[Model] Use upstream GLM-5.3 and Qwen4-Exp configs and processor (#57387)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Harry Mellor
2026-10-02 02:25:18 +00:00
committed by GitHub
co-authored by Claude Opus 5
parent 866f02db8a
commit e3cae8d2ac
36 changed files with 366 additions and 2385 deletions
+5 -5
View File
@@ -7,16 +7,13 @@ from unittest.mock import patch
import pytest
import torch
from transformers import Qwen4ExpConfig, Qwen4ExpTextConfig
from vllm.config.speculative import SpeculativeConfig
from vllm.model_executor.models.config import (
Qwen3_5ForConditionalGenerationConfig,
Qwen4ExpForConditionalGenerationConfig,
)
from vllm.models.qwen4_exp.config import (
Qwen4ExpConfig,
Qwen4ExpTextConfig,
)
from vllm.models.qwen4_exp.nvidia.model_state import Qwen4ExpModelState
from vllm.v1.worker.gpu.model_states.mamba_hybrid import MambaHybridModelState
@@ -37,7 +34,10 @@ def _text_config(**kwargs) -> Qwen4ExpTextConfig:
"linear_num_value_heads": 2,
"linear_key_head_dim": 8,
"linear_value_head_dim": 8,
"num_experts": 0,
"num_experts": 4,
"num_experts_per_tok": 2,
# `Qwen4ExpTextConfig` requires an EOS token whenever PLE is enabled.
"eos_token_id": 1,
"hc_count": 2,
"hc_lowrank": 4,
"ple_layer_ids": [1],
+1 -1
View File
@@ -10,6 +10,7 @@ import pytest
import torch
from torch import nn
from torch.nn import functional as F
from transformers import Qwen4ExpTextConfig
import vllm.model_executor.layers.vocab_parallel_embedding as embedding_module
import vllm.model_executor.parameter as parameter_module
@@ -33,7 +34,6 @@ from vllm.models.qwen4_exp.common.ple import (
compute_ple_shard_overlap,
copy_ple_embedding_shard_,
)
from vllm.models.qwen4_exp.config import Qwen4ExpTextConfig
from vllm.models.qwen4_exp.nvidia.ngram_embedding import (
Qwen4ExpNGramEmbedding,
Qwen4ExpPLEDeviceEmbedding,