(py312) root@10-60-132-120:/home# python speech_to_text_cache_aware_streaming_infer.py \ model_path="/home/nemotron-speech-streaming-en-0.6b.nemo" \ audio_file="/home/alan_podcast2_convert_last5m.wav" \ compare_vs_offline=true \ att_context_size="[70,13]" \ amp=false \ debug_mode=false \ rnnt_decoding.greedy.use_cuda_graph_decoder=false OneLogger: Setting error_handling_strategy to DISABLE_QUIETLY_AND_REPORT_METRIC_ERROR for rank (rank=0) with OneLogger disabled. To override: explicitly set error_handling_strategy parameter. No exporters were provided. This means that no telemetry data will be collected. [NeMo W 2026-09-19 02:30:27 nemo_logging:365] /usr/local/miniconda3/envs/py312/lib/python3.12/site-packages/hydra/_internal/hydra.py:119: UserWarning: Future Hydra versions will no longer change working directory at job runtime by default. See https://hydra.cc/docs/1.2/upgrades/1.1_to_1.2/changes_to_job_working_dir/ for more information. ret = run_job( [NeMo I 2026-09-19 02:30:27 speech_to_text_cache_aware_streaming_infer:321] Hydra config: model_path: /home/nemotron-speech-streaming-en-0.6b.nemo pretrained_name: null audio_dir: null audio_type: wav audio_file: /home/alan_podcast2_convert_last5m.wav dataset_manifest: null output_path: null batch_size: 32 random_seed: null chunk_size: -1 shift_size: -1 left_chunks: 2 online_normalization: false pad_and_drop_preencoded: false att_context_size: - 70 - 13 compare_vs_offline: true cuda: null allow_mps: false amp: false amp_dtype: float16 compute_dtype: float32 matmul_precision: high use_cuda_graphs: false ctc_decoding: strategy: greedy_batch preserve_alignments: null compute_timestamps: null word_seperator: ' ' segment_seperators: - . - '!' - '?' segment_gap_threshold: null ctc_timestamp_type: all batch_dim_index: 0 greedy: preserve_alignments: false compute_timestamps: false preserve_frame_confidence: false confidence_method_cfg: name: entropy entropy_type: tsallis alpha: 0.33 entropy_norm: exp temperature: DEPRECATED ngram_lm_model: null ngram_lm_alpha: 0.0 boosting_tree: model_path: null key_phrases_file: null key_phrases_list: null key_phrase_items_list: null context_score: 1.0 depth_scaling: 2.0 unk_score: 0.0 final_eos_score: 1.0 score_per_phrase: 0.0 source_lang: en use_triton: null uniform_weights: false bpe_mode: default use_bpe_dropout: null num_of_transcriptions: 5 bpe_alpha: 0.3 var_bpe_scoring_temp: 10.0 var_bpe_penalize_subsplits: true boosting_tree_alpha: 0.0 allow_cuda_graphs: true beam: beam_size: 4 search_type: default preserve_alignments: false compute_timestamps: false return_best_hypothesis: true allow_cuda_graphs: true beam_alpha: null beam_beta: 1.0 beam_threshold: 20.0 kenlm_path: null ngram_lm_alpha: 1.0 ngram_lm_model: null boosting_tree: model_path: null key_phrases_file: null key_phrases_list: null key_phrase_items_list: null context_score: 1.0 depth_scaling: 2.0 unk_score: 0.0 final_eos_score: 1.0 score_per_phrase: 0.0 source_lang: en use_triton: null uniform_weights: false bpe_mode: default use_bpe_dropout: null num_of_transcriptions: 5 bpe_alpha: 0.3 var_bpe_scoring_temp: 10.0 var_bpe_penalize_subsplits: true boosting_tree_alpha: 0.0 flashlight_cfg: lexicon_path: null boost_path: null beam_size_token: 16 beam_threshold: 20.0 unk_weight: -.inf sil_weight: 0.0 pyctcdecode_cfg: beam_prune_logp: -10.0 token_min_logp: -5.0 prune_history: false hotwords: null hotword_weight: 10.0 wfst: beam_size: 4 search_type: riva return_best_hypothesis: true preserve_alignments: false compute_timestamps: false decoding_mode: nbest open_vocabulary_decoding: false beam_width: 10.0 lm_weight: 1.0 device: cuda arpa_lm_path: null wfst_lm_path: null riva_decoding_cfg: {} confidence_cfg: preserve_frame_confidence: false preserve_token_confidence: false preserve_word_confidence: false exclude_blank: true aggregation: min tdt_include_duration: false method_cfg: name: entropy entropy_type: tsallis alpha: 0.33 entropy_norm: exp temperature: DEPRECATED temperature: 1.0 rnnt_decoding: model_type: rnnt strategy: greedy_batch compute_hypothesis_token_set: false preserve_alignments: null tdt_include_token_duration: null confidence_cfg: preserve_frame_confidence: false preserve_token_confidence: false preserve_word_confidence: false exclude_blank: true aggregation: min tdt_include_duration: false method_cfg: name: entropy entropy_type: tsallis alpha: 0.33 entropy_norm: exp temperature: DEPRECATED fused_batch_size: -1 compute_timestamps: null compute_langs: false word_seperator: ' ' segment_seperators: - . - '!' - '?' segment_gap_threshold: null rnnt_timestamp_type: all greedy: max_symbols_per_step: 10 preserve_alignments: false preserve_frame_confidence: false tdt_include_token_duration: false tdt_include_duration_confidence: false confidence_method_cfg: name: entropy entropy_type: tsallis alpha: 0.33 entropy_norm: exp temperature: DEPRECATED loop_labels: true use_cuda_graph_decoder: false ngram_lm_model: null ngram_lm_alpha: 0.0 boosting_tree: model_path: null key_phrases_file: null key_phrases_list: null key_phrase_items_list: null context_score: 1.0 depth_scaling: 2.0 unk_score: 0.0 final_eos_score: 1.0 score_per_phrase: 0.0 source_lang: en use_triton: null uniform_weights: false bpe_mode: default use_bpe_dropout: null num_of_transcriptions: 5 bpe_alpha: 0.3 var_bpe_scoring_temp: 10.0 var_bpe_penalize_subsplits: true boosting_tree_alpha: 0.0 enable_per_stream_biasing: false beam: beam_size: 4 search_type: default score_norm: true return_best_hypothesis: true tsd_max_sym_exp_per_step: 50 alsd_max_target_len: 1.0 nsc_max_timesteps_expansion: 1 nsc_prefix_alpha: 1 maes_num_steps: 2 maes_prefix_alpha: 1 maes_expansion_gamma: 2.3 maes_expansion_beta: 2 language_model: null softmax_temperature: 1.0 preserve_alignments: false ngram_lm_model: null ngram_lm_alpha: 0.0 boosting_tree: model_path: null key_phrases_file: null key_phrases_list: null key_phrase_items_list: null context_score: 1.0 depth_scaling: 2.0 unk_score: 0.0 final_eos_score: 1.0 score_per_phrase: 0.0 source_lang: en use_triton: null uniform_weights: false bpe_mode: default use_bpe_dropout: null num_of_transcriptions: 5 bpe_alpha: 0.3 var_bpe_scoring_temp: 10.0 var_bpe_penalize_subsplits: true boosting_tree_alpha: 0.0 hat_subtract_ilm: false hat_ilm_weight: 0.0 max_symbols_per_step: 10 blank_lm_score_mode: LM_WEIGHTED_FULL pruning_mode: LATE allow_cuda_graphs: true enable_per_stream_biasing: false preserve_frame_confidence: false tdt_include_duration_confidence: false confidence_method_cfg: name: entropy entropy_type: tsallis alpha: 0.33 entropy_norm: exp temperature: DEPRECATED temperature: 1.0 durations: [] big_blank_durations: [] strip_lang_tags: false lang_tag_pattern: null decoder_type: null debug_mode: false target_lang: null strip_lang_tags: false lang_tag_pattern: null [NeMo I 2026-09-19 02:30:27 transcribe_utils:323] Restoring model : EncDecRNNTBPEModel [NeMo I 2026-09-19 02:30:31 mixins:195] Tokenizer SentencePieceTokenizer initialized with 1024 tokens [NeMo W 2026-09-19 02:30:31 modelPT:176] If you intend to do training or fine-tuning, please call the ModelPT.setup_training_data() method and provide a valid configuration file to setup the train data loader. Train config : use_lhotse: true skip_missing_manifest_entries: true input_cfg: null tarred_audio_filepaths: null manifest_filepath: null sample_rate: 16000 shuffle: true num_workers: 2 pin_memory: true max_duration: 40.0 min_duration: 0.1 text_field: answer batch_duration: null use_bucketing: true max_tps: null bucket_duration_bins: null bucket_batch_size: null num_buckets: null bucket_buffer_size: null shuffle_buffer_size: null augmentor: null [NeMo W 2026-09-19 02:30:31 modelPT:183] If you intend to do validation, please call the ModelPT.setup_validation_data() or ModelPT.setup_multiple_validation_data() method and provide a valid configuration file to setup the validation data loader(s). Validation config : use_lhotse: true manifest_filepath: /data/ASR/en/librispeech/test-other.json sample_rate: 16000 batch_size: 32 shuffle: false max_duration: 40.0 min_duration: 0.1 num_workers: 2 pin_memory: true text_field: answer tarred_audio_filepaths: null [NeMo I 2026-09-19 02:30:36 rnnt_models:232] Using RNNT Loss : warprnnt_numba Loss warprnnt_numba_kwargs: {'fastemit_lambda': 0.005, 'clamp': -1.0} [NeMo I 2026-09-19 02:30:36 rnnt_models:232] Using RNNT Loss : warprnnt_numba Loss warprnnt_numba_kwargs: {'fastemit_lambda': 0.005, 'clamp': -1.0} [NeMo I 2026-09-19 02:30:36 rnnt_models:232] Using RNNT Loss : warprnnt_numba Loss warprnnt_numba_kwargs: {'fastemit_lambda': 0.005, 'clamp': -1.0} [NeMo I 2026-09-19 02:30:39 save_restore_connector:288] Model EncDecRNNTBPEModel was successfully restored from /home/nemotron-speech-streaming-en-0.6b.nemo. [NeMo I 2026-09-19 02:30:39 speech_to_text_cache_aware_streaming_infer:356] CacheAwareStreamingConfig(chunk_size=[105, 112], shift_size=[105, 112], cache_drop_size=0, last_channel_cache_size=70, valid_out_len=14, pre_encode_cache_size=[0, 9], drop_extra_pre_encoded=2, last_channel_num=0, last_time_num=0) [NeMo I 2026-09-19 02:30:39 rnnt_models:232] Using RNNT Loss : warprnnt_numba Loss warprnnt_numba_kwargs: {'fastemit_lambda': 0.005, 'clamp': -1.0} [NeMo I 2026-09-19 02:30:39 rnnt_decoding:1041] Joint fused batch size <= 0; Will temporarily disable fused batch step in the Joint. [NeMo I 2026-09-19 02:30:39 rnnt_bpe_models:437] Changed decoding strategy to model_type: rnnt strategy: greedy_batch compute_hypothesis_token_set: false preserve_alignments: null tdt_include_token_duration: null confidence_cfg: preserve_frame_confidence: false preserve_token_confidence: false preserve_word_confidence: false exclude_blank: true aggregation: min tdt_include_duration: false method_cfg: name: entropy entropy_type: tsallis alpha: 0.33 entropy_norm: exp temperature: DEPRECATED fused_batch_size: -1 compute_timestamps: null compute_langs: false word_seperator: ' ' segment_seperators: - . - '!' - '?' segment_gap_threshold: null rnnt_timestamp_type: all greedy: max_symbols_per_step: 10 preserve_alignments: false preserve_frame_confidence: false tdt_include_token_duration: false tdt_include_duration_confidence: false confidence_method_cfg: name: entropy entropy_type: tsallis alpha: 0.33 entropy_norm: exp temperature: DEPRECATED loop_labels: true use_cuda_graph_decoder: false ngram_lm_model: null ngram_lm_alpha: 0.0 boosting_tree: model_path: null key_phrases_file: null key_phrases_list: null key_phrase_items_list: null context_score: 1.0 depth_scaling: 2.0 unk_score: 0.0 final_eos_score: 1.0 score_per_phrase: 0.0 source_lang: en use_triton: null uniform_weights: false bpe_mode: default use_bpe_dropout: null num_of_transcriptions: 5 bpe_alpha: 0.3 var_bpe_scoring_temp: 10.0 var_bpe_penalize_subsplits: true boosting_tree_alpha: 0.0 enable_per_stream_biasing: false beam: beam_size: 4 search_type: default score_norm: true return_best_hypothesis: true tsd_max_sym_exp_per_step: 50 alsd_max_target_len: 1.0 nsc_max_timesteps_expansion: 1 nsc_prefix_alpha: 1 maes_num_steps: 2 maes_prefix_alpha: 1 maes_expansion_gamma: 2.3 maes_expansion_beta: 2 language_model: null softmax_temperature: 1.0 preserve_alignments: false ngram_lm_model: null ngram_lm_alpha: 0.0 boosting_tree: model_path: null key_phrases_file: null key_phrases_list: null key_phrase_items_list: null context_score: 1.0 depth_scaling: 2.0 unk_score: 0.0 final_eos_score: 1.0 score_per_phrase: 0.0 source_lang: en use_triton: null uniform_weights: false bpe_mode: default use_bpe_dropout: null num_of_transcriptions: 5 bpe_alpha: 0.3 var_bpe_scoring_temp: 10.0 var_bpe_penalize_subsplits: true boosting_tree_alpha: 0.0 hat_subtract_ilm: false hat_ilm_weight: 0.0 max_symbols_per_step: 10 blank_lm_score_mode: LM_WEIGHTED_FULL pruning_mode: LATE allow_cuda_graphs: true enable_per_stream_biasing: false preserve_frame_confidence: false tdt_include_duration_confidence: false confidence_method_cfg: name: entropy entropy_type: tsallis alpha: 0.33 entropy_norm: exp temperature: DEPRECATED temperature: 1.0 durations: [] big_blank_durations: [] strip_lang_tags: false lang_tag_pattern: null [NeMo I 2026-09-19 02:30:47 speech_to_text_cache_aware_streaming_infer:258] Final offline transcriptions: ["Musical type shit, but it's cool yeah it no for real I agree with that anyways do you have a girlfriend I had a girlfriend but she left me for a black dude that's not very cool you should kill her no I should kill him no kill her yeah you're not supposed to do that you're not supposed to nation do for a black dude like that's not very nice that's not very nice at all It's not cool I'm not coming all the way to America to have my girlfriend stolen like that's not chill exactly like what the fuck I come here for like come on man like I came here for like uh interracial relationship you know I'm saying but um but she chose the other race so I took that personal in conclusion American girls are really weird except for me do y'all have like a racial preference when it comes to dating you know my boyfriend's Indian I'm Mexican nothing personal but like I feel like Indian guys in America you know when it comes to sexual market value you know what I'm saying kind of low on the tier nothing personal Asian guys as well I think that's like the stereotype it is but I mean it kind of exists I think it just depends on the communication between your partner also depends on the person like if the person howla cool if it's like six five like jacked as out like he's not jacked but I love him and he loves me and we communicate very well so our sex is pretty good he was actually my tutor oh that's how you guys met yeah so like his smartness attracted you kind of we met last year because I needed a chemistry tutor and then he was a year older than me and someone was like hey like if you're struggling with chemistry he can help you yo he smooth he just slide in as a tutor smooth over there yo I need to start tutoring people start tutoring Yeah what's good yo yo what's up what's your ethnicity bruh I'm love Laos that's cool uh do you live in the Us yeah I do so like uh immigrants nah no my parents yeah so they were the immigrants how's the immigration life uh I mean that's alright I mean to be honest I wouldn't know I was born and raised here but like do you get girls though like bro like I picked up so many girls at the club like these past couple weekends oh going to the club is the easiest way to pick up why why do you think it is why do you think it's easy for you because I live in Texas in the town that I'm at there's like no Asians I'm like the only Asian guy at the club is that like an advantage or like a disadvantage I think it's an advantage because like you're different you're exotic I see that but like you know Texas is like uh kind of a conservative state I just saw like uh white girls might not be into like you know people with color I mean that's like everywhere though like white girls are like hard to get but not hard for you maybe in some places California it might be a little bit easier Texas is gonna be a little bit harder you have no problem getting girls though no because like all the Hispanic girls here in Texas they'll be into some type of Asian I see I see so it's like a Asian Hispanic match yeah any tips for like a fellow Asian boy who's trying to game white woman it don't matter what race she is at the club if you're drunk just go up and talk to her I was super nervous at first uh huh you just go up say hey like you're really pretty do you wanna dance they'll start dancing with you right away for real is that easy that's how easy it is at the club Oh shit in person it's not that easy damn