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

54
vllm/model_executor/models/gpt_bigcode.py
View File

@@ -18,7 +18,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
"""Inference-only GPTBigCode model compatible with HuggingFace weights."""
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 GPTBigCodeConfig
from vllm.attention import Attention, AttentionMetadata
from vllm.config import CacheConfig, LoRAConfig
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.sampler import Sampler, SamplerOutput
from vllm.model_executor.layers.vocab_parallel_embedding import (
ParallelLMHead, VocabParallelEmbedding)
@@ -41,7 +40,9 @@ 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 SupportsLoRA
from .interfaces import SupportsLoRA, SupportsPP
from .utils import (is_pp_missing_parameter,
make_empty_intermediate_tensors_factory, make_layers)
class GPTBigCodeAttention(nn.Module):
@@ -194,6 +195,7 @@ class GPTBigCodeModel(nn.Module):
cache_config: Optional[CacheConfig] = None,
quant_config: Optional[QuantizationConfig] = None,
lora_config: Optional[LoRAConfig] = None,
prefix: str = "",
):
super().__init__()
self.config = config
@@ -207,11 +209,15 @@ class GPTBigCodeModel(nn.Module):
self.embed_dim,
org_num_embeddings=config.vocab_size)
self.wpe = nn.Embedding(config.max_position_embeddings, self.embed_dim)
self.h = nn.ModuleList([
GPTBigCodeBlock(config, cache_config, quant_config)
for _ in range(config.num_hidden_layers)
])
self.start_layer, self.end_layer, self.h = make_layers(
config.num_hidden_layers,
lambda prefix: GPTBigCodeBlock(config, cache_config, quant_config),
prefix=f"{prefix}.h",
)
self.ln_f = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_epsilon)
self.make_empty_intermediate_tensors = (
make_empty_intermediate_tensors_factory(["hidden_states"],
config.n_embd))
def forward(
self,
@@ -219,20 +225,28 @@ class GPTBigCodeModel(nn.Module):
position_ids: torch.Tensor,
kv_caches: List[torch.Tensor],
attn_metadata: AttentionMetadata,
) -> torch.Tensor:
inputs_embeds = self.wte(input_ids)
position_embeds = self.wpe(position_ids)
hidden_states = inputs_embeds + position_embeds
intermediate_tensors: Optional[IntermediateTensors],
) -> Union[torch.Tensor, IntermediateTensors]:
if get_pp_group().is_first_rank:
inputs_embeds = self.wte(input_ids)
position_embeds = self.wpe(position_ids)
hidden_states = inputs_embeds + position_embeds
else:
hidden_states = intermediate_tensors["hidden_states"]
for i in range(len(self.h)):
for i in range(self.start_layer, self.end_layer):
layer = self.h[i]
hidden_states = layer(hidden_states, kv_caches[i], attn_metadata)
hidden_states = layer(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.ln_f(hidden_states)
return hidden_states
class GPTBigCodeForCausalLM(nn.Module, SupportsLoRA):
class GPTBigCodeForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
packed_modules_mapping = {"c_attn": ["c_attn"]}
supported_lora_modules = ["c_fc", "c_proj", "wte", "c_attn"]
@@ -272,6 +286,8 @@ class GPTBigCodeForCausalLM(nn.Module, SupportsLoRA):
self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
config.vocab_size)
self.sampler = Sampler()
self.make_empty_intermediate_tensors = (
self.transformer.make_empty_intermediate_tensors)
def forward(
self,
@@ -280,9 +296,9 @@ class GPTBigCodeForCausalLM(nn.Module, SupportsLoRA):
kv_caches: List[torch.Tensor],
attn_metadata: AttentionMetadata,
intermediate_tensors: Optional[IntermediateTensors] = None,
) -> torch.Tensor:
) -> Union[torch.Tensor, IntermediateTensors]:
hidden_states = self.transformer(input_ids, positions, kv_caches,
attn_metadata)
attn_metadata, intermediate_tensors)
return hidden_states
def compute_logits(
@@ -311,6 +327,8 @@ class GPTBigCodeForCausalLM(nn.Module, SupportsLoRA):
# Skip attention mask.
# NOTE: "c_attn.bias" should not be skipped.
continue
if is_pp_missing_parameter(name, self):
continue
param = params_dict[name]
weight_loader = getattr(param, "weight_loader",
default_weight_loader)