cudaComputeCapabilitiesMajor, &cudaComputeCapabilitiesMinor) == NVML_SUCCESS) {
SPDLOG_INFO(FMT_STRING("Detected sm_{:d}{:d} compute capabilities"), cudaComputeCapabilitiesMajor, cudaComputeCapabilitiesMinor);
}
}
2024-07-08 16:08:49 -06:00
// Single engine (TP = PP = 1) -> using leader mode (no MPI involved)
2024-07-08 16:08:49 -06:00
if(config["/pretrained_config/mapping/world_size"_json_pointer].get<uint8_t>() == 1){
SPDLOG_INFO("Detected single engine deployment, using leader mode");
execConfig.setParallelConfig(tle::ParallelConfig(
tle::CommunicationType::kMPI,
tle::CommunicationMode::kLEADER,
std::nullopt,
std::nullopt,
std::nullopt
));
} else { // Multiple engines -> using orchestrator mode (MPI involved)
2024-07-08 16:08:49 -06:00
SPDLOG_INFO("Detected sharded engine deployment, using orchestrator mode");
execConfig.setParallelConfig(tle::ParallelConfig(
tle::CommunicationType::kMPI,
tle::CommunicationMode::kORCHESTRATOR,
std::nullopt,
std::nullopt,
tle::OrchestratorConfig(true, workerPath)
));
}
// Define some configuration variables
execConfig.setKvCacheConfig(tle::KvCacheConfig(true));
execConfig.setEnableChunkedContext(cudaComputeCapabilitiesMajor >= 8);
return execConfig;
}
huggingface::tgi::backends::TensorRtLlmBackend::TensorRtLlmBackend(
2024-07-08 16:08:49 -06:00
const std::filesystem::path &enginesFolder,
const std::filesystem::path &executorWorker
):
2024-07-08 16:08:49 -06:00
config(json::parse(std::ifstream(enginesFolder / "config.json"))),
executor(
enginesFolder,
tensorrt_llm::executor::ModelType::kDECODER_ONLY,
GetExecutorConfig(config, executorWorker.string()
))
{
2024-07-08 16:08:49 -06:00
SPDLOG_INFO(FMT_STRING("Engine (version={})"), config["/version"_json_pointer].get_ref<const std::string&>());
}
2024-07-03 02:27:53 -06:00
[[nodiscard("Returned request id needs to be provided back to gather generated tokens")]]
2024-07-03 02:27:53 -06:00
tle::IdType huggingface::tgi::backends::TensorRtLlmBackend::Submit(
2024-07-08 16:08:49 -06:00
const std::vector<tle::TokenIdType> &tokens,
2024-07-03 02:27:53 -06:00
const int32_t maxNewTokens,
2024-07-08 16:08:49 -06:00
const int32_t topK,
2024-07-03 02:27:53 -06:00
const float_t topP,
const float_t temperature,
const int32_t minLength,
2024-07-08 16:08:49 -06:00
std::optional<float_t> repetitionPenalty,
std::optional<float_t> frequencyPenalty,
std::optional<uint32_t> seed,
std::optional<uint32_t> nTopTokens
2024-07-03 02:27:53 -06:00
) {
SPDLOG_DEBUG(
FMT_STRING("Submitting inference over {:d} tokens to the executor ({:d} already in-flight)"),
2024-07-08 16:08:49 -06:00
tokens.size(),
executor.getLatestIterationStats().back().numActiveRequests
);
const auto sampling = tle::SamplingConfig{
1,
topK,
topP,
std::nullopt,
std::nullopt,
std::nullopt,
seed,
temperature,
minLength,
std::nullopt,
repetitionPenalty,
std::nullopt,
frequencyPenalty,
};
const auto output = tle::OutputConfig{false, false, nTopTokens.value_or(1) > 1};
const auto request = tle::Request{tokens, maxNewTokens, true, sampling, output};
return executor.enqueueRequest(request);
2024-07-03 02:27:53 -06:00
}
2024-07-08 16:08:49 -06:00
size_t huggingface::tgi::backends::TensorRtLlmBackend::Stream(const tle::IdType reqId, const std::function<TokenStreamingCallback>& cb) {
bool isFinal = false;
size_t generatedTokens = 0;
do {
const auto responses = executor.awaitResponses(reqId);
for (const auto &response: responses){
if(response.hasError()) {
SPDLOG_WARN("Caught error during generation: {}", response.getErrorMsg());
isFinal = true;
} else {
const auto generation = response.getResult();
const auto token = generation.outputTokenIds[0][0];
// Update the end of stream detection and overall number of generated tokens
isFinal = generation.isFinal;
++generatedTokens;
// Send the token back through the callback function for further processing
cb(token);
}
}
} while(!isFinal);
// Return the number of generated tokens
return generatedTokens;
2024-07-08 16:08:49 -06:00
}