fix weights_not_loaded

update weights_not_loaded and flash_mla_with_kvcache
update paged_mqa_logits

vllm/model_executor/model_loader/default_loader.py

@@ -313,7 +313,7 @@ class DefaultModelLoader(BaseModelLoader):
         # We only enable strict check for non-quantized models
         # that have loaded weights tracking currently.
         if model_config.quantization is None and loaded_weights is not None:
-            weights_not_loaded = weights_to_load - loaded_weights
+            weights_not_loaded = {k for k in weights_not_loaded if not k.endswith("indexer.weights_proj.bias")}
             if weights_not_loaded:
                 raise ValueError(
                     "Following weights were not initialized from "

vllm/v1/attention/backends/mla/flashmla.py

@@ -336,6 +336,8 @@ class FlashMLAImpl(MLACommonImpl[FlashMLAMetadata]):
                 causal=True,
                 descale_q=layer._q_scale.reshape(1),
                 descale_k=layer._k_scale.reshape(1),
+                is_fp8_kvcache=False,
+                indices= None,
             )
 
         o = reshape_attn_output_for_spec_decode(o)

vllm/v1/attention/backends/mla/indexer.py

@@ -315,7 +315,7 @@ class DeepseekV32IndexerMetadataBuilder(AttentionMetadataBuilder):
             seq_lens = common_attn_metadata.seq_lens[:num_decodes]
             if is_deep_gemm_supported():
                 if current_platform.is_rocm():
-                    self.scheduler_metadata_buffer[:] = gemmopt.get_paged_mqa_logits_metadata(
+                    self.scheduler_metadata_buffer= gemmopt.get_paged_mqa_logits_metadata(
                         seq_lens, self.kv_cache_spec.block_size, self.num_sms
                     )
                 else: