from megatron.core.fusions.fused_bias_dropout import get_bias_dropout_add
from megatron.core.tensor_parallel.layers import ColumnParallelLinear, RowParallelLinear
from megatron.core.transformer.attention import SelfAttentionSubmodules
from megatron.core.extensions.transformer_engine import (
TENorm,
)
from megatron.core.transformer.dot_product_attention import DotProductAttention
from megatron.core.transformer.enums import AttnMaskType
from megatron.core.transformer.identity_op import IdentityOp
from megatron.core.transformer.spec_utils import ModuleSpec
from megatron.core.transformer.transformer_layer import (
TransformerLayer,
TransformerLayerSubmodules,
)
from megatron.core.models.gpt.gpt_layer_specs import _get_mlp_module_spec
from megatron.training import get_args
from mindspeed_mm.models.common.module_spec.llava_layer_spec import get_mlp_module_spec
from mindspeed_mm.models.vision.vision_encoders.qwen2vl_vit_model import Qwen2vlVitSelfAttention, Qwen2vlSelfAttention, Qwen2_5VitDotProductAttention
from mindspeed_mm.patchs.canonical_layer_patch import (
PatchSplitQKVSelfAttention,
_patch_get_mlp_module_spec,
PatchViTSelfAttention,
SplitQKVSelfAttentionSubmodules
)
def get_qwen2vl_llm_layer_spec(config=None, *args, **kwargs) -> ModuleSpec:
if get_args().hetero_parallel:
from mindspeed_mm.utils.hetero_utils.hetero_CP_utils import get_hetero_dotproductattention
DOTPRODUCTATTENTION = get_hetero_dotproductattention(config)
else:
DOTPRODUCTATTENTION = DotProductAttention
qk_layernorm = False
canonical_model = getattr(config, 'canonical_model', False)
if canonical_model:
mlp = _patch_get_mlp_module_spec(use_te=False)
else:
mlp = _get_mlp_module_spec(use_te=False)
return ModuleSpec(
module=TransformerLayer,
submodules=TransformerLayerSubmodules(
input_layernorm=TENorm,
self_attention=ModuleSpec(
module=Qwen2vlSelfAttention if not canonical_model else PatchSplitQKVSelfAttention,
params={"attn_mask_type": AttnMaskType.causal},
submodules=SelfAttentionSubmodules(
linear_qkv=ColumnParallelLinear,
core_attention=DOTPRODUCTATTENTION,
linear_proj=RowParallelLinear,
q_layernorm=TENorm if qk_layernorm else IdentityOp,
k_layernorm=TENorm if qk_layernorm else IdentityOp,
) if not canonical_model else
SplitQKVSelfAttentionSubmodules(
linear_qkv=ColumnParallelLinear,
q_proj=ColumnParallelLinear,
k_proj=ColumnParallelLinear,
v_proj=ColumnParallelLinear,
core_attention=DOTPRODUCTATTENTION,
linear_proj=RowParallelLinear,
q_layernorm=TENorm if qk_layernorm else IdentityOp,
k_layernorm=TENorm if qk_layernorm else IdentityOp,
),
),
self_attn_bda=get_bias_dropout_add,
pre_mlp_layernorm=TENorm,
mlp=mlp,
mlp_bda=get_bias_dropout_add,
sharded_state_dict_keys_map={
'input_layernorm.': 'self_attention.linear_qkv.layer_norm_',
'pre_mlp_layernorm.': 'mlp.linear_fc1.layer_norm_',
},
),
)
def get_qwen2vl_layer_spec(config=None, is_vit=True, *args, **kwargs) -> ModuleSpec:
attn_mask_type = AttnMaskType.no_mask if is_vit else AttnMaskType.causal
if get_args().hetero_parallel:
from mindspeed_mm.utils.hetero_utils.hetero_CP_utils import get_hetero_dotproductattention
DOTPRODUCTATTENTION = get_hetero_dotproductattention(config)
else:
DOTPRODUCTATTENTION = DotProductAttention
canonical_model = getattr(config, 'canonical_model', False)
if canonical_model:
mlp = _patch_get_mlp_module_spec(use_te=False)
else:
mlp = _get_mlp_module_spec(use_te=False)
return ModuleSpec(
module=TransformerLayer,
submodules=TransformerLayerSubmodules(
input_layernorm=TENorm,
self_attention=ModuleSpec(
module=Qwen2vlVitSelfAttention if not canonical_model else PatchViTSelfAttention,
params={
"attn_mask_type": attn_mask_type
},
submodules=SelfAttentionSubmodules(
linear_qkv=ColumnParallelLinear,
core_attention=DOTPRODUCTATTENTION,
linear_proj=RowParallelLinear,
q_layernorm=IdentityOp,
k_layernorm=IdentityOp,
),
),
self_attn_bda=get_bias_dropout_add,
pre_mlp_layernorm=TENorm,
mlp=mlp,
mlp_bda=get_bias_dropout_add,
),
)
def get_qwen2_5_vit_layer_spec(config=None, is_vit=True, *args, **kwargs) -> ModuleSpec:
if hasattr(config, "use_vit_dp") and config.use_vit_dp:
core_attention = Qwen2_5VitDotProductAttention
else:
core_attention = DotProductAttention
attn_mask_type = AttnMaskType.no_mask if is_vit else AttnMaskType.causal
mlp = get_mlp_module_spec(use_te=False)
return ModuleSpec(
module=TransformerLayer,
submodules=TransformerLayerSubmodules(
input_layernorm=TENorm,
self_attention=ModuleSpec(
module=Qwen2vlVitSelfAttention,
params={
"attn_mask_type": attn_mask_type
},
submodules=SelfAttentionSubmodules(
linear_qkv=ColumnParallelLinear,
core_attention=core_attention,
linear_proj=RowParallelLinear,
q_layernorm=IdentityOp,
k_layernorm=IdentityOp,
),
),
self_attn_bda=get_bias_dropout_add,
pre_mlp_layernorm=TENorm,
mlp=mlp,
mlp_bda=get_bias_dropout_add,
),
)
