""" # Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License" # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """ import os import time from typing import Dict, List, Optional import numpy as np import paddle import paddle.nn as nn from paddleformers.utils.log import logger from fastdeploy.config import FDConfig from fastdeploy.engine.request import Request # from fastdeploy.spec_decode import MTPProposer, NgramProposer from fastdeploy.model_executor.forward_meta import HPUForwardMeta from fastdeploy.model_executor.guided_decoding import get_guided_backend from fastdeploy.model_executor.guided_decoding.base_guided_decoding import ( LogitsProcessorBase, ) from fastdeploy.model_executor.layers.attention import get_attention_backend from fastdeploy.model_executor.layers.attention.base_attention_backend import ( AttentionBackend, ) from fastdeploy.model_executor.layers.rotary_embedding import get_rope from fastdeploy.model_executor.layers.sample.meta_data import SamplingMetadata from fastdeploy.model_executor.layers.sample.sampler import Sampler, SpeculativeSampler from fastdeploy.model_executor.model_loader import get_model_loader from fastdeploy.model_executor.ops.intel_hpu import ( recover_block, save_output, step_paddle, update_inputs_v3, ) from fastdeploy.utils import get_logger from fastdeploy.worker.model_runner_base import ModelRunnerBase from fastdeploy.worker.output import ModelOutputData, ModelRunnerOutput hpu_model_runner_profile_logger = get_logger("hpu_model_runner_profile", "hpu_model_runner_profile.log") def post_process_hpu(sampled_token_ids: paddle.Tensor, model_output: ModelOutputData, is_warmuping: bool) -> None: """Post-processing steps after completing a single token generation.""" start_time = time.time() not_need_stop_hpu = model_output.not_need_stop.to(sampled_token_ids.place) is_block_step_hpu = model_output.is_block_step.to(sampled_token_ids.place) update_inputs_v3( model_output.stop_flags, model_output.step_idx, not_need_stop_hpu, model_output.seq_lens_this_time, model_output.seq_lens_encoder, model_output.seq_lens_decoder, model_output.max_dec_len, model_output.input_ids, model_output.stop_nums, sampled_token_ids, is_block_step_hpu, model_output.eos_token_id, model_output.next_tokens, ) model_output.not_need_stop[:] = not_need_stop_hpu.cpu() model_output.is_block_step[:] = is_block_step_hpu.cpu() end_time = time.time() execution_time = (end_time - start_time) * 1000 hpu_model_runner_profile_logger.info(f"post_process_hpu::update_inputs_v3 execution time(ms): {execution_time}") if is_warmuping: return start_time = time.time() save_output( sampled_token_ids, model_output.not_need_stop, model_output.mp_rank, ) end_time = time.time() execution_time = (end_time - start_time) * 1000 hpu_model_runner_profile_logger.info(f"post_process_hpu::save_output execution time(ms): {execution_time}") def recover_block_hpu( recover_block_list, # cpu recover_len, # cpu stop_flags, # hpu seq_lens_this_time, # hpu ori_seq_lens_encoder, # cpu seq_lens_encoder, # hpu block_tables, # cpu free_list, # cpu free_list_len, # cpu input_ids, # hpu pre_ids, # hpu step_idx, # hpu encoder_block_lens, # cpu used_list_len, # cpu next_tokens, # hpu first_token_ids, ): # hpu for bid in range(recover_len.item()): recover_id = recover_block_list[bid].item() ori_seq_len_encoder = ori_seq_lens_encoder[recover_id].item() step_idx_now = step_idx[recover_id].item() seq_len = ori_seq_len_encoder + step_idx_now encoder_block_len = encoder_block_lens[recover_id].item() decoder_used_len = used_list_len[recover_id].item() seq_lens_this_time[recover_id] = seq_len seq_lens_encoder[recover_id] = seq_len stop_flags[recover_id] = False ori_free_list_len = free_list_len[0] free_list_len[0] -= decoder_used_len for i in range(decoder_used_len): block_tables[recover_id, encoder_block_len + i] = free_list[ori_free_list_len - i - 1] recover_block(input_ids, first_token_ids, pre_ids, next_tokens, recover_id, ori_seq_len_encoder, step_idx_now) def step_intel_hpu(share_inputs: Dict[str, paddle.Tensor], block_size: int, max_model_len: int) -> None: """ step cuda """ step_paddle( share_inputs["stop_flags"], share_inputs["seq_lens_this_time"], share_inputs["seq_lens_encoder"], share_inputs["seq_lens_decoder"], share_inputs["block_tables"], share_inputs["encoder_block_lens"], share_inputs["is_block_step"], share_inputs["step_block_list"], share_inputs["step_lens"], share_inputs["recover_block_list"], share_inputs["recover_lens"], share_inputs["need_block_list"], share_inputs["need_block_len"], share_inputs["used_list_len"], share_inputs["free_list"], share_inputs["free_list_len"], share_inputs["first_token_ids"], block_size, max_model_len, ) if share_inputs["recover_lens"].item() > 0: recover_block_hpu( share_inputs["recover_block_list"], share_inputs["recover_lens"], share_inputs["stop_flags"], share_inputs["seq_lens_this_time"], share_inputs["ori_seq_lens_encoder"], share_inputs["seq_lens_encoder"], share_inputs["block_tables"], share_inputs["free_list"], share_inputs["free_list_len"], share_inputs["input_ids"], share_inputs["pre_ids"], share_inputs["step_idx"], share_inputs["encoder_block_lens"], share_inputs["used_list_len"], share_inputs["next_tokens"], share_inputs["first_token_ids"], ) share_inputs["recover_lens"] = paddle.full([1], 0, dtype="int32").cpu() # TODO: replace rebuild_padding_v3 in CustomDevice if we adopt this version pp optimization def rebuild_padding_v3_1( tmp_out, batch_ids, total_batch, seq_lens_encoder, is_prompt=None, ): dim_emb = tmp_out.shape[-1] output_data = paddle.zeros((total_batch, dim_emb)) if is_prompt is True: # context tmp_out = tmp_out.reshape([total_batch, -1, dim_emb]) for i in range(batch_ids.shape[0]): seq_len = seq_lens_encoder[batch_ids[i]].item() output_data[i] = tmp_out[i, seq_len - 1] elif is_prompt is False: output_data[0 : batch_ids.shape[0], :] = tmp_out[: batch_ids.shape[0], :] return output_data from fastdeploy.model_executor.layers.linear import QKVParallelLinear, RowParallelLinear from fastdeploy.model_executor.ops.intel_hpu import fused_mlp def fused_attention_forward( self, src: paddle.Tensor = None, qkv_proj: QKVParallelLinear = None, o_proj: RowParallelLinear = None, forward_meta: HPUForwardMeta = None, ): """ The forward function of attention layer. args: src: the hidden states tensor residual_input: the residual tensor forward_meta: the forward meta data """ return forward_meta.attn_backend.forward( src, qkv_proj, o_proj, self, forward_meta, ) def fused_self_atten_forward( self, forward_meta: HPUForwardMeta, hidden_states: paddle.Tensor, ): """ """ atten_out = self.attn( src=hidden_states, qkv_proj=self.qkv_proj, o_proj=self.o_proj, forward_meta=forward_meta, ) return atten_out def fused_mlp_forward(self, x): """ """ out = fused_mlp( x, self.up_gate_proj.weight, None, self.down_proj.weight, ) # all_reduce if self.nranks > 1: from fastdeploy.distributed.communication import ( tensor_model_parallel_all_reduce_custom, ) tensor_model_parallel_all_reduce_custom(out) return out import types from fastdeploy.model_executor.layers.attention.attention import Attention from fastdeploy.model_executor.models.ernie4_5_moe import ( Ernie4_5_Attention, Ernie4_5_MLP, ) from fastdeploy.model_executor.models.qwen2 import Qwen2Attention, Qwen2MLP def convert_model(model): """ """ for name, module in model.named_children(): if len(list(module.named_children())) > 0: # print(f"********** model {model.