search

X-LANCE / SLAM-LLM

Speech, Language, Audio, Music Processing with Large Language Model
MIT License
591 stars 52 forks source link

[Question] How do I inferece on custom data with pretrained model? #167

Open ylee1123 opened 2 weeks ago

ylee1123 commented 2 weeks ago

System Info

PyTorch version: 2.6.0.dev20241101+cu121 Is debug build: False CUDA used to build PyTorch: 12.1 ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.4 LTS (x86_64) GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 Clang version: Could not collect CMake version: version 3.25.0 Libc version: glibc-2.35

Python version: 3.10.12 (main, Sep 11 2024, 15:47:36) [GCC 11.4.0] (64-bit runtime) Python platform: Linux-6.8.0-45-generic-x86_64-with-glibc2.35 Is CUDA available: True CUDA runtime version: 12.2.91 CUDA_MODULE_LOADING set to: LAZY GPU models and configuration: GPU 0: NVIDIA GeForce RTX 4090 GPU 1: NVIDIA GeForce RTX 4090

Nvidia driver version: 535.183.01 cuDNN version: Could not collect HIP runtime version: N/A MIOpen runtime version: N/A Is XNNPACK available: True

CPU: Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Address sizes: 43 bits physical, 48 bits virtual Byte Order: Little Endian CPU(s): 32 On-line CPU(s) list: 0-31 Vendor ID: AuthenticAMD Model name: AMD Ryzen Threadripper PRO 3955WX 16-Cores CPU family: 23 Model: 49 Thread(s) per core: 2 Core(s) per socket: 16 Socket(s): 1 Stepping: 0 Frequency boost: enabled CPU max MHz: 4402.7339 CPU min MHz: 2200.0000 BogoMIPS: 7785.33 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif v_spec_ctrl umip rdpid overflow_recov succor smca sev sev_es Virtualization: AMD-V L1d cache: 512 KiB (16 instances) L1i cache: 512 KiB (16 instances) L2 cache: 8 MiB (16 instances) L3 cache: 64 MiB (4 instances) NUMA node(s): 1 NUMA node0 CPU(s): 0-31 Vulnerability Gather data sampling: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling: Not affected Vulnerability Retbleed: Mitigation; untrained return thunk; SMT enabled with STIBP protection Vulnerability Spec rstack overflow: Mitigation; Safe RET Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization Vulnerability Spectre v2: Mitigation; Retpolines; IBPB conditional; STIBP always-on; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected Vulnerability Srbds: Not affected Vulnerability Tsx async abort: Not affected

Versions of relevant libraries: [pip3] mypy-extensions==1.0.0 [pip3] numpy==1.26.4 [pip3] nvidia-cublas-cu11==11.10.3.66 [pip3] nvidia-cublas-cu12==12.1.3.1 [pip3] nvidia-cuda-cupti-cu11==11.7.101 [pip3] nvidia-cuda-cupti-cu12==12.1.105 [pip3] nvidia-cuda-nvrtc-cu11==11.7.99 [pip3] nvidia-cuda-nvrtc-cu12==12.1.105 [pip3] nvidia-cuda-runtime-cu11==11.7.99 [pip3] nvidia-cuda-runtime-cu12==12.1.105 [pip3] nvidia-cudnn-cu11==8.5.0.96 [pip3] nvidia-cudnn-cu12==9.1.0.70 [pip3] nvidia-cufft-cu11==10.9.0.58 [pip3] nvidia-cufft-cu12==11.0.2.54 [pip3] nvidia-curand-cu11==10.2.10.91 [pip3] nvidia-curand-cu12==10.3.2.106 [pip3] nvidia-cusolver-cu11==11.4.0.1 [pip3] nvidia-cusolver-cu12==11.4.5.107 [pip3] nvidia-cusparse-cu11==11.7.4.91 [pip3] nvidia-cusparse-cu12==12.1.0.106 [pip3] nvidia-nccl-cu11==2.14.3 [pip3] nvidia-nccl-cu12==2.21.5 [pip3] nvidia-nvjitlink-cu12==12.4.127 [pip3] nvidia-nvtx-cu11==11.7.91 [pip3] nvidia-nvtx-cu12==12.1.105 [pip3] pytorch-triton==3.1.0+cf34004b8a [pip3] torch==2.6.0.dev20241101+cu121 [pip3] torchaudio==2.5.0.dev20241101+cu121 [pip3] torchvision==0.20.0.dev20241101+cu121 [pip3] triton==2.1.0

Information

🐛 Describe the bug

After downloading pretrained EAT encoder file from the github repository, I tried to run 'inference_clotho_bs.sh' file and this error popped out. I wonder the reason for this error and the ``proper'' way to implement pretrained model to inference on both clotho dataset and custom data.

