Layers that will be compiled: group_conv_0 single_conv_0 group_conv_1 single_conv_1 group_pre_2 group_post_2 single_pre_2 single_post_2 group_conv_3 single_conv_3 group_conv_4 single_conv_4 group_pre_5 group_post_5 single_pre_5 single_post_5 group_conv_6 single_conv_6 group_conv_7 single_conv_7 group_conv_8 single_conv_8 group_pre_9 group_post_9 single_pre_9 single_post_9 group_conv_10 single_conv_10 group_conv_11 single_conv_11 group_conv_12 single_conv_12 group_pre_13 group_post_13 single_pre_13 single_post_13 group_conv_14 single_conv_14 group_conv_15 single_conv_15 group_conv_16 single_conv_16 group_pre_17 group_post_17 single_pre_17 single_post_17 group_conv_18 single_conv_18 group_conv_19 single_conv_19 group_conv_20 single_conv_20 group_pre_21 group_post_21 single_pre_21 single_post_21 group_conv_22 single_conv_22 group_conv_23 single_conv_23 group_pre_24 group_post_24 single_pre_24 single_post_24 group_conv_25 single_conv_25 group_conv_26 single_conv_26 group_pre_27 group_post_27 single_pre_27 single_post_27 group_conv_28 single_conv_28 group_conv_29 single_conv_29 group_cache_0 group_cache_128 group_cache_256 group_cache_384 group_cache_512 group_cache_640 group_cache_768 group_cache_896 group_cache_1024 group_cache_1152 group_cache_1280 group_cache_1408 group_cache_1536 group_cache_1664 group_cache_1792 group_cache_1920 single_cache_127 single_cache_255 single_cache_383 single_cache_511 single_cache_639 single_cache_767 single_cache_895 single_cache_1023 single_cache_1151 single_cache_1279 single_cache_1407 single_cache_1535 single_cache_1663 single_cache_1791 single_cache_1919 single_cache_2047 conv_post_final_29 vision_0 2026-03-07 08:50:55,104 - sima_lmm.model.vision_language_model - INFO - Generating FileGenMode.DEVKIT files... Generated all mode=DEVKIT files 2026-03-07 08:50:55,116 - sima_lmm.model.vision_language_model - INFO - Generating FileGenMode.SOURCE_TO_ONNX files... 2026-03-07 08:50:59,994 - sima_lmm.model.vision_language_model - INFO - FileGenMode.SOURCE_TO_ONNX files generation completed. Generated all mode=SOURCE_TO_ONNX files 2026-03-07 08:50:59,994 - sima_lmm.model.vision_language_model - INFO - Generating FileGenMode.ONNX_TO_QUANT files... 2026-03-07 08:51:25,616 - sima_lmm.model.vision_language_model - INFO - FileGenMode.ONNX_TO_QUANT files generation completed. Generated all mode=ONNX_TO_QUANT files 2026-03-07 08:51:25,616 - sima_lmm.model.vision_language_model - INFO - Generating FileGenMode.MODEL_SDK_COMPILE files... 2026-03-07 08:51:49,571 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1024.sima 2026-03-07 08:51:49,585 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:51:49,585 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1024" 2026-03-07 08:51:49,590 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:51:49,591 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:51:49,592 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:51:49,592 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:51:49,592 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:51:49,592 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:51:49,621 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1152.sima 2026-03-07 08:51:49,632 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:51:49,632 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1152" 2026-03-07 08:51:49,636 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:51:49,636 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:51:49,638 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:51:49,638 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:51:49,638 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:51:49,638 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:51:49,650 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1280.sima 2026-03-07 08:51:49,663 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:51:49,663 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1280" 2026-03-07 08:51:49,663 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1408.sima 2026-03-07 08:51:49,667 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:51:49,667 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:51:49,668 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:51:49,669 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:51:49,669 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:51:49,669 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:51:49,675 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:51:49,675 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1408" 2026-03-07 08:51:49,678 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:51:49,679 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:51:49,680 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:51:49,680 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:51:49,680 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:51:49,680 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:51:49,824 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n128_layer26_conv.sima 2026-03-07 08:51:49,929 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:51:49,929 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n128_layer26_conv" 2026-03-07 08:51:49,935 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:51:49,935 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:51:49,985 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:51:49,985 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:51:49,986 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:51:49,986 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:51:50,070 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1536.sima 2026-03-07 08:51:50,085 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:51:50,085 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1536" 2026-03-07 08:51:50,090 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:51:50,090 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:51:50,091 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:51:50,091 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:51:50,091 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:51:50,091 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:51:52,505 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/tuple_einsum_1, 3 2026-03-07 08:51:52,505 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:51:52,532 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/tuple_einsum_1, 3 2026-03-07 08:51:52,532 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:51:52,673 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/tuple_einsum_1, 3 2026-03-07 08:51:52,673 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:51:52,679 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/tuple_einsum_1, 3 2026-03-07 08:51:52,679 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:51:52,900 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/slice_concat_0, 2 2026-03-07 08:51:52,900 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:51:56,518 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:51:56,939 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:51:57,005 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:51:57,373 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:51:57,374 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:51:57,390 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:52:05,958 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:52:06,133 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:52:06,173 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:52:06,262 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:52:06,376 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:52:06,416 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:52:06,728 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:52:06,865 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:52:06,904 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:52:06,932 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:52:06,933 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:52:06,953 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:52:07,307 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:52:07,308 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:52:07,341 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:52:07,728 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:52:07,728 