__class__.__name__} has submodule: name={name}, module={module.__class__.__name__}") if isinstance(module, Ernie4_5_Attention): module.forward = types.MethodType(fused_self_atten_forward, module) if isinstance(module, Qwen2Attention): module.forward = types.MethodType(fused_self_atten_forward, module) if isinstance(module, Ernie4_5_MLP): module.forward = types.MethodType(fused_mlp_forward, module) if isinstance(module, Qwen2MLP): module.forward = types.MethodType(fused_mlp_forward, module) convert_model(module) else: # print(f"*********[ Leaf node] Loading submodule: name={name} -- module: {module.__class__.__name__}") if isinstance(module, Attention): module.forward = types.MethodType(fused_attention_forward, module) return model class HPUModelRunner(ModelRunnerBase): """ """ def __init__( self, fd_config: FDConfig, device: str, # logic device device_id: int, # physical device id rank: int, local_rank: int, ): super().__init__(fd_config=fd_config, device=device) self.rank = rank self.local_rank = local_rank self.device_id = device_id self.speculative_method = self.fd_config.speculative_config.method self.speculative_decoding = self.speculative_method is not None self.guided_backend = None if self.fd_config.parallel_config.guided_decoding_backend != "off": self.guided_backend = get_guided_backend(fd_config=self.fd_config) # Sampler if not self.speculative_decoding: self.sampler = Sampler() else: self.sampler = SpeculativeSampler(fd_config) # Lazy initialize kv cache after model loading # self.kv_caches: list[paddle.Tensor] = [] # Cuda Graph self.use_cudagraph = self.graph_opt_config.use_cudagraph self.cudagraph_capture_sizes = list(reversed(self.graph_opt_config.cudagraph_capture_sizes)) self.cudagraph_num_of_warmups = self.graph_opt_config.cudagraph_num_of_warmups self.input_ids = paddle.zeros(self.scheduler_config.max_num_seqs, dtype="int32") # Initialize share inputs self._init_share_inputs(self.scheduler_config.max_num_seqs) self.infer_seed_increment = paddle.full( shape=[self.scheduler_config.max_num_seqs, 1], fill_value=4, dtype="int64" ).cpu() self.restore_chunked_prefill_request = dict() # Initialize attention Backend # Note(gonshaotian): Currently, all attention layers share one attention backend instance. # In the future, we will expand it as a list. self.attn_backends: list[AttentionBackend] = [] # self.attn_metadatas: list[AttentionMetadata] = [] self.initialize_attn_backend() # Forward meta store the global meta information of the forward self.forward_meta: HPUForwardMeta = None self.is_warmuping = False self.is_hpu_perf_breakdown_sync_mode = int(os.environ.get("HPU_PERF_BREAKDOWN_SYNC_MODE", 1)) == 1 # Postprocess Env params os.environ["INFERENCE_MSG_QUEUE_ID"] = str( self.local_rank + int(self.parallel_config.engine_worker_queue_port) ) if int(os.environ.get("HABANA_PROFILE", 0)) == 1: step_start = int(os.environ.get("PROFILE_START", 0)) step_end = int(os.environ.get("PROFILE_END", 4)) import paddle.profiler as profiler self.prof = profiler.Profiler( targets=[profiler.ProfilerTarget.CPU, profiler.ProfilerTarget.CUSTOM_DEVICE], scheduler=(step_start, step_end), on_trace_ready=profiler.export_chrome_tracing("./profile"), ) self.prof.start() def exist_prefill(self): """ check whether prefill stage finished """ if int(paddle.max(self.share_inputs["seq_lens_encoder"])) != 0: return 1 else: return 0 def init_speculative_proposer(self): """ Init speculative proposer """ # if self.speculative_method == "ngram": # self.proposer = NgramProposer(self.fd_config) # elif self.speculative_method == "mtp": # self.proposer = MTPProposer(self.fd_config, self.get_model(), # self.local_rank, self.device_id, # self.share_inputs) # else: # self.proposer = None pass def _init_logits_processor(self, request): """ init logits processor for guided decoding """ assert self.guided_backend is not None, ( "guided_backend is None, use " "--guided-decoding-backend to specify the backend at server startup." ) if request.guided_json is not None: schemata_key = ("json", request.guided_json) elif request.guided_regex is not None: schemata_key = ("regex", request.guided_regex) elif request.guided_grammar is not None: schemata_key = ("grammar", request.guided_grammar) elif request.structural_tag is not None: schemata_key = ("structural_tag", request.structural_tag) return self.guided_backend.get_logits_processor(schemata_key=schemata_key), schemata_key def insert_prefill_inputs(self, req_dicts: List[Request], num_running_requests: int = None): """ Process inputs for prefill tasks and insert it to share_inputs buffer req_dict: A list of Request dict num_running_requests: batch_size """ # NOTE(luotingdan): Lazy initialize kv cache if "caches" not in self.share_inputs: self.initialize_kv_cache() # NOTE(luotingdan): Set environment variable of prefill node if req_dicts[-1].disaggregate_info is not None and req_dicts[-1].disaggregate_info["role"] == "prefill": os.environ["PREFILL_NODE_ONE_STEP_STOP"] = "1" req_len = len(req_dicts) for i in range(req_len): request = req_dicts[i] idx = request.idx length = len(request.prompt_token_ids) prefill_tokens = [] if ( request.guided_json is not None or request.guided_regex is not None or request.structural_tag is not None or request.guided_grammar is not None ): logits_info, schemata_key = self._init_logits_processor(request) request.logits_processor, request.logits_cached = logits_info request.schemata_key = schemata_key # Is Decode Node if req_dicts[i].disaggregate_info is not None and req_dicts[i].disaggregate_info["role"] == "decode": prefill_tokens.append(request.prompt_token_ids[0]) self.share_inputs["pre_ids"][idx : idx + 1] = request.prompt_token_ids[-1] self.share_inputs["input_ids"][idx : idx + 1, 0] = request.prompt_token_ids[0] self.share_inputs["seq_lens_encoder"][idx : idx + 1] = 0 self.share_inputs["seq_lens_decoder"][idx : idx + 1] = length self.share_inputs["seq_lens_this_time"][idx : idx + 1] = 1 self.share_inputs["step_seq_lens_encoder"][idx : idx + 1] = 0 self.share_inputs["step_seq_lens_decoder"][idx : idx + 1] = length self.share_inputs["step_idx"][idx : idx + 1] = 1 if self.speculative_decoding: num_prefill_send_token = self.speculative_config.num_speculative_tokens + 1 self.share_inputs["draft_tokens"][idx : idx + 1, 0:num_prefill_send_token] = paddle.to_tensor( request.draft_token_ids[0:num_prefill_send_token], dtype="int64" ) self.share_inputs["seq_lens_this_time"][idx : idx + 1] = num_prefill_send_token else: self.share_inputs["pre_ids"][idx : idx + 1] = -1 self.share_inputs["step_idx"][idx : idx + 1] = 0 self.share_inputs["input_ids"][idx : idx + 1, :length] = np.array(request.prompt_token_ids) # Use chunked prefill if self.cache_config.enable_chunked_prefill: request.set("chunk_idx", 1) logger.info(f"prefill_chunk_info: {request.prefill_chunk_info}") token_chunk_size = request.prefill_chunk_info[0] self.share_inputs["seq_lens_this_time"][idx : idx + 1] = token_chunk_size self.share_inputs["input_ids"][idx, :token_chunk_size] = np.array( request.prompt_token_ids[:token_chunk_size] ) self.share_inputs["step_seq_lens_encoder"][idx : idx + 1] = token_chunk_size self.share_inputs["seq_lens_encoder"][idx : idx + 1] = token_chunk_size self.share_inputs["seq_lens_decoder"][idx : idx + 1] = request.get("seq_lens_decoder", 0) self.share_inputs["step_seq_lens_decoder"][idx : idx + 1] = request.get("seq_lens_decoder", 0) else: self.share_inputs["seq_lens_decoder"][idx : idx + 1] = request.get("seq_lens_decoder", 0) self.share_inputs["step_seq_lens_decoder"][idx : idx + 1] = request.get("seq_lens_decoder", 0) self.share_inputs["seq_lens_this_time"][idx : idx + 1] = length self.share_inputs["step_seq_lens_encoder"][idx : idx + 1] = length self.share_inputs["seq_lens_encoder"][idx : idx + 1] = length if len(request.eos_token_ids) < self.model_config.eos_tokens_lens: request.eos_token_ids.append(request.eos_token_ids[0]) self.share_inputs["eos_token_id"][:] = np.array(request.eos_token_ids, dtype="int64").reshape(-1, 1) self.share_inputs["top_p"][idx : idx + 1] = request.get("top_p", 0.7) self.share_inputs["temperature"][idx : idx + 1] = request.get("temperature", 0.95) self.share_inputs["penalty_score"][idx : idx + 1] = request.get("repetition_penalty", 1.0) self.share_inputs["frequency_score"][idx : idx + 1] = request.get("frequency_penalty", 0.0) self.share_inputs["presence_score"][idx : idx + 1] = request.get("presence_penalty", 0.0) self.share_inputs["min_dec_len"][idx : idx + 1] = request.get("min_tokens", 1) self.share_inputs["max_dec_len"][idx : idx + 1] = request.get( "max_tokens", self.model_config.max_model_len ) self.share_inputs["stop_flags"][idx : idx + 1] = False self.share_inputs["first_token_ids"][idx : idx + 1] = self.share_inputs["input_ids"][idx : idx + 1, :1] self.share_inputs["ori_seq_lens_encoder"][idx : idx + 1] = length if request.get("seed") is not None: self.share_inputs["infer_seed"][idx : idx + 1] = request.get("seed") encoder_block_num = len(request.get("block_tables")) self.share_inputs["encoder_block_lens"][idx : idx + 1] = encoder_block_num self.share_inputs["block_tables"][idx : idx + 1, :] = -1 self.share_inputs["block_tables"][idx : idx + 1, :encoder_block_num] = np.array( request.block_tables, dtype="int32" ) if request.get("stop_token_ids") is not None and request.get("stop_seqs_len") is not None: stop_seqs_num = len(request.get("stop_seqs_len")) for i in range(stop_seqs_num, self.model_config.max_stop_seqs_num): request.stop_seqs_len.append(0) self.share_inputs["stop_seqs_len"][:] = np.array(request.stop_seqs_len, dtype="int32") self.share_inputs["stop_seqs"][:stop_seqs_num, : len(request.get("stop_token_ids")[0])] = np.array( request.get("stop_token_ids"), dtype="int64" ) self.sampler.apply_logits_processor(idx, request.get("logits_processor"), prefill_tokens) self.share_inputs["not_need_stop"][0] = True if self.speculative_method in ["mtp"]: self.proposer.insert_prefill_inputs(req_dicts, num_running_requests) def _dummy_prefill_inputs(self, num_tokens: int, batch_size: int, expected_decode_len: int): """Set dummy prefill inputs to share_inputs""" # NOTE(gongshaotian): The maximum decoding length is equal to the expected decoded tokens plus the eos token max_dec_len = expected_decode_len + 1 full_length = min(num_tokens // batch_size, self.parallel_config.max_model_len - max_dec_len) input_length = int(full_length * self.cache_config.kv_cache_ratio) block_num = ( input_length + self.cache_config.block_size - 1 ) // self.cache_config.block_size + self.cache_config.enc_dec_block_num for i in range(batch_size): idx = i self.share_inputs["input_ids"][idx : idx + 1, :input_length] = np.array([5] * input_length) self.share_inputs["eos_token_id"][:] = np.array([2], dtype="int64").reshape(-1, 1) self.share_inputs["seq_lens_this_time"][idx : idx + 1] = input_length self.share_inputs["step_seq_lens_encoder"][idx : idx + 1] = input_length self.share_inputs["seq_lens_encoder"][idx : idx + 1] = input_length self.share_inputs["seq_lens_decoder"][idx : idx + 1] = 0 self.share_inputs["step_idx"][idx : idx + 1] = 0 self.share_inputs["max_dec_len"][idx : idx + 1] = max_dec_len self.share_inputs["stop_flags"][idx : idx + 1] = False self.share_inputs["first_token_ids"][idx : idx + 1] = self.share_inputs["input_ids"][idx : idx + 1, :1] self.share_inputs["ori_seq_lens_encoder"][idx : idx + 1] = input_length self.share_inputs["encoder_block_lens"][idx : idx + 1] = block_num self.share_inputs["block_tables"][idx : idx + 1, :block_num] = np.arange( idx * block_num, (idx + 1) * block_num, 1 ) def _init_share_inputs(self, max_num_seqs: int): """Initialize all share buffers for model inputs. Note: In the future, we may abandon share buffers. """ self.MAX_INFER_SEED = 9223372036854775806 self.share_inputs = {} self.share_inputs["pre_ids"] = paddle.full( [max_num_seqs, self.parallel_config.max_model_len], -1, dtype="int64" ) self.share_inputs["input_ids"] = paddle.full( [max_num_seqs, self.parallel_config.max_model_len], self.model_config.pad_token_id, dtype="int64" ) self.share_inputs["eos_token_id"] = paddle.full([self.model_config.eos_tokens_lens, 1], 0, dtype="int64") self.share_inputs["top_p"] = paddle.full([max_num_seqs, 1], self.model_config.top_p, dtype="float32") self.share_inputs["temperature"] = paddle.full( [max_num_seqs, 1], self.model_config.temperature, dtype="float32" ) self.share_inputs["penalty_score"] = paddle.full( [max_num_seqs, 1], self.model_config.penalty_score, dtype="float32" ) self.share_inputs["frequency_score"] = paddle.full( [max_num_seqs, 