biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Models] Add remaining model PP support (#7168)

Browse Source 
Signed-off-by: Muralidhar Andoorveedu <muralidhar.andoorveedu@centml.ai>
Signed-off-by: Murali Andoorveedu <muralidhar.andoorveedu@centml.ai>
Co-authored-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Murali Andoorveedu
2024-10-03 19:56:58 -07:00
committed by GitHub
parent 303d44790a
commit 0f6d7a9a34
69 changed files with 2585 additions and 1344 deletions
Download Patch File Download Diff File Expand all files Collapse all files

57
vllm/model_executor/models/starcoder2.py
View File

@@ -18,7 +18,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
""" PyTorch Starcoder2 model."""
from typing import Iterable, List, Optional, Tuple
from typing import Iterable, List, Optional, Tuple, Union
import torch
from torch import nn
@@ -26,14 +26,13 @@ from transformers import Starcoder2Config
from vllm.attention import Attention, AttentionMetadata
from vllm.config import CacheConfig
from vllm.distributed import get_tensor_model_parallel_world_size
from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
from vllm.model_executor.layers.activation import get_act_fn
from vllm.model_executor.layers.linear import (ColumnParallelLinear,
QKVParallelLinear,
RowParallelLinear)
from vllm.model_executor.layers.logits_processor import LogitsProcessor
from vllm.model_executor.layers.quantization.base_config import (
QuantizationConfig)
from vllm.model_executor.layers.quantization import QuantizationConfig
from vllm.model_executor.layers.rotary_embedding import get_rope
from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
from vllm.model_executor.layers.vocab_parallel_embedding import (
@@ -42,6 +41,10 @@ from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.sequence import IntermediateTensors
from .interfaces import SupportsPP
from .utils import (is_pp_missing_parameter,
make_empty_intermediate_tensors_factory, make_layers)
class Starcoder2Attention(nn.Module):
@@ -195,7 +198,8 @@ class Starcoder2Model(nn.Module):
def __init__(self,
config: Starcoder2Config,
cache_config: Optional[CacheConfig] = None,
quant_config: Optional[QuantizationConfig] = None):
quant_config: Optional[QuantizationConfig] = None,
prefix: str = ""):
super().__init__()
self.config = config
self.padding_idx = config.pad_token_id
@@ -204,13 +208,16 @@ class Starcoder2Model(nn.Module):
# TODO: consider padding_idx (currently removed)
self.embed_tokens = VocabParallelEmbedding(config.vocab_size,
config.hidden_size)
self.layers = nn.ModuleList([
Starcoder2DecoderLayer(config,
cache_config,
quant_config=quant_config)
for _ in range(config.num_hidden_layers)
])
self.start_layer, self.end_layer, self.layers = make_layers(
config.num_hidden_layers,
lambda prefix: Starcoder2DecoderLayer(
config, cache_config, quant_config=quant_config),
prefix=f"{prefix}.layers",
)
self.norm = nn.LayerNorm(config.hidden_size, eps=config.norm_epsilon)
self.make_empty_intermediate_tensors = (
make_empty_intermediate_tensors_factory(["hidden_states"],
config.hidden_size))
def forward(
self,
@@ -218,17 +225,25 @@ class Starcoder2Model(nn.Module):
positions: torch.Tensor,
kv_caches: List[torch.Tensor],
attn_metadata: AttentionMetadata,
) -> torch.Tensor:
hidden_states = self.embed_tokens(input_ids)
for i in range(len(self.layers)):
intermediate_tensors: Optional[IntermediateTensors],
) -> Union[torch.Tensor, IntermediateTensors]:
if get_pp_group().is_first_rank:
hidden_states = self.embed_tokens(input_ids)
else:
assert intermediate_tensors is not None
hidden_states = intermediate_tensors["hidden_states"]
for i in range(self.start_layer, self.end_layer):
layer = self.layers[i]
hidden_states = layer(positions, hidden_states, kv_caches[i],
hidden_states = layer(positions, hidden_states,
kv_caches[i - self.start_layer],
attn_metadata)
if not get_pp_group().is_last_rank:
return IntermediateTensors({"hidden_states": hidden_states})
hidden_states = self.norm(hidden_states)
return hidden_states
class Starcoder2ForCausalLM(nn.Module):
class Starcoder2ForCausalLM(nn.Module, SupportsPP):
def __init__(self,
config: Starcoder2Config,
@@ -255,6 +270,8 @@ class Starcoder2ForCausalLM(nn.Module):
self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
config.vocab_size)
self.sampler = Sampler()
self.make_empty_intermediate_tensors = (
self.model.make_empty_intermediate_tensors)
def forward(
self,
@@ -263,9 +280,9 @@ class Starcoder2ForCausalLM(nn.Module):
kv_caches: List[torch.Tensor],
attn_metadata: AttentionMetadata,
intermediate_tensors: Optional[IntermediateTensors] = None,
) -> torch.Tensor:
) -> Union[torch.Tensor, IntermediateTensors]:
hidden_states = self.model(input_ids, positions, kv_caches,
attn_metadata)
attn_metadata, intermediate_tensors)
return hidden_states
def compute_logits(
@@ -302,6 +319,8 @@ class Starcoder2ForCausalLM(nn.Module):
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, shard_id)
@@ -309,6 +328,8 @@ class Starcoder2ForCausalLM(nn.Module):
else:
if self.config.tie_word_embeddings and "lm_head.weight" in name:
continue
if is_pp_missing_parameter(name, self):
continue
param = params_dict[name]
weight_loader = getattr(param, "weight_loader",
default_weight_loader)