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:52:07,759 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:52:10,519 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:52:10,757 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:52:10,816 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:52:11,789 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:52:11,791 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:52:11,862 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:52:12,230 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:52:12,377 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:52:15,421 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:52:15,748 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:52:15,833 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:52:16,883 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:52:16,884 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:52:16,984 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:52:25,136 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:52:26,630 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:52:27,362 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:52:27,473 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:52:44,732 - mlc.test_util.test_context - INFO - Compression done in 17s. Compression ratio: 0.966 2026-03-07 08:52:44,732 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:52:44,947 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:52:45,306 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:52:45,306 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:52:47,159 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:52:47,395 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:52:49,744 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:52:50,102 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:52:50,851 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:52:51,178 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:52:54,835 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:52:55,182 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:53:06,683 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:53:06,862 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:53:07,054 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:53:08,084 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:53:09,037 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:53:09,115 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:53:09,159 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.233 2026-03-07 08:53:09,159 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:53:09,173 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:53:09,413 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:53:09,414 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:53:14,414 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:53:14,419 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 7s. 2026-03-07 08:53:18,477 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:53:20,058 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:53:20,652 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:53:21,362 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:53:21,459 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:53:21,507 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.23 2026-03-07 08:53:21,507 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:53:21,526 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:53:21,822 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:53:21,822 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:53:22,613 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - EV74: 4 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:53:23,256 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:23,257 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1024_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1024_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1024_stage1_mla.elf, llima-compile 2026-03-07 08:53:23,528 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1664.sima 2026-03-07 08:53:23,539 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:53:23,539 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1664" 2026-03-07 08:53:23,540 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:53:23,540 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:53:23,541 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:53:23,542 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:53:23,542 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:53:23,542 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:53:24,056 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:53:24,186 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:53:24,240 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.223 2026-03-07 08:53:24,241 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:53:24,284 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:53:24,678 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:53:24,678 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:53:24,773 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:53:26,401 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:53:26,413 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/slice_concat_0, 73 2026-03-07 08:53:26,413 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:53:27,780 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:53:27,888 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:53:27,950 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.225 2026-03-07 08:53:27,950 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:53:27,971 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:53:28,299 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:53:28,299 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:53:28,723 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:53:28,729 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 8s. 2026-03-07 08:53:32,649 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:53:32,659 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 11s. 2026-03-07 08:53:34,183 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:53:34,198 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 14s. 2026-03-07 08:53:35,209 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:53:35,215 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 9s. 2026-03-07 08:53:38,936 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:53:39,205 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:53:39,205 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:53:39,205 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:39,205 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:53:39,205 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:53:39,206 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:39,206 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:53:39,206 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:53:39,206 - afe.backends.mpk.interface - INFO - EV74: 4 2026-03-07 08:53:39,206 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:53:39,206 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:39,206 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1152_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1152_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1152_mpk.json, llima-compile 2026-03-07 08:53:39,485 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1792.sima 2026-03-07 08:53:39,500 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:53:39,500 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1792" 2026-03-07 08:53:39,502 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:53:39,502 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:53:39,503 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:53:39,503 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:53:39,503 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:53:39,504 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:53:41,820 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:53:42,063 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/slice_concat_0, 73 2026-03-07 08:53:42,063 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:53:43,346 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:53:43,477 