Thanks in advance.

Error logs

''' Error executing job with overrides: ['++model_config.llm_name=vicuna-7b-v1.5', '++model_config.llm_path=~/AAC/SLAM-LLM/models/pretrain/model.pt', '++model_config.llm_dim=4096', '++model_config.encoder_name=eat', '++model_config.encoder_path=~AAC/SLAM-LLM/models/pretrain/model.pt', '++model_config.encoder_dim=768', '++model_config.encoder_projector=linear', '++model_config.encoder_projector_ds_rate=5', '++model_config.normalize=true', '++dataset_config.encoder_projector_ds_rate=5', '++dataset_config.dataset=audio_dataset', '++dataset_config.val_data_path=~/AAC/SLAM-LLM/dataset/clotho-dataset/clotho/evaluation_single.jsonl', '++dataset_config.fbank_mean=-4.268', '++dataset_config.fbank_std=4.569', '++dataset_config.model_name=eat', '++dataset_config.inference_mode=true', '++dataset_config.normalize=true', '++dataset_config.input_type=mel', '++dataset_config.fixed_length=true', '++dataset_config.target_length=1024', '++train_config.model_name=aac', '++train_config.batching_strategy=custom', '++train_config.num_epochs=1', '++train_config.val_batch_size=4', '++train_config.num_workers_dataloader=0', '++train_config.output_dir=~/AAC/SLAM-LLM/exp/clotho/aac_epoch_1_step_4500', '++train_config.freeze_encoder=true', '++train_config.freeze_llm=true', '++train_config.use_peft=false', '++ckpt_path=~/AAC/SLAM-LLM/exp/clotho/aac_epoch_1_step_4500/model.pt', '++peft_ckpt=~/AAC/SLAM-LLM/exp/clotho/aac_epoch_1_step_4500', '++decode_log=~/AAC/SLAM-LLM/exp/clotho/aac_epoch_1_step_4500/decode_beam4', '++model_config.num_beams=4'] Traceback (most recent call last): File "~/AAC/SLAM-LLM/examples/slam_aac/inference_aac_batch.py", line 49, in main_hydra inference(cfg) File "~/AAC/SLAM-LLM/src/slam_llm/pipeline/inference_batch.py", line 100, in main model, tokenizer = model_factory(train_config, model_config, kwargs) File "examples/slam_aac/model/slam_model_aac.py", line 28, in model_factory encoder = setup_encoder(train_config, model_config, kwargs) File "~/AAC/SLAM-LLM/src/slam_llm/models/slam_model.py", line 85, in setup_encoder encoder = EATEncoder.load(model_config) File "~/AAC/SLAM-LLM/src/slam_llm/models/encoder.py", line 68, in load EATEncoder, cfg, task = fairseq.checkpoint_utils.load_model_ensemble_and_task([model_config.encoder_path]) File "~/AAC/fairseq/fairseq/checkpoint_utils.py", line 451, in load_model_ensemble_and_task state = load_checkpoint_to_cpu(filename, arg_overrides) File "~/AAC/fairseq/fairseq/checkpoint_utils.py", line 368, in load_checkpoint_to_cpu state = _upgrade_state_dict(state) File "~/AAC/fairseq/fairseq/checkpoint_utils.py", line 618, in _upgrade_state_dict {"criterion_name": "CrossEntropyCriterion", "best_loss": state["best_loss"]} KeyError: 'best_loss' '''

Expected behavior

I was hoping to get results on evaluation dataset (as written in the official github README.md)

ddlBoJack commented 2 weeks ago

You may need to download the EAT checkpoint from the README page of SLAM-AAC

ylee1123 commented 2 weeks ago

When I downloaded the checkpoint file and specified the path in bash file, the following error appears.

