# Copyright (c) 2026, NVIDIA CORPORATION. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # Not making this deterministic because the HuggingFace Qwen2 rotary embedding uses a CuBLAS matmul # that is incompatible with torch.use_deterministic_algorithms(True) on CUDA >= 10.2. # Setting CUBLAS_WORKSPACE_CONFIG is not sufficient since the error originates inside the # transformers library (modeling_qwen2.py Qwen2RotaryEmbedding.forward). # This legacy EasyMagpieTTS checkpoint trained context text without adding CAS embeddings. CUBLAS_WORKSPACE_CONFIG=:4096:8 HF_HUB_OFFLINE=1 TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD=1 coverage run -a --data-file=/workspace/.coverage --source=/workspace/nemo examples/tts/magpietts_inference.py \ --deterministic \ --codecmodel_path /home/TestData/tts/25fps_spectral_codec_with_bandwidth_extension.nemo \ --nemo_files /home/TestData/tts/2603_EasyMagpieTTS/EMTTS_Pretraining_Qwen_WithCrossLingual_3_5_Delay.nemo \ --out_dir ./emp_zs_0 \ --model_type easy_magpie \ --batch_size 4 \ --datasets_json_path examples/tts/evalset_config.json \ --datasets an4_val_ci_nemotronTokenizer \ --use_local_transformer \ --topk 80 \ --use_cfg \ --cfg_scale 2.5 \ --phoneme_input_type predicted \ --disable_cas_for_context_text \ --run_evaluation \ --disable_fcd \ --phoneme_tokenizer_path /home/TestData/tts/2603_EasyMagpieTTS/bpe_ipa_tokenizer_2048_en_de_es_fr_hi_it_vi_zh.json \ --temperature 0.6 \ --clean_up_disk \ --cer_target 0.10 \ --ssim_target 0.70 \ --asr_model_name /home/TestData/tts/pretrained_models/parakeet-tdt-1.1b/parakeet-tdt-1.1b.nemo \ --eou_model_name /home/TestData/tts/pretrained_models/wav2vec2-base-960h