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:53:43,536 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.219 2026-03-07 08:53:43,536 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:53:43,612 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:53:43,986 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:53:43,986 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:53:47,210 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:53:47,210 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:53:47,210 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:47,210 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:53:47,211 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:53:47,211 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:47,211 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:53:47,211 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:53:47,211 - afe.backends.mpk.interface - INFO - EV74: 4 2026-03-07 08:53:47,211 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:53:47,211 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:47,211 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1408_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1408_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1408_stage1_mla_stats.yaml, llima-compile 2026-03-07 08:53:47,483 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1920.sima 2026-03-07 08:53:47,499 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:53:47,499 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1920" 2026-03-07 08:53:47,501 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:53:47,501 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:53:47,502 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:53:47,502 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:53:47,502 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:53:47,502 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:47,519 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n128_layer26_conv_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n128_layer26_conv_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n128_layer26_conv_stage1_mla.elf, llima-compile 2026-03-07 08:53:47,807 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token127.sima 2026-03-07 08:53:47,821 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:53:47,821 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token127" 2026-03-07 08:53:47,823 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:53:47,823 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:53:47,824 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:53:47,824 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:53:47,824 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:53:47,825 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - EV74: 4 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:53:48,065 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1280_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1280_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1280_stage1_mla.elf, llima-compile 2026-03-07 08:53:48,350 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token255.sima 2026-03-07 08:53:48,364 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:53:48,364 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token255" 2026-03-07 08:53:48,366 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:53:48,366 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:53:48,367 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:53:48,367 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:53:48,367 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:53:48,367 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:53:48,998 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:53:49,124 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:53:49,136 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:53:49,213 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:53:49,214 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:53:49,220 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:53:49,508 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:53:49,682 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:53:49,815 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:53:49,816 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:53:49,828 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:53:49,903 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:53:49,972 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:53:49,973 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:53:49,978 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:53:50,466 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/slice_concat_0, 106 2026-03-07 08:53:50,466 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:53:51,025 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:53:51,026 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:53:51,135 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:53:51,777 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:53:51,793 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 11s. 2026-03-07 08:53:58,035 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:53:58,081 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:54:01,948 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:54:02,345 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:54:02,504 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:54:02,515 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:54:02,518 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:54:02,518 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:54:02,518 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:54:02,554 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:54:02,554 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:54:03,334 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:54:03,678 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:54:03,764 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - EV74: 4 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:05,251 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1536_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1536_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1536_stage1_mla_stats.yaml, llima-compile 2026-03-07 08:54:05,533 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token383.sima 2026-03-07 08:54:05,545 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:54:05,545 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token383" 2026-03-07 08:54:05,546 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:54:05,547 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:54:05,548 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:54:05,548 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:54:05,548 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:54:05,548 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:54:06,238 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:54:06,569 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:54:06,600 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:54:06,663 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:54:06,729 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:54:06,741 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:54:06,956 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:54:06,957 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:54:06,962 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:54:07,853 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:54:07,854 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:54:07,931 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:54:11,509 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:54:12,046 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:54:12,379 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:54:12,408 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:54:12,411 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:54:12,411 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:54:12,412 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:54:12,502 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:54:12,503 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:54:14,419 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:54:15,430 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:54:15,561 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 0s. 2026-03-07 08:54:15,672 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 0s. 2026-03-07 08:54:15,940 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:54:16,036 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:54:16,486 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:54:16,486 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:54:16,486 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:16,487 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:54:16,487 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:54:16,487 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:16,487 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:54:16,487 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:54:16,487 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:54:16,487 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:54:16,487 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:16,487 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token127_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token127_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token127_stage1_mla_stats.yaml, llima-compile 2026-03-07 08:54:16,768 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token511.sima 2026-03-07 08:54:16,779 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:54:16,779 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token511" 2026-03-07 08:54:16,781 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:54:16,781 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:54:16,782 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:54:16,782 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:54:16,782 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:54:16,782 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:54:17,319 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:54:17,322 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:54:17,449 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:54:18,670 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:18,734 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token255_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token255_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token255_stage1_mla_stats.yaml, llima-compile 2026-03-07 08:54:18,800 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:54:18,812 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:54:19,014 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token639.sima 2026-03-07 08:54:19,024 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:54:19,024 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token639" 2026-03-07 08:54:19,026 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:54:19,026 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:54:19,027 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:54:19,027 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:54:19,027 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:54:19,028 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:54:19,092 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:54:19,093 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:54:19,100 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:54:19,434 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:54:19,507 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:54:20,873 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:54:20,984 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:54:20,996 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:54:21,333 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:54:21,334 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:54:21,339 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:54:25,981 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:54:26,535 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:54:26,819 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:54:26,845 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:54:26,847 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:54:26,847 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:54:26,848 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:54:26,929 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:54:26,929 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:54:28,673 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:54:28,674 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 0s. 2026-03-07 08:54:31,401 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:31,402 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token383_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token383_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token383_stage1_mla_stats.yaml, llima-compile 2026-03-07 08:54:31,691 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token767.sima 2026-03-07 08:54:31,701 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:54:31,701 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token767" 2026-03-07 08:54:31,703 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:54:31,703 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:54:31,704 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:54:31,705 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:54:31,705 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:54:31,705 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:54:33,152 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:54:33,244 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:54:33,610 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:54:33,729 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:54:33,741 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:54:34,143 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:54:34,144 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:54:34,150 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:54:36,392 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:54:36,473 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:54:41,380 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:54:41,972 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:54:42,332 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:54:42,362 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:54:42,365 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:54:42,365 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:54:42,366 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:54:42,463 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:54:42,463 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:54:43,111 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:54:43,122 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:54:43,518 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:54:43,649 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:54:43,963 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:54:43,990 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:54:43,993 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:54:43,993 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:54:43,993 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:54:44,083 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:54:44,083 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:54:44,532 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:54:44,534 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 0s. 2026-03-07 08:54:45,921 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:54:45,922 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 0s. 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:47,751 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token511_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token511_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token511_mpk.json, llima-compile 2026-03-07 08:54:48,028 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token895.sima 2026-03-07 08:54:48,038 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:54:48,038 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token895" 2026-03-07 08:54:48,040 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:54:48,040 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:54:48,041 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:54:48,041 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:54:48,041 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:54:48,041 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:54:48,945 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:54:48,946 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:54:48,946 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:48,946 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:54:48,946 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:54:48,946 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:48,946 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:54:48,946 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:54:48,947 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:54:48,947 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:54:48,947 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:54:48,947 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token639_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token639_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token639_stage1_mla.elf, llima-compile 2026-03-07 08:54:49,227 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1023.sima 2026-03-07 08:54:49,237 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:54:49,237 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1023" 2026-03-07 08:54:49,239 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:54:49,240 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:54:49,241 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:54:49,241 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:54:49,241 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:54:49,241 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:54:49,892 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:54:49,987 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:54:49,999 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:54:50,461 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:54:50,470 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:54:50,471 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:54:50,476 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:54:50,558 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:54:51,083 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:54:51,179 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:54:51,191 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:54:51,717 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:54:51,718 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:54:51,725 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:54:57,722 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:54:58,085 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:54:58,904 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:54:59,516 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:54:59,891 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:54:59,923 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:54:59,926 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:54:59,926 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:54:59,926 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:55:00,028 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:55:00,028 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:55:02,049 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:55:02,051 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 0s. 2026-03-07 08:55:05,449 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:05,455 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token767_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token767_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token767_stage1_mla_stats.yaml, llima-compile 2026-03-07 08:55:05,529 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:55:05,741 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1151.sima 2026-03-07 08:55:05,750 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:55:05,750 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1151" 2026-03-07 08:55:05,752 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:55:05,752 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:55:05,753 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:55:05,753 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:55:05,753 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:55:05,753 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:55:07,257 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:55:07,340 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:55:07,353 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:55:07,946 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:55:07,947 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:55:07,953 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:55:07,988 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:55:08,085 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:55:09,565 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:55:09,975 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:55:12,707 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:55:13,234 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:55:13,548 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:55:13,575 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:55:13,578 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:55:13,578 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:55:13,578 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:55:13,665 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:55:13,665 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:55:15,489 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:55:15,490 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 0s. 2026-03-07 08:55:16,418 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:55:16,859 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:55:17,063 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:55:17,436 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:55:17,466 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:55:17,470 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:55:17,470 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:55:17,470 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:55:17,568 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:55:17,568 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:55:18,637 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:55:18,637 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:55:18,637 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:18,637 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:55:18,638 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:55:18,638 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:18,638 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:55:18,638 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:55:18,638 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:55:18,638 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:55:18,638 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:18,638 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token895_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token895_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token895_mpk.json, llima-compile 2026-03-07 08:55:18,916 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1279.sima 2026-03-07 08:55:18,926 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:55:18,926 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1279" 2026-03-07 08:55:18,928 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:55:18,928 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:55:18,929 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:55:18,929 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:55:18,929 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:55:18,929 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:55:19,594 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:55:19,596 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 0s. 2026-03-07 08:55:19,863 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:55:20,465 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:55:20,555 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:55:20,568 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:55:21,231 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:55:21,232 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:55:21,237 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:55:21,455 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:55:21,593 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:55:21,655 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.218 2026-03-07 08:55:21,655 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:55:21,734 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:55:22,133 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:55:22,133 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:55:23,130 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:55:23,130 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:55:23,130 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:23,130 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:55:23,130 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:55:23,130 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:23,130 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:55:23,130 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:55:23,130 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:55:23,131 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:55:23,131 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:23,131 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1023_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1023_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1023_stage1_mla_stats.yaml, llima-compile 2026-03-07 08:55:23,410 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1407.sima 2026-03-07 08:55:23,419 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:55:23,419 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1407" 2026-03-07 08:55:23,420 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:55:23,420 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:55:23,421 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:55:23,422 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:55:23,422 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:55:23,422 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:55:24,727 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:55:24,809 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:55:24,900 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:55:24,973 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:55:24,985 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:55:25,706 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:55:25,707 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:55:25,712 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:55:27,823 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:55:30,234 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:55:30,251 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 12s. 2026-03-07 08:55:31,138 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:55:32,101 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:55:32,654 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:55:32,665 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:55:32,784 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:55:32,850 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.215 2026-03-07 08:55:32,850 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:55:32,944 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:55:32,985 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:55:33,012 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:55:33,015 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:55:33,015 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:55:33,015 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:55:33,103 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:55:33,103 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:55:33,326 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:55:33,327 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:55:34,943 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:55:34,944 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 0s. 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:38,843 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1151_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1151_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1151_stage1_mla_stats.yaml, llima-compile 2026-03-07 08:55:39,127 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1535.sima 2026-03-07 08:55:39,136 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:55:39,137 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1535" 2026-03-07 08:55:39,138 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:55:39,138 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:55:39,139 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:55:39,139 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:55:39,139 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:55:39,139 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:55:39,398 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:55:39,493 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:55:40,662 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:55:40,738 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:55:40,751 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:55:41,064 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:55:41,080 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 13s. 2026-03-07 08:55:41,542 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:55:41,543 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:55:41,549 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:55:41,977 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:55:42,052 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - EV74: 4 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:44,024 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1664_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1664_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1664_stage1_mla.elf, llima-compile 2026-03-07 08:55:44,307 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:55:44,310 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1663.sima 2026-03-07 08:55:44,332 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:55:44,332 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1663" 2026-03-07 08:55:44,334 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:55:44,334 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:55:44,335 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:55:44,335 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:55:44,335 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:55:44,335 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:55:45,460 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/slice_concat_0, 215 2026-03-07 08:55:45,460 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:55:47,926 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:55:48,032 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:55:48,555 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:55:48,940 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:55:48,972 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:55:48,975 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:55:48,975 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:55:48,975 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:55:49,074 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:55:49,075 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:55:49,339 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:55:49,758 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:55:49,865 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:55:49,908 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:55:49,986 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.214 2026-03-07 08:55:49,986 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:55:50,081 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:55:50,179 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:55:50,212 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:55:50,214 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:55:50,214 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:55:50,215 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:55:50,324 