File "/PATH/TO/AAC/SLAM-LLM/examples/slam_aac/inference_aac_batch.py", line 49, in main_hydra inference(cfg) File "/PATH/TO/AAC/SLAM-LLM/src/slam_llm/pipeline/inference_batch.py", line 100, in main model, tokenizer = model_factory(train_config, model_config, kwargs) File "examples/slam_aac/model/slam_model_aac.py", line 28, in model_factory encoder = setup_encoder(train_config, model_config, kwargs) File "/PATH/TO/AAC/SLAM-LLM/src/slam_llm/models/slam_model.py", line 85, in setup_encoder encoder = EATEncoder.load(model_config) File "/PATH/TO/AAC/SLAM-LLM/src/slam_llm/models/encoder.py", line 68, in load EATEncoder, cfg, task = fairseq.checkpoint_utils.load_model_ensemble_and_task([model_config.encoder_path]) File "/PATH/TO/AAC/fairseq/fairseq/checkpoint_utils.py", line 463, in load_model_ensemble_and_task task = tasks.setup_task(cfg.task, from_checkpoint=True) File "/PATH/TO/AAC/fairseq/fairseq/tasks/init.py", line 44, in setup_task task is not None

AssertionError: Could not infer task type from {'_name': 'mae_image_classification', 'data': '/hpc_stor03/sjtu_home/wenxi.chen/mydata/audio/AS2M', 'multi_data': None, 'input_size': 224, 'local_cache_path': None, 'key': 'imgs', 'beit_transforms': False, 'target_transform': False, 'no_transform': False, 'rebuild_batches': True, 'precompute_mask_config': None, 'subsample': 1.0, 'seed': 1, 'dataset_type': 'imagefolder', 'audio_mae': True, 'h5_format': True, 'downsr_16hz': True, 'target_length': 1024, 'flexible_mask': False, 'esc50_eval': False, 'spcv2_eval': False, 'AS2M_finetune': True, 'spcv1_finetune': False, 'roll_aug': True, 'noise': False, 'weights_file': '/hpc_stor03/sjtu_home/wenxi.chen/mydata/audio/AS2M/weight_train_all.csv', 'num_samples': 200000, 'is_finetuning': False, 'label_descriptors': 'label_descriptors.csv', 'labels': 'lbl'}. Available argparse tasks: dict_keys(['multilingual_masked_lm', 'translation_multi_simple_epoch', 'speech_dlm_task', 'speech_to_text', 'text_to_speech', 'language_modeling', 'masked_lm', 'sentence_prediction', 'sentence_prediction_adapters', 'audio_pretraining', 'audio_finetuning', 'nlu_finetuning', 'multilingual_language_modeling', 'audio_classification', 'span_masked_lm', 'hubert_pretraining', 'speech_unit_modeling', 'multires_hubert_pretraining', 'translation', 'online_backtranslation', 'denoising', 'multilingual_denoising', 'simul_speech_to_text', 'simul_text_to_text', 'translation_from_pretrained_xlm', 'translation_lev', 'frm_text_to_speech', 'speech_to_speech', 'legacy_masked_lm', 'translation_from_pretrained_bart', 'multilingual_translation', 'sentence_ranking', 'semisupervised_translation', 'cross_lingual_lm', 'dummy_lm', 'dummy_masked_lm', 'dummy_mt']). Available hydra tasks: dict_keys(['speech_dlm_task', 'language_modeling', 'masked_lm', 'sentence_prediction', 'sentence_prediction_adapters', 'audio_pretraining', 'audio_finetuning', 'nlu_finetuning', 'multilingual_language_modeling', 'audio_classification', 'span_masked_lm', 'hubert_pretraining', 'speech_unit_modeling', 'multires_hubert_pretraining', 'translation', 'denoising', 'multilingual_denoising', 'simul_text_to_text', 'translation_from_pretrained_xlm', 'translation_lev', 'dummy_lm', 'dummy_masked_lm'])

how do I setup the task for fairseq library?

cwx-worst-one commented 1 week ago

You should set up the necessary environment for each model component as described in the README. Please refer to the environment setup instructions in the EAT repository for guidance. Here’s an example setup (from that repository):

git clone https://github.com/pytorch/fairseq
cd fairseq
pip install --editable ./
git clone https://github.com/cwx-worst-one/EAT