1], self.model_config.frequency_score, dtype="float32" ) self.share_inputs["presence_score"] = paddle.full( [max_num_seqs, 1], self.model_config.presence_score, dtype="float32" ) self.share_inputs["min_dec_len"] = paddle.full([max_num_seqs, 1], self.model_config.min_length, dtype="int64") self.share_inputs["max_dec_len"] = paddle.full( [max_num_seqs, 1], self.model_config.max_model_len, dtype="int64" ) self.share_inputs["min_length"] = paddle.full([max_num_seqs, 1], self.model_config.min_length, dtype="int64") self.share_inputs["max_length"] = paddle.full( [max_num_seqs, 1], self.model_config.max_model_len, dtype="int64" ) self.share_inputs["seq_lens_this_time"] = paddle.full(max_num_seqs, 0, dtype="int32") self.share_inputs["seq_lens_encoder"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") self.share_inputs["seq_lens_decoder"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") self.share_inputs["step_seq_lens_encoder"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") self.share_inputs["step_seq_lens_decoder"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") self.share_inputs["step_idx"] = paddle.full([max_num_seqs, 1], 0, dtype="int64") self.share_inputs["not_need_stop"] = paddle.full( [1], False, dtype="bool" ).cpu() # TODO(gongshaotian): move to pinnd memory self.share_inputs["stop_flags"] = paddle.full([max_num_seqs, 1], True, dtype="bool") self.share_inputs["stop_nums"] = paddle.full([1], max_num_seqs, dtype="int64") self.share_inputs["bad_tokens"] = paddle.full([1], -1, dtype="int64") self.share_inputs["next_tokens"] = paddle.full([max_num_seqs, 1], -1, dtype="int64") self.share_inputs["is_block_step"] = paddle.full([max_num_seqs], False, dtype="bool").cpu() self.share_inputs["encoder_block_lens"] = paddle.full([max_num_seqs], 0, dtype="int32").cpu() self.share_inputs["step_block_list"] = paddle.full([max_num_seqs], -1, dtype="int32").cpu() self.share_inputs["step_lens"] = paddle.full([1], 0, dtype="int32").cpu() self.share_inputs["recover_block_list"] = paddle.full([max_num_seqs], -1, dtype="int32").cpu() self.share_inputs["recover_lens"] = paddle.full([1], 0, dtype="int32").cpu() self.share_inputs["need_block_list"] = paddle.full([max_num_seqs], -1, dtype="int32").cpu() self.share_inputs["need_block_len"] = paddle.full([1], 0, dtype="int32").cpu() self.share_inputs["used_list_len"] = paddle.full([max_num_seqs], 0, dtype="int32").cpu() self.share_inputs["infer_seed"] = paddle.full([max_num_seqs, 1], 0, dtype="int64").cpu() self.share_inputs["first_token_ids"] = paddle.full([max_num_seqs, 1], -1, dtype="int64") self.share_inputs["ori_seq_lens_encoder"] = paddle.full([max_num_seqs, 1], 0, dtype="int32").cpu() self.share_inputs["system_lens"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") self.share_inputs["system_ids"] = paddle.full([max_num_seqs, 1], -1, dtype="int32") self.share_inputs["ids_remove_padding"] = paddle.full( [max_num_seqs * self.parallel_config.max_model_len], 0, dtype="int64" ) self.share_inputs["cum_offsets"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") self.share_inputs["padding_offset"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") self.share_inputs["cu_seqlens_q"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") self.share_inputs["cu_seqlens_k"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") # AttentionBackend buffers self.share_inputs["decoder_batch_ids"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") self.share_inputs["decoder_tile_ids_per_batch"] = paddle.full([max_num_seqs, 1], 0, dtype="int32") # Initialize rotary position embedding tmp_position_ids = paddle.arange(self.parallel_config.max_model_len).reshape((1, -1)) # TODO(gongshaotian): move to models self.share_inputs["rope_emb"] = get_rope( rotary_dim=self.model_config.head_dim, position_ids=tmp_position_ids, base=self.model_config.rope_theta, model_config=self.model_config, ) # Set block tables pre_max_block_num = ( self.parallel_config.max_model_len + self.cache_config.block_size - 1 ) // self.cache_config.block_size + self.cache_config.enc_dec_block_num self.share_inputs["block_tables"] = paddle.full([max_num_seqs, pre_max_block_num], -1, dtype="int32").cpu() # Initialize free list free_list = list( range( self.parallel_config.total_block_num - 2, int(self.parallel_config.total_block_num * self.cache_config.kv_cache_ratio) - 1, -1, ) ) self.free_list_len = len(free_list) self.share_inputs["free_list"] = paddle.to_tensor(free_list, dtype="int32").cpu() self.share_inputs["free_list_len"] = paddle.full([1], self.free_list_len, dtype="int32").cpu() # Initialize stop seqs self.share_inputs["stop_seqs_len"] = paddle.full([self.model_config.max_stop_seqs_num], 0, dtype="int32") self.share_inputs["stop_seqs"] = paddle.full( [self.model_config.max_stop_seqs_num, self.model_config.stop_seqs_max_len], -1, dtype="int32" ) if self.speculative_decoding: max_draft_token_num = self.speculative_config.num_speculative_tokens self.share_inputs["input_ids_cpu"] = paddle.full( shape=[max_num_seqs, self.parallel_config.max_model_len], fill_value=1, dtype="int64" ).cpu() self.share_inputs["accept_tokens"] = paddle.full( shape=[max_num_seqs, max_draft_token_num + 1], fill_value=0, dtype="int64" ) self.share_inputs["accept_num"] = paddle.full(shape=[max_num_seqs], fill_value=0, dtype="int32") self.share_inputs["draft_tokens"] = paddle.full( shape=[max_num_seqs, max_draft_token_num + 1], fill_value=0, dtype="int64" ) self.share_inputs["actual_draft_token_num"] = paddle.full( shape=[max_num_seqs], fill_value=max_draft_token_num, dtype="int32" ) self.share_inputs["output_cum_offsets"] = paddle.full(shape=[max_num_seqs, 1], fill_value=0, dtype="int32") self.share_inputs["output_padding_offset"] = paddle.full( shape=[max_num_seqs * (max_draft_token_num + 1)], fill_value=0, dtype="int32" ) def _prepare_inputs(self) -> None: """prepare the model inputs""" from fastdeploy.model_executor.ops.intel_hpu import prepare_block_metadata ( ids_remove_padding, rotary_embs, block_groups, block_list, block_indices, block_offsets, block_mapping, attention_mask, batch_ids, total_batch, is_prompt, ) = prepare_block_metadata( self.share_inputs["input_ids"], self.share_inputs["rope_emb"], self.share_inputs["block_tables"], self.share_inputs["seq_lens_encoder"], self.share_inputs["seq_lens_decoder"], self.cache_config.block_size, self.parallel_config.dtype, ) is_prompt = is_prompt.item() == 1 if is_prompt.item() > 0 else None if is_prompt is True: attention_mask = None # cum_offsets = None self.share_inputs["ids_remove_padding"] = ids_remove_padding