From a9bd832fc5bd62e7de739ff9e715b44a129634dc Mon Sep 17 00:00:00 2001 From: Aaron Ang <67321817+aaron-ang@users.noreply.github.com> Date: Thu, 10 Apr 2025 02:01:00 -0400 Subject: [PATCH] [Model] use AutoWeightsLoader for deepseek_v2, internlm2 (#16383) Signed-off-by: Aaron Ang --- vllm/model_executor/models/deepseek_v2.py | 174 +++++++++++----------- vllm/model_executor/models/internlm2.py | 75 +++++----- 2 files changed, 127 insertions(+), 122 deletions(-) diff --git a/vllm/model_executor/models/deepseek_v2.py b/vllm/model_executor/models/deepseek_v2.py index 23b450ae..62714f88 100644 --- a/vllm/model_executor/models/deepseek_v2.py +++ b/vllm/model_executor/models/deepseek_v2.py @@ -53,7 +53,7 @@ from vllm.model_executor.sampling_metadata import SamplingMetadata from vllm.sequence import IntermediateTensors from .interfaces import SupportsPP -from .utils import (PPMissingLayer, is_pp_missing_parameter, +from .utils import (AutoWeightsLoader, PPMissingLayer, is_pp_missing_parameter, make_empty_intermediate_tensors_factory, make_layers, maybe_prefix) @@ -668,6 +668,91 @@ class DeepseekV2Model(nn.Module): hidden_states, _ = self.norm(hidden_states, residual) return hidden_states + def load_weights(self, weights: Iterable[Tuple[str, + torch.Tensor]]) -> Set[str]: + stacked_params_mapping = [ + # (param_name, shard_name, shard_id) + ("gate_up_proj", "gate_proj", 0), + ("gate_up_proj", "up_proj", 1), + ] + + # Params for weights, fp8 weight scales, fp8 activation scales + # (param_name, weight_name, expert_id, shard_id) + expert_params_mapping = FusedMoE.make_expert_params_mapping( + ckpt_gate_proj_name="gate_proj", + ckpt_down_proj_name="down_proj", + ckpt_up_proj_name="up_proj", + num_experts=self.config.n_routed_experts) + + params_dict = dict(self.named_parameters()) + loaded_params: Set[str] = set() + for name, loaded_weight in weights: + spec_layer = get_spec_layer_idx_from_weight_name(self.config, name) + if spec_layer is not None: + continue # skip spec decode layers for main model + + for (param_name, weight_name, shard_id) in stacked_params_mapping: + # Skip non-stacked layers and experts (experts handled below). + if weight_name not in name: + continue + # We have mlp.experts[0].gate_proj in the checkpoint. + # Since we handle the experts below in expert_params_mapping, + # we need to skip here BEFORE we update the name, otherwise + # name will be updated to mlp.experts[0].gate_up_proj, which + # will then be updated below in expert_params_mapping + # for mlp.experts[0].gate_gate_up_proj, which breaks load. + if (("mlp.experts." in name) and name not in params_dict): + continue + name = name.replace(weight_name, param_name) + # Skip loading extra bias for GPTQ models. + if name.endswith(".bias") and name not in params_dict: + continue + + if is_pp_missing_parameter(name, self): + continue + + param = params_dict[name] + weight_loader = param.weight_loader + weight_loader(param, loaded_weight, shard_id) + break + else: + for mapping in expert_params_mapping: + param_name, weight_name, expert_id, shard_id = mapping + if weight_name not in name: + continue + name = name.replace(weight_name, param_name) + + if is_pp_missing_parameter(name, self): + continue + + param = params_dict[name] + weight_loader = param.weight_loader + weight_loader(param, + loaded_weight, + name, + shard_id=shard_id, + expert_id=expert_id) + break + else: + # Skip loading extra bias for GPTQ models. + if name.endswith(".bias") and name not in params_dict: + continue + + # Remapping the name of FP8 kv-scale. + name = maybe_remap_kv_scale_name(name, params_dict) + if name is None: + continue + + if is_pp_missing_parameter(name, self): + continue + + param = params_dict[name] + weight_loader = getattr(param, "weight_loader", + default_weight_loader) + weight_loader(param, loaded_weight) + loaded_params.add(name) + return loaded_params + class DeepseekV2ForCausalLM(nn.Module, SupportsPP): @@ -737,91 +822,8 @@ class DeepseekV2ForCausalLM(nn.Module, SupportsPP): def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]) -> Set[str]: - stacked_params_mapping = [ - # (param_name, shard_name, shard_id) - ("gate_up_proj", "gate_proj", 0), - ("gate_up_proj", "up_proj", 1), - ] - - # Params for weights, fp8 weight scales, fp8 activation scales - # (param_name, weight_name, expert_id, shard_id) - expert_params_mapping = FusedMoE.make_expert_params_mapping( - ckpt_gate_proj_name="gate_proj", - ckpt_down_proj_name="down_proj", - ckpt_up_proj_name="up_proj", - num_experts=self.config.n_routed_experts) - - params_dict = dict(self.named_parameters()) - loaded_params: Set[str] = set() - for name, loaded_weight in weights: - if "rotary_emb.inv_freq" in name: - continue - - spec_layer = get_spec_layer_idx_from_weight_name(self.config, name) - if spec_layer is not None: - continue # skip spec decode layers for main model - - for (param_name, weight_name, shard_id) in stacked_params_mapping: - # Skip non-stacked layers and experts (experts handled below). - if weight_name not in name: - continue - # We have mlp.experts[0].gate_proj in the checkpoint. - # Since we handle the experts below