[Mypy Fix] Mypy fix for "vllm/model_executor/models/[tT]" (#53466)

Signed-off-by: Jensen Chen <a1043904820@gmail.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Jensen Chen
2026-08-26 14:12:11 +00:00
committed by GitHub
co-authored by Claude Opus 5 mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 903a02192f
commit c47ca4aaeb
3 changed files with 14 additions and 9 deletions
+1 -1
View File
@@ -124,7 +124,7 @@ EXCLUDE = [
r"vllm/model_executor/models/[qQ]", r"vllm/model_executor/models/[qQ]",
r"vllm/model_executor/models/[rR]", r"vllm/model_executor/models/[rR]",
r"vllm/model_executor/models/[sS]", r"vllm/model_executor/models/[sS]",
r"vllm/model_executor/models/[tT]", r"vllm/model_executor/models/transformers",
r"vllm/model_executor/models/[uU]", r"vllm/model_executor/models/[uU]",
r"vllm/model_executor/models/[vV]", r"vllm/model_executor/models/[vV]",
r"vllm/model_executor/models/[wW]", r"vllm/model_executor/models/[wW]",
+12 -7
View File
@@ -28,7 +28,7 @@ import torch.nn as nn
from vllm.config import VllmConfig from vllm.config import VllmConfig
from vllm.model_executor.models.llama import LlamaForCausalLM, LlamaModel from vllm.model_executor.models.llama import LlamaForCausalLM, LlamaModel
from .llama import LlamaDecoderLayer from .llama import LlamaAttention, LlamaDecoderLayer, LlamaMLP
from .utils import AutoWeightsLoader, PPMissingLayer, WeightsMapper from .utils import AutoWeightsLoader, PPMissingLayer, WeightsMapper
@@ -46,7 +46,7 @@ class TeleChat2Model(LlamaModel):
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""): def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
hf_config = vllm_config.model_config.hf_config hf_config = vllm_config.model_config.hf_config
vllm_config.model_config.hf_config.attribute_map = { vllm_config.model_config.hf_config.attribute_map = { # type: ignore[misc]
"num_hidden_layers": "n_layer", "num_hidden_layers": "n_layer",
"num_attention_heads": "n_head", "num_attention_heads": "n_head",
"intermediate_size": "ffn_hidden_size", "intermediate_size": "ffn_hidden_size",
@@ -63,11 +63,16 @@ class TeleChat2Model(LlamaModel):
# Telechat2's gate_up_proj and qkv_proj don't have bias # Telechat2's gate_up_proj and qkv_proj don't have bias
# see: https://github.com/vllm-project/vllm/pull/10311#issuecomment-2490297566 # see: https://github.com/vllm-project/vllm/pull/10311#issuecomment-2490297566
for layer in self.layers: for layer in self.layers:
if not isinstance(layer, PPMissingLayer): if isinstance(layer, PPMissingLayer):
layer.self_attn.qkv_proj.bias = None continue
layer.self_attn.qkv_proj.skip_bias_add = True self_attn = layer.self_attn
layer.mlp.gate_up_proj.bias = None mlp = layer.mlp
layer.mlp.gate_up_proj.skip_bias_add = True assert isinstance(self_attn, LlamaAttention)
assert isinstance(mlp, LlamaMLP)
self_attn.qkv_proj.register_parameter("bias", None)
self_attn.qkv_proj.skip_bias_add = True
mlp.gate_up_proj.register_parameter("bias", None)
mlp.gate_up_proj.skip_bias_add = True
def _split_key_value( def _split_key_value(
self, weights: Iterable[tuple[str, torch.Tensor]] self, weights: Iterable[tuple[str, torch.Tensor]]
+1 -1
View File
@@ -289,7 +289,7 @@ class Terratorch(nn.Module, IsAttentionFree, SupportsMultiModal):
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]: def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
params_list = [] params_list = []
model_buffers = dict(self.named_buffers()) model_buffers = dict(self.named_buffers())
loaded_buffers = [] loaded_buffers: list[str] = []
for key, value in weights: for key, value in weights:
if isinstance(value, (dict, OrderedDict)): if isinstance(value, (dict, OrderedDict)):
if key == "state_dict": if key == "state_dict":