self.share_inputs["rotary_embs"] = rotary_embs self.share_inputs["block_groups"] = block_groups self.share_inputs["block_list"] = block_list self.share_inputs["block_indices"] = block_indices self.share_inputs["block_offsets"] = block_offsets self.share_inputs["block_mapping"] = block_mapping self.share_inputs["block_bias"] = attention_mask self.share_inputs["block_size"] = self.cache_config.block_size self.share_inputs["batch_ids"] = batch_ids self.share_inputs["total_batch"] = total_batch.item() self.share_inputs["is_prompt"] = is_prompt self.initialize_forward_meta() def _prepare_sampler_inputs(self, sampled_ids) -> None: if self.forward_meta.total_batch == self.share_inputs["temperature"].shape[0]: self.sampling_metadata = SamplingMetadata( temperature=self.share_inputs["temperature"], top_p=self.share_inputs["top_p"], step_idx=self.share_inputs["step_idx"], prompt_ids=self.share_inputs["input_ids"], pre_token_ids=self.share_inputs["pre_ids"], stop_flags=self.share_inputs["stop_flags"], seq_lens_encoder=self.share_inputs["seq_lens_encoder"], seq_lens_decoder=self.share_inputs["seq_lens_decoder"], frequency_penalties=self.share_inputs["frequency_score"], presence_penalties=self.share_inputs["presence_score"], repetition_penalties=self.share_inputs["penalty_score"], min_dec_lens=self.share_inputs["min_dec_len"], bad_words_token_ids=self.share_inputs["bad_tokens"], eos_token_ids=self.share_inputs["eos_token_id"], ) else: from fastdeploy.model_executor.ops.intel_hpu import fused_index_select ( temperature, top_p, step_idx, prompt_token_ids, pre_token_ids, stop_flags, seq_lens_encoder, seq_lens_decoder, frequency_penalties, presence_penalties, repetition_penalties, min_dec_lens, ) = fused_index_select( self.share_inputs["temperature"], self.share_inputs["top_p"], self.share_inputs["step_idx"], self.share_inputs["input_ids"], self.share_inputs["pre_ids"], self.share_inputs["stop_flags"], self.share_inputs["seq_lens_encoder"], self.share_inputs["seq_lens_decoder"], self.share_inputs["frequency_score"], self.share_inputs["presence_score"], self.share_inputs["penalty_score"], self.share_inputs["min_dec_len"], sampled_ids, self.forward_meta.total_batch, ) self.sampling_metadata = SamplingMetadata( temperature=temperature, top_p=top_p, step_idx=step_idx, prompt_ids=prompt_token_ids, pre_token_ids=pre_token_ids, stop_flags=stop_flags, seq_lens_encoder=seq_lens_encoder, seq_lens_decoder=seq_lens_decoder, frequency_penalties=frequency_penalties, presence_penalties=presence_penalties, repetition_penalties=repetition_penalties, min_dec_lens=min_dec_lens, bad_words_token_ids=self.share_inputs["bad_tokens"], eos_token_ids=self.share_inputs["eos_token_id"], ) def load_model(self) -> None: """load or download model""" logger.info(f"Starting to load model {self.model_config.architectures[0]}") time_before_load = time.perf_counter() # 1. Load original model model_loader = get_model_loader(load_config=self.fd_config.load_config) self.model = model_loader.load_model(fd_config=self.fd_config) # 1.1 Load RL dynamic model if self.fd_config.load_config.dynamic_load_weight: from fastdeploy.rl.dynamic_weight_manager import DynamicWeightManager self.dynamic_weight_manager = DynamicWeightManager(self.fd_config, self.model) # 2. Load lora model # 3. Load drafter model(for speculative decoding) # 4. Convert model to HPU format self.model = convert_model(self.model) time_after_load = time.perf_counter() logger.info(f"Model loading took {time_after_load - time_before_load} seconds") # 4. Init proposer for speculative method self.init_speculative_proposer() def get_model(self) -> nn.Layer: """get current model""" return self.model def initialize_forward_meta(self): """ Initialize forward meta and attention meta data """ # Initialize forward meta self.forward_meta = HPUForwardMeta.init_forward_meta(self.share_inputs, self.attn_backends[0]) # Initialzie attention meta data for attn_backend in self.attn_backends: attn_backend.init_attention_metadata(self.forward_meta) def clear_cache(self): """Clear cached data from shared inputs and forward metadata.""" self.share_inputs.pop("caches", None) if self.forward_meta is not None: self.forward_meta.clear_caches() def initialize_kv_cache(self) -> None: """ Initialize kv cache """ cache_kvs = {} max_block_num = self.num_gpu_blocks kv_cache_shape = self.attn_backends[0].get_kv_cache_shape(max_num_blocks=max_block_num) for i in range(self.model_config.num_hidden_layers): cache_type = self.parallel_config.dtype cache_kvs["key_caches_{}".format(i)] = paddle.full( shape=kv_cache_shape, fill_value=0, dtype=cache_type, ) cache_kvs["value_caches_{}".format(i)] = paddle.full( shape=kv_cache_shape, fill_value=0, dtype=cache_type, ) self.share_inputs["caches"] = list(cache_kvs.values()) for value in cache_kvs.values(): del value def initialize_attn_backend(self) -> None: """ Initialize attention backends and forward metadata """ assert len(self.attn_backends) == 0 # TODO(gongshaotian): Get rank from config num_heads = self.model_config.num_attention_heads // self.parallel_config.tensor_parallel_size self.model_config.kv_num_heads = ( int(self.model_config.num_key_value_heads) // self.parallel_config.tensor_parallel_size ) head_dim = self.model_config.head_dim # Get the attention backend attn_cls = get_attention_backend() attn_backend = attn_cls( self.fd_config, kv_num_heads=self.model_config.kv_num_heads, num_heads=num_heads, head_dim=head_dim ) if attn_backend is None: raise NotImplementedError( "Attention backend which you specified is not supported, please set FD_ATTENTION_BACKEND correctly." ) self.attn_backends.append(attn_backend) def _dummy_run( self, num_tokens: paddle.Tensor, batch_size: paddle.Tensor, expected_decode_len: int = 1, in_capturing: bool = False, ) -> paddle.Tensor: """ Use dummy inputs to run before formal execution. Args: num_tokens: expected_decode_len: Expected number of tokens generated """ self._dummy_prefill_inputs( num_tokens=num_tokens, batch_size=batch_size, expected_decode_len=expected_decode_len ) if self.speculative_method in ["mtp"]: raise NotImplementedError("speculative sampling is not supported on Intel HPU.") while True: # 1. Compute real num_tokens self._prepare_inputs() # 2. Initialize attention backend and forward meta data model_output = self.model(self.share_inputs["ids_remove_padding"], self.forward_meta) hiddden_states = rebuild_padding_v3_1( model_output, self.forward_meta.batch_ids, self.forward_meta.total_batch, self.forward_meta.seq_lens_encoder, self.forward_meta.is_prompt, ) # 5. Execute spec decode logits = self.model.compute_logits(hiddden_states) self._prepare_sampler_inputs(self.forward_meta.batch_ids) sampled_token_ids = self.sampler( logits, self.sampling_metadata, self.forward_meta.batch_ids, self.forward_meta.seq_lens_encoder.shape[0], self.rank, self.local_rank, ) if self.parallel_config.tensor_parallel_size > 1: dtype = sampled_token_ids.dtype sampled_token_ids = sampled_token_ids.to("float32") paddle.distributed.broadcast(sampled_token_ids, 0) sampled_token_ids = sampled_token_ids.to(dtype) # 6. post process model_output_data = ModelOutputData( next_tokens=self.share_inputs["next_tokens"], stop_flags=self.share_inputs["stop_flags"], step_idx=self.share_inputs["step_idx"], max_dec_len=self.share_inputs["max_dec_len"], pre_ids=self.share_inputs["pre_ids"], seq_lens_this_time=self.share_inputs["seq_lens_this_time"], eos_token_id=self.share_inputs["eos_token_id"], not_need_stop=self.share_inputs["not_need_stop"], input_ids=self.share_inputs["input_ids"], stop_nums=self.share_inputs["stop_nums"], seq_lens_encoder=self.share_inputs["seq_lens_encoder"], seq_lens_decoder=self.share_inputs["seq_lens_decoder"], is_block_step=self.share_inputs["is_block_step"], full_hidden_states=model_output, msg_queue_id=self.parallel_config.msg_queue_id, mp_rank=self.local_rank, use_ep=self.parallel_config.use_ep, draft_tokens=self.share_inputs["draft_tokens"] if self.speculative_decoding else None, actual_draft_token_num=( self.share_inputs["actual_draft_token_num"] if self.speculative_decoding else None ), accept_tokens=self.share_inputs["accept_tokens"] if self.speculative_decoding else None, accept_num=self.share_inputs["accept_num"] if self.speculative_decoding else None, ) post_process_hpu( sampled_token_ids=sampled_token_ids, model_output=model_output_data, is_warmuping=self.is_warmuping ) # 7. Updata 'infer_seed' and step_cuda() self.share_inputs["infer_seed"].add_(self.infer_seed_increment) self.share_inputs["infer_seed"][:] %= self.MAX_INFER_SEED step_intel_hpu(self.share_inputs, self.cache_config.block_size, self.parallel_config.max_model_len) if int((self.share_inputs["seq_lens_this_time"] > 0).sum()) == 0: break def _update_chunked_prefill(self, tasks): """ 更新chunked prefill相关参数 """ if not self.cache_config.enable_chunked_prefill: return for task in tasks: if task.get("prefill_chunk_info", None) is None: continue if task.chunk_idx > len(task.prefill_chunk_info): continue self.restore_chunked_prefill_request[task.request_id] = task for id, task in list(self.restore_chunked_prefill_request.items()): idx = task.idx logger.debug(f"{task.request_id} chunked prefill {task.chunk_idx}/{len(task.prefill_chunk_info)}") start_idx = sum(task.prefill_chunk_info[: task.chunk_idx]) if task.chunk_idx == len(task.prefill_chunk_info): self.share_inputs["seq_lens_this_time"][idx : idx + 1] = 1 self.share_inputs["seq_lens_encoder"][idx : idx + 1] = 0 self.share_inputs["step_idx"][idx : idx + 1] = 1 self.share_inputs["seq_lens_decoder"][idx : idx + 1] = start_idx + task.get("seq_lens_decoder", 0) del self.restore_chunked_prefill_request[task.request_id] else: token_chunk_size = task.prefill_chunk_info[task.chunk_idx] self.share_inputs["seq_lens_this_time"][idx : idx + 1] = token_chunk_size self.share_inputs["input_ids"][idx, :token_chunk_size] = np.array( task.prompt_token_ids[start_idx : start_idx + token_chunk_size] ) self.share_inputs["seq_lens_encoder"][idx : idx + 1] = token_chunk_size self.share_inputs["step_idx"][idx : idx + 1] = 0 self.share_inputs["seq_lens_decoder"][idx : idx + 1] = start_idx + task.get("seq_lens_decoder", 0) if self.speculative_decoding and self.proposer.is_chunk_prefill_enabled(): self.proposer.update_task_chunk_prefill(task) task.chunk_idx += 1 def _dummy_sampler_run(self) -> paddle.Tensor: """ """ pass def update_warmup_inputs(self, requests, is_decode=False): for i in range(len(requests)): request = requests[i] idx = request["idx"] length = len(request["input_ids"]) self.share_inputs["input_ids"][idx : idx + 1, :length] = np.array(request["input_ids"]) if is_decode: self.share_inputs["seq_lens_encoder"][idx : idx + 1] = 0 self.share_inputs["seq_lens_decoder"][idx : idx + 1] = length self.share_inputs["seq_lens_this_time"][idx : idx + 1] = 1 self.share_inputs["step_seq_lens_encoder"][idx : idx + 1] = 0 self.share_inputs["step_seq_lens_decoder"][idx : idx + 1] = length self.share_inputs["step_idx"][idx : idx + 1] = 1 else: self.share_inputs["seq_lens_encoder"][idx : idx + 1] = length self.share_inputs["seq_lens_decoder"][idx : idx + 1] = 0 self.share_inputs["seq_lens_this_time"][idx : idx + 1] = length self.share_inputs["step_seq_lens_encoder"][idx : idx + 1] = length self.share_inputs["step_seq_lens_decoder"][idx : idx + 1] = 0 self.share_inputs["step_idx"][idx : idx + 1] = 0 if len(request["eos_token_ids"]) < self.model_config.eos_tokens_lens: request["eos_token_ids"].append(request["eos_token_ids"][0]) self.share_inputs["eos_token_id"][:] = np.array(request["eos_token_ids"], dtype="int64").reshape(-1, 1) self.share_inputs["top_p"][idx : idx + 1] = request.get("top_p", 0.7) self.share_inputs["temperature"][idx : idx + 1] = request.get("temperature", 0.95) self.share_inputs["penalty_score"][idx : idx + 1] = request.get("repetition_penalty", 1.0) self.share_inputs["frequency_score"][idx : idx + 1] = request.get("frequency_penalty", 0.0) self.share_inputs["presence_score"][idx : idx + 1] = request.get("presence_penalty", 0.0) self.share_inputs["min_dec_len"][idx : idx + 1] = request.get("min_tokens", 1) self.share_inputs["max_dec_len"][idx : idx + 1] = request.get("max_tokens", 1) self.share_inputs["stop_flags"][idx : idx + 1] = False self.share_inputs["first_token_ids"][idx : idx + 1] = self.share_inputs["input_ids"][idx : idx + 1, :1] self.share_inputs["ori_seq_lens_encoder"][idx : idx + 1] = length if request.get("seed") is not None: self.share_inputs["infer_seed"][idx : idx + 1] = request.get("seed") encoder_block_num = len(request["block_tables"]) self.share_inputs["encoder_block_lens"][idx : idx + 1] = encoder_block_num self.share_inputs["block_tables"][idx : idx + 1, :] = -1 self.share_inputs["block_tables"][idx : idx + 1, :encoder_block_num] = np.array( request["block_tables"], dtype="int32" ) self.share_inputs["not_need_stop"][0] = True def warm_up_bucket(self) -> None: max_prefill_batch = 3 # Hard-Code in FastDeploy/fastdeploy/engine/config.py warmup_max_model_len = min( int(os.environ.get("HPU_WARMUP_MODEL_LEN", 4096)), self.parallel_config.max_model_len ) prefill_batchs = [] prefill_batch_step = int(os.environ.get("BATCH_STEP_PREFILL", 1)) current_prefill_batch = prefill_batch_step while current_prefill_batch <= max_prefill_batch: prefill_batchs.append(int(current_prefill_batch)) current_prefill_batch += prefill_batch_step max_prefill_length = self.cache_config.block_size + warmup_max_model_len for prefill_batch in prefill_batchs: for prefill_length in range( self.cache_config.block_size, max_prefill_length, self.cache_config.block_size ): if prefill_length * prefill_batch > self.scheduler_config.max_num_batched_tokens: continue logger.info(f"Warmup prefill_batch: {prefill_batch}, prefill_length: {prefill_length} start") requests = [ { "idx": i, "input_ids": [5] * (prefill_length - 1), "block_tables": list(range(prefill_length // self.cache_config.block_size)), "eos_token_ids": [2], } for i in range(prefill_batch) ] self.update_warmup_inputs(requests, is_decode=False) self.execute_model() logger.info(f"warmup prefill_batch: {prefill_batch}, prefill_length: {prefill_length} done") decode_batchs = [] decode_batch_step = int(os.environ.get("BATCH_STEP_DECODE", 4)) current_decode_batch = decode_batch_step while current_decode_batch <= self.scheduler_config.max_num_seqs: decode_batchs.append(int(current_decode_batch)) current_decode_batch += decode_batch_step decode_block_nums = [] decode_block_num_step = int(os.environ.get("BLOCK_STEP_DECODE", 16)) current_decode_block_num = decode_block_num_step pre_max_block_num = ( warmup_max_model_len + self.cache_config.block_size - 1 ) // self.cache_config.block_size + self.cache_config.enc_dec_block_num while current_decode_block_num <= min( self.num_gpu_blocks, pre_max_block_num * self.scheduler_config.max_num_seqs ): decode_block_nums.append(int(current_decode_block_num)) current_decode_block_num += decode_block_num_step logger.info(f"warmup decode_batchs: {decode_batchs}, decode_block_nums: {decode_block_nums} start") for decode_batch in decode_batchs: for decode_block_num in decode_block_nums: if decode_block_num < decode_batch: continue if decode_block_num // decode_batch * self.cache_config.block_size > warmup_max_model_len: continue blocks = [decode_block_num // decode_batch for _ in range(decode_batch)] remain_block_num = decode_block_num % decode_batch b = 0 while remain_block_num > 0: blocks[b] += 1 remain_block_num -= 1 b += 1 if blocks[0] * self.cache_config.block_size > warmup_max_model_len: continue logger.info(f"warmup decode_batch: {decode_batch}, decode_block_num: {decode_block_num} start") requests = [ { "idx": i, "input_ids": [5] * (blocks[i] * self.cache_config.block_size - 1), "block_tables": list(range(blocks[i])), "eos_token_ids": [2], } for i in range(decode_batch) ] self.update_warmup_inputs(requests, is_decode=True) self.execute_model() logger.info(f"Warmup decode_batch: {decode_batch}, decode_block_num: {decode_block_num} done") self.share_inputs["not_need_stop"][0] = False logger.info("Warmup bucket done") def capture_model(self) -> None: """ Trigger CUDA Graph capture for all shapes in 'CudaGraphConfig.cudagraph_capture_sizes' """ if not self.use_cudagraph: logger.info("Skipping CUDA graph capture. Please check GraphOptimizationConfig") return time_before_capture = time.perf_counter() expected_decode_len = 1 capture_sizes = self.cudagraph_capture_sizes.copy() for batch_size in sorted(capture_sizes, reverse=True): self._dummy_run( num_tokens=self.parallel_config.max_model_len, batch_size=batch_size, in_capturing=True, expected_decode_len=expected_decode_len, ) logger.info(f"Warm up the model with the batch size:{batch_size}, num tokens:{expected_decode_len}") time_after_capture = time.perf_counter() logger.info(f"Cuda Graph capturing took {time_after_capture - time_before_capture} seconds") def _get_skip_idx(self, model_forward_batch): """ Get the index of the request that needs to be skipped during execution. Args: model_forward_batch: A list of requests to be executed by this runner. Returns: A list of indices corresponding to the requests that need to be skipped. """ skip_idx_list = [] if not self.parallel_config.enable_chunked_prefill or self.guided_backend is None: return skip_idx_list for task in model_forward_batch: if task.get("prefill_chunk_info", None) is None or task.chunk_idx >= len(task.prefill_chunk_info): continue skip_idx_list.append(task.idx) for task in self.restore_chunked_prefill_request.values(): if task.idx in skip_idx_list or task.chunk_idx >= len(task.prefill_chunk_info): continue skip_idx_list.append(task.idx) return skip_idx_list def execute_model( self, model_forward_batch: Optional[List[Request]] = None, ) -> Optional[ModelRunnerOutput]: """ The Entrance of model execute. Args: model_forward_batch: 'Request' contains information related to prompt and is an abstract class at the server level, which is too granular for ModelRunner. We plan to replace it with 'ModelForwardBatch'. intermediate_tensors: """ # # 1. Prepare inputs of model and decoder. start_time = time.time() self._prepare_inputs() # self.share_inputs["ids_remove_padding"].cpu() # # 2. Padding inputs for cuda grph end_time = time.time() execution_time = (end_time - start_time) * 1000 real_bs = self.share_inputs["ids_remove_padding"].shape[0] hpu_model_runner_profile_logger.info(f"_prepare_inputs time(ms): {execution_time}, BT={real_bs}") start_time = time.time() # # 3. Execute model model_output = self.model(self.share_inputs["ids_remove_padding"], self.forward_meta) if self.is_hpu_perf_breakdown_sync_mode: model_output.cpu() end_time = time.time() execution_time = (end_time - start_time) * 1000 hpu_model_runner_profile_logger.info( f"Model execution time(ms): {execution_time}, BT={real_bs}, block_list_shape={self.share_inputs['block_list'].shape}, block_indices_shape={self.share_inputs['block_indices'].shape}" ) start_time = time.time() start_time0 = time.time() hiddden_states = rebuild_padding_v3_1( model_output, self.forward_meta.batch_ids, self.forward_meta.total_batch, self.forward_meta.seq_lens_encoder, self.forward_meta.is_prompt, ) end_time0 = time.time() execution_time0 = (end_time0 - start_time0) * 1000 hpu_model_runner_profile_logger.info(f"RebuildPadding execution time(ms): {execution_time0}, BT={real_bs}") # # 4. Compute logits, Sample start_time1 = time.time() logits = self.model.compute_logits(hiddden_states) end_time1 = time.time() execution_time1 = (end_time1 - start_time1) * 1000 hpu_model_runner_profile_logger.info(f"ComputeLogits execution time(ms): {execution_time1}, BT={real_bs}") # data = np.random.rand(self.scheduler_config.max_num_seqs, self.model_config.vocab_size).astype(np.float32) # logits = paddle.to_tensor(data, dtype='bfloat16') start_time2 = time.time() self._prepare_sampler_inputs(self.forward_meta.batch_ids) sampled_token_ids = self.sampler( logits, self.sampling_metadata, self.forward_meta.batch_ids, self.forward_meta.seq_lens_encoder.shape[0], self.rank, self.local_rank, ) if self.parallel_config.tensor_parallel_size > 1: dtype = sampled_token_ids.dtype sampled_token_ids = sampled_token_ids.to("float32") paddle.distributed.broadcast(sampled_token_ids, 0) sampled_token_ids = sampled_token_ids.to(dtype) if self.is_hpu_perf_breakdown_sync_mode: sampled_token_ids.cpu() end_time2 = time.time() execution_time2 = (end_time2 - start_time2) * 1000 hpu_model_runner_profile_logger.info(f"Sampler execution time(ms): {execution_time2}, BT={real_bs}") # 5. Post Process start_time3 = time.time() model_output_data = ModelOutputData( next_tokens=self.share_inputs["next_tokens"], stop_flags=self.share_inputs["stop_flags"], step_idx=self.share_inputs["step_idx"], max_dec_len=self.share_inputs["max_dec_len"], pre_ids=self.share_inputs["pre_ids"], seq_lens_this_time=self.share_inputs["seq_lens_this_time"], eos_token_id=self.share_inputs["eos_token_id"], not_need_stop=self.share_inputs["not_need_stop"], input_ids=self.share_inputs["input_ids"], stop_nums=self.share_inputs["stop_nums"], seq_lens_encoder=self.share_inputs["seq_lens_encoder"], seq_lens_decoder=self.share_inputs["seq_lens_decoder"], is_block_step=self.share_inputs["is_block_step"], full_hidden_states=model_output, msg_queue_id=self.parallel_config.msg_queue_id, mp_rank=self.local_rank, use_ep=self.parallel_config.use_ep, draft_tokens=self.share_inputs["draft_tokens"] if self.speculative_decoding else None, actual_draft_token_num=self.share_inputs["actual_draft_token_num"] if self.speculative_decoding else None, accept_tokens=self.share_inputs["accept_tokens"] if self.speculative_decoding else None, accept_num=self.share_inputs["accept_num"] if self.speculative_decoding else None, ) # if self.speculative_config.method in ["mtp"] and self.parallel_config.splitwise_role == "prefill": # skip_save_output = True # else: # skip_save_output = False post_process_hpu( sampled_token_ids=sampled_token_ids, model_output=model_output_data, is_warmuping=self.is_warmuping ) end_time3 = time.time() execution_time3 = (end_time3 - start_time3) * 1000 hpu_model_runner_profile_logger.info(f"PostProcessHpu execution time(ms): {execution_time3}, BT={real_bs}") end_time = time.time() execution_time = (end_time - start_time) * 1000 hpu_model_runner_profile_logger.info(f"PostProcessing execution time(ms): {execution_time}, BT={real_bs}") # 6. Speculative decode if self.speculative_decoding: if self.speculative_method == "mtp": self.proposer.run(full_hidden_states=hiddden_states) else: self.proposer.run(share_inputs=self.share_inputs) # 7. Updata 'infer_seed' and step_cuda() self.share_inputs["infer_seed"].add_(self.infer_seed_increment) self.share_inputs["infer_seed"][:] %= self.MAX_INFER_SEED start_time = time.time() step_intel_hpu(self.share_inputs, self.cache_config.block_size, self.parallel_config.max_model_len) end_time = time.time() execution_time = (end_time - start_time) * 1000 hpu_model_runner_profile_logger.info(f"StepPaddle execution time(ms): {execution_time}, BT={real_bs}") self._update_chunked_prefill(model_forward_batch) self._add_cache(model_forward_batch) if int(os.environ.get("HABANA_PROFILE", 0)) == 1: self.prof.step() return None def _add_cache(self, model_forward_batch) -> None: """ Add cache for guided decoding. """ if self.guided_backend is None: return for request in model_forward_batch: logits_cached = request.get("logits_cached", None) if logits_cached is None or logits_cached: continue request.logits_cached = True if isinstance(request.logits_processor, LogitsProcessorBase): self.guided_backend.add_cache(request.schemata_key, request.logits_processor) else: self.guided_backend.add_cache(request.schemata_key, request.logits_processor.result()) def _execute_empty_input(self) -> None: """ In certain scenarios, such as during EP, the runner needs to execute partial modules of the model without input data. This requires the model to implement the `empty_input_forward` method. """ if hasattr(self.model, "empty_input_forward"): self.model.empty_input_forward() else: raise ValueError(f"{type(self.model)} has no attribute 'empty_input_forward") def profile_run(self) -> None: """Execute a forward pass with dummy inputs to profile the memory usage of the model.""" # Initialize kv cache for profile run. After profile run kv cache will be reset. # TODO(gongshaotian): Optimize the management logic of kvcache self.num_gpu_blocks = self.parallel_config.total_block_num self.initialize_kv_cache() # 1. Profile with multimodal encoder & encoder cache # 2. Dummy run self._dummy_run( num_tokens=self.scheduler_config.max_num_batched_tokens, batch_size=min(self.scheduler_config.max_num_seqs, 3), ) # 3. gc self.clear_cache() if self.speculative_method in ["mtp"]: self.proposer.clear_dummy_input() def update_share_input_block_num(self, num_gpu_blocks: int) -> None: """ Set a globally unified block number and update the model's shared input. Args: num_gpu_blocks: """ self.num_gpu_blocks = num_gpu_blocks # Reset block table and kv cache with global block num self.initialize_kv_cache() # Reset free list free_list = list( range(self.num_gpu_blocks - 2, int(self.num_gpu_blocks * self.cache_config.kv_cache_ratio) - 1, -1) ) self.free_list_len = len(free_list) self.share_inputs.update( { "free_list": paddle.to_tensor(free_list, dtype="int32").cpu(), "free_list_len": paddle.full([1], self.free_list_len, dtype="int32").cpu(), } ) self.parallel_config.do_profile = False if self.speculative_method in ["mtp"]: self.proposer.update_block_num(num_gpu_blocks) def cal_theortical_kvcache(self): """ Calculate the total block memory required at the model level TODO(gongshaotian): Move to Attention Backend """ """ Byte of dtype: - default(bf16): 2 - cache_int8: 1 - cache_int4: """ cache_quant_dtype = None if ( self.quant_config and hasattr(self.quant_config, "kv_cache_quant_type") and self.quant_config.kv_cache_quant_type is not None ): cache_quant_dtype = self.quant_config.kv_cache_quant_type if cache_quant_dtype is not None: # int8, int8_zp, fp8, fp8_zp byte_of_dtype = 1 else: # default byte_of_dtype = 2 hidden_dim = self.model_config.head_dim * self.model_config.kv_num_heads # NOTE(liuzichang): Implement multi-layer MTP architecture in the future num_layers = ( self.model_config.num_hidden_layers + self.speculative_config.num_gpu_block_expand_ratio if self.speculative_method in ["mtp"] else self.model_config.num_hidden_layers ) required_memory = byte_of_dtype * 2 * (self.cache_config.block_size * hidden_dim) * num_layers # k + v return required_memory def not_need_stop(self) -> bool: """ """ return self.share_inputs["not_need_stop"][0]