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:55:50,324 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:55:50,518 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:55:50,518 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:55:51,159 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:55:51,160 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 1s. 2026-03-07 08:55:52,665 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:55:52,666 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 1s. 2026-03-07 08:55:52,804 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:55:53,030 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:55:53,064 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:55:53,938 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:55:53,939 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:55:53,951 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:55:54,256 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - EV74: 4 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:54,257 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1792_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1792_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1792_mpk.json, llima-compile 2026-03-07 08:55:54,542 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1791.sima 2026-03-07 08:55:54,551 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:55:54,551 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1791" 2026-03-07 08:55:54,553 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:55:54,553 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:55:54,554 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:55:54,554 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:55:54,554 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:55:54,554 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:54,844 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1279_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1279_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1279_stage1_mla.elf, llima-compile 2026-03-07 08:55:55,123 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1919.sima 2026-03-07 08:55:55,136 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:55:55,136 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1919" 2026-03-07 08:55:55,137 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:55:55,137 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:55:55,138 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:55:55,139 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:55:55,139 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:55:55,139 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:55:56,156 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/slice_concat_0, 215 2026-03-07 08:55:56,156 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:55:56,644 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/slice_concat_0, 171 2026-03-07 08:55:56,644 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:55:57,160 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1407_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1407_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1407_stage1_mla_stats.yaml, llima-compile 2026-03-07 08:55:57,439 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_cache_token2047.sima 2026-03-07 08:55:57,449 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:55:57,449 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_cache_token2047" 2026-03-07 08:55:57,450 - afe.core.compile_networks - INFO - The model is split into 1 segments for MLA and APU 2026-03-07 08:55:57,450 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 1: compiling for MLA 2026-03-07 08:55:57,451 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:55:57,452 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:55:57,452 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:55:57,452 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:55:59,002 - mlc.compiler.model_graph.l1_based - INFO - Trigger large tensor support due to: MLA_0/slice_concat_0, 149 2026-03-07 08:55:59,003 - mlc.compiler.model_graph.l1_based - INFO - Segmenting layers by spatial/channel dimension 2026-03-07 08:55:59,161 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:55:59,176 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 14s. 2026-03-07 08:55:59,470 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:55:59,558 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:56:03,851 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:56:04,091 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:56:04,126 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:56:04,357 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:56:04,584 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:56:04,617 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:56:05,085 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:56:05,086 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:56:05,099 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:56:05,633 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:56:05,634 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:56:05,647 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:56:06,485 - mlc.compiler.model_graph.large_tensor_helper - INFO - Setting tile layouts 2026-03-07 08:56:06,719 - mlc.compiler.model_graph.large_tensor_helper - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:56:06,753 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:56:07,828 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:56:07,828 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:56:07,842 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:56:07,959 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:56:08,585 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:56:08,953 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:56:08,989 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:56:08,992 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:56:08,992 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:56:08,993 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:56:09,114 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:56:09,114 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:56:11,726 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:56:11,727 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 1s. 2026-03-07 08:56:14,579 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:56:14,579 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:56:14,580 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:14,580 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:56:14,580 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:56:14,580 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:14,580 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:56:14,580 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:56:14,580 - afe.backends.mpk.interface - INFO - EV74: 4 2026-03-07 08:56:14,580 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:56:14,580 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:14,580 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1920_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1920_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n128_cache_token1920_stage1_mla_stats.yaml, llima-compile 2026-03-07 08:56:15,395 - afe.ir.serializer.api - INFO - Loading model from file: CompiledModels/models--LiquidAI--LFM2-VL-3B/sima_files/sdk/models--LiquidAI--LFM2-VL-3B_language_n1_post_layer29_conv_final.sima 2026-03-07 08:56:15,623 - afe.apis.model - WARNING - A deepcopy is disabled hence model graph will be mutated and no other APIs can be called after the compile step. 2026-03-07 08:56:15,623 - afe.apis.model - INFO - Compiling quantized net "models--LiquidAI--LFM2-VL-3B_language_n1_post_layer29_conv_final" 2026-03-07 08:56:15,625 - afe.core.compile_networks - INFO - The model is split into 2 segments for MLA and APU 2026-03-07 08:56:15,626 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 1 of 2: compiling for MLA 2026-03-07 08:56:15,700 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties 2026-03-07 08:56:15,700 - mlc.compiler.model_graph.l1_based - INFO - Checking if layers fit into memory without splitting 2026-03-07 08:56:15,700 - mlc.compiler.model_graph.l1_based - INFO - Generating the list of feasible layouts 2026-03-07 08:56:15,700 - mlc.compiler.model_graph.l1_based - INFO - Setting layer parameters 2026-03-07 08:56:15,743 - mlc.compiler.model_graph.l1_based - INFO - Setting tile layouts 2026-03-07 08:56:15,749 - mlc.compiler.model_graph.l1_based - INFO - Allocating memory for IFM/OFM tensors 2026-03-07 08:56:15,763 - mlc.compiler.model_graph.l1_based - INFO - Get model compilation properties done 2026-03-07 08:56:16,271 - afe.backends.mla.afe_to_n2a_compiler.n2a_backend_runner - INFO - Start evaluate process to generate check file 2026-03-07 08:56:16,272 - mlc.kernel.layout - INFO - L2 caching mode: L2CachingMode.NONE 2026-03-07 08:56:16,279 - mlc.compiler.model_graph.l1_based - INFO - Generating model code 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:16,719 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1535_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1535_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1535_mpk.json, llima-compile 2026-03-07 08:56:18,353 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:56:18,486 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:56:29,588 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:56:29,681 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:56:29,721 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:56:30,361 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:56:30,897 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:56:30,943 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:56:30,946 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:56:30,946 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:56:30,946 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:56:31,091 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:56:31,091 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:56:31,574 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:56:31,708 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:56:33,712 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:56:33,799 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:56:33,877 - mlc.compiler.model_graph.l1_based - INFO - Generating model code done 2026-03-07 08:56:34,012 - mlc.test_util.test_context - INFO - Scheduling instructions 2026-03-07 08:56:34,079 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:56:34,080 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 1s. 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:39,352 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1663_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1663_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1663_mpk.json, llima-compile 2026-03-07 08:56:40,940 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:56:41,061 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:56:41,611 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:56:42,147 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:56:42,193 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:56:42,195 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:56:42,196 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:56:42,196 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:56:42,340 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:56:42,341 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:56:42,975 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:56:43,044 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:56:43,504 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:56:43,588 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:56:43,659 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:56:44,194 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:56:44,239 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:56:44,242 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:56:44,242 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:56:44,242 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:56:44,386 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:56:44,386 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:56:45,239 - mlc.test_util.test_context - INFO - Performing DRAM/L2 synchronization 2026-03-07 08:56:45,280 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:56:45,281 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 1s. 2026-03-07 08:56:45,981 - mlc.test_util.test_context - INFO - Inserting and merging NOPs 2026-03-07 08:56:46,529 - mlc.test_util.test_context - INFO - Setting IQ sync bits 2026-03-07 08:56:46,576 - mlc.test_util.test_context - INFO - Run compression 2026-03-07 08:56:46,579 - mlc.test_util.test_context - INFO - Compression done in 0s. Compression ratio: 0.404 2026-03-07 08:56:46,579 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:56:46,579 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:56:46,726 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:56:46,726 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:56:47,352 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:56:47,354 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 1s. 2026-03-07 08:56:49,715 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:56:49,718 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 1s. 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:50,670 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1919_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1919_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1919_stage1_mla.elf, llima-compile 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:52,651 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1791_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1791_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token1791_stage1_mla.elf, llima-compile 2026-03-07 08:56:55,142 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - EV74: 5 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - A65 : 0 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:56:55,143 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_cache_token2047_stage1_mla.elf, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token2047_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n1_cache_token2047_mpk.json, llima-compile 2026-03-07 08:57:05,364 - mlc.test_util.test_context - INFO - Compression done in 21s. Compression ratio: 0.97 2026-03-07 08:57:05,365 - mlc.test_util.test_context - INFO - Re-allocate dram memory 2026-03-07 08:57:06,025 - mlc.test_util.test_context - INFO - Generating metrics 2026-03-07 08:57:06,301 - mlc.test_util.test_context - INFO - Writing report to MLC file 2026-03-07 08:57:06,301 - mlc.test_util.test_context - INFO - Writing instructions to MLC file 2026-03-07 08:58:01,769 - mlc.test_util.test_context - INFO - Code generation done 2026-03-07 08:58:01,770 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Evaluate model and writing chk file done in 2s. 2026-03-07 08:58:05,309 - afe.backends.mla.afe_to_n2a_compiler.n2a_compiler_operations - INFO - Stage 2 of 2: backend is APU 2026-03-07 08:58:05,311 - afe.core.compile_networks - INFO - Stage 2 of 2: compiling for APU 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - ============================== 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - Compilation summary: 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - Desired batch size: 1 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - Achieved batch size: 1 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - Plugin distribution per backend: 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - MLA : 1 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - EV74: 2 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - A65 : 1 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - ------------------------------ 2026-03-07 08:58:16,997 - afe.backends.mpk.interface - INFO - Generated files: models--LiquidAI--LFM2-VL-3B_language_n1_post_layer29_conv_final_stage2_a65.so, models--LiquidAI--LFM2-VL-3B_language_n1_post_layer29_conv_final_mpk.json, models--LiquidAI--LFM2-VL-3B_language_n1_post_layer29_conv_final_stage1_mla_stats.yaml, models--LiquidAI--LFM2-VL-3B_language_n1_post_layer29_conv_final_stage1_mla.elf, llima-compile 2026-03-07 08:58:19,239 - sima_lmm.model.vision_language_model - INFO - FileGenMode.MODEL_SDK_COMPILE files generation completed. Generated all mode=MODEL_SDK_COMPILE files