in expert_params_mapping, - # we need to skip here BEFORE we update the name, otherwise - # name will be updated to mlp.experts[0].gate_up_proj, which - # will then be updated below in expert_params_mapping - # for mlp.experts[0].gate_gate_up_proj, which breaks load. - if (("mlp.experts." in name) and name not in params_dict): - continue - name = name.replace(weight_name, param_name) - # Skip loading extra bias for GPTQ models. - if name.endswith(".bias") and name not in params_dict: - continue - - if is_pp_missing_parameter(name, self): - continue - - param = params_dict[name] - weight_loader = param.weight_loader - weight_loader(param, loaded_weight, shard_id) - break - else: - for mapping in expert_params_mapping: - param_name, weight_name, expert_id, shard_id = mapping - if weight_name not in name: - continue - name = name.replace(weight_name, param_name) - - if is_pp_missing_parameter(name, self): - continue - - param = params_dict[name] - weight_loader = param.weight_loader - weight_loader(param, - loaded_weight, - name, - shard_id=shard_id, - expert_id=expert_id) - break - else: - # Skip loading extra bias for GPTQ models. - if name.endswith(".bias") and name not in params_dict: - continue - - # Remapping the name of FP8 kv-scale. - name = maybe_remap_kv_scale_name(name, params_dict) - if name is None: - continue - - if is_pp_missing_parameter(name, self): - continue - - param = params_dict[name] - weight_loader = getattr(param, "weight_loader", - default_weight_loader) - weight_loader(param, loaded_weight) - loaded_params.add(name) - return loaded_params + loader = AutoWeightsLoader(self, skip_prefixes=["rotary_emb.inv_freq"]) + return loader.load_weights(weights) class DeepseekV3ForCausalLM(DeepseekV2ForCausalLM): diff --git a/vllm/model_executor/models/internlm2.py b/vllm/model_executor/models/internlm2.py index 520b85c0..bf544ed3 100644 --- a/vllm/model_executor/models/internlm2.py +++ b/vllm/model_executor/models/internlm2.py @@ -32,7 +32,7 @@ from vllm.model_executor.sampling_metadata import SamplingMetadata from vllm.sequence import IntermediateTensors, PoolerOutput from .interfaces import SupportsLoRA, SupportsPP -from .utils import (is_pp_missing_parameter, +from .utils import (AutoWeightsLoader, is_pp_missing_parameter, make_empty_intermediate_tensors_factory, make_layers, maybe_prefix) @@ -306,6 +306,42 @@ class InternLM2Model(nn.Module): hidden_states, _ = self.norm(hidden_states, residual) return hidden_states + def load_weights(self, weights: Iterable[Tuple[str, + torch.Tensor]]) -> Set[str]: + stacked_params_mapping = [ + # (param_name, shard_name, shard_id) + ("gate_up_proj", "w1", 0), + ("gate_up_proj", "w3", 1), + ] + params_dict = dict(self.named_parameters()) + loaded_params: Set[str] = set() + for name, loaded_weight in weights: + for (param_name, weight_name, shard_id) in stacked_params_mapping: + if weight_name not in name: + continue + name = name.replace(weight_name, param_name) + # Skip loading extra bias for GPTQ models. + if name.endswith(".bias") and name not in params_dict: + continue + if is_pp_missing_parameter(name, self): + continue + param = params_dict[name] + weight_loader = param.weight_loader + weight_loader(param, loaded_weight, shard_id) + break + else: + # Skip loading extra bias for GPTQ models. + if name.endswith(".bias") and name not in params_dict: + continue + if is_pp_missing_parameter(name, self): + continue + param = params_dict[name] + weight_loader = getattr(param, "weight_loader", + default_weight_loader) + weight_loader(param, loaded_weight) + loaded_params.add(name) + return loaded_params + class InternLM2ForCausalLM(nn.Module, SupportsPP, SupportsLoRA): packed_modules_mapping = { @@ -373,41 +409,8 @@ class InternLM2ForCausalLM(nn.Module, SupportsPP, SupportsLoRA): def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]) -> Set[str]: - stacked_params_mapping = [ - # (param_name, shard_name, shard_id) - ("gate_up_proj", "w1", 0), - ("gate_up_proj", "w3", 1), - ] - params_dict = dict(self.named_parameters()) - loaded_params: Set[str] = set() - for name, loaded_weight in weights: - if "rotary_emb.inv_freq" in name: - continue - for (param_name, weight_name, shard_id) in stacked_params_mapping: - if weight_name not in name: - continue - name = name.replace(weight_name, param_name) - # Skip loading extra bias for GPTQ models. - if name.endswith(".bias") and name not in params_dict: - continue - if is_pp_missing_parameter(name, self): - continue - param = params_dict[name] - weight_loader = param.weight_loader - weight_loader(param, loaded_weight, shard_id) - break - else: - # Skip loading extra bias for GPTQ models. - if name.endswith(".bias") and name not in params_dict: - continue - if is_pp_missing_parameter(name, self): - continue - param = params_dict[name] - weight_loader = getattr(param, "weight_loader", - default_weight_loader) - weight_loader(param, loaded_weight) - loaded_params.add(name) - return loaded_params + loader = AutoWeightsLoader(self, skip_prefixes=["rotary_emb.inv_freq"]) + return loader.load_weights(weights) class InternLM2ForRewardModel(InternLM2ForCausalLM):