that's why people go to the clubs but like sometimes like I don't even get into the club I just straight up get rejected wait how old are you a sixteen sixteen Yeah that's why you can't get to the club there's one club here in Texas that allows sixteen year olds to go are you serious Yeah yeah there's a lot in the United States that like sixteen plus eighteen plus no do they drink there they can't drink unless you're twenty one but you can get in you can drink before and go oh shit as long as like security don't know can you imagine though like a bunch of sixteen years old just like hanging around the club hell yeah that's what I'm gonna do w my brother when I go back home he's not twenty one so I'm gonna have him drink before uh shit I'ma take him to the eighteen plus club so he can pick up some girls crazy where are you from I'm from China Yeah no so you guys can drink there right Yeah yeah so like all you gotta do is just find some friends here get some drinks before you go to the club just don't act stupid when you're going in if I see like a really attractive uh black woman cause that's what I'm into I just go up to her I just be like yo what's up Shaniqua uh you really pretty you wanna dance Yeah that's it that's it that's all you gotta do she'll either say yes or no if she says no okay cool just leave it and then go find someone else yo you know like low key is kind of a numbers game you know what I'm saying yeah it's like if you try enough times like one girl is gonna say yes yeah I got you girl like this is my second time at the club yo what yo how do you find the courage to do so man like holy shit it was the first time I've ever picked up a girl so it was really nerve wracking but my friends they kept on telling me to pick up a girl and I got annoyed so I was like alright fine whatever and I went up to a girl I said yo you're really pretty would you like to dance? She's like yeah sure let's dance and I told her I was like yo like you're looking really pretty and I wanna kiss you and she's like let's kiss then yo my man smooth you just gotta say that's all you gotta do but when you're drunk it's easier to say it true cause you don't care as much Yeah exact that's why people drink I guess here I'll give you my Instagram if you ever hear in Texas I'll take you out yo let's go let's do it if you ever hear in Texas I'll buy you drinks you can drink them before and I'll take you to the club yo I got my homie in Texas let's go alright man good luck with uh pulling more girls and uh yeah you have kill one man got you peace you have a good night"] [NeMo I 2026-09-19 02:31:04 speech_to_text_cache_aware_streaming_infer:299] Final streaming transcriptions: ["Musical type shit, but it's cool yeah it no for real I agree with that anyways do you have a girlfriend I had a girlfriend but she left me for a black dude that's not very cool you should kill her no I should kill him no kill her yeah you're not supposed to do that you're not supposed to nation do for a black dude like that's not very nice that's not very nice at all It's not cool I'm not coming all the way to America to have my girlfriend stolen like that's not chill exactly like what the fuck I come here for like come on man like I came here for like uh interracial relationship you know I'm saying but um but she chose the other race so I took that personal in conclusion American girls are really weird except for me do y'all have like a racial preference when it comes to dating you know my boyfriend's Indian I'm Mexican nothing personal but like I feel like Indian guys in America you know when it comes to sexual market value you know what I'm saying kind of low on the tier nothing personal Asian guys as well I think that's like the stereotype it is but I mean it kind of exists I think it just depends on the communication between your partner also depends on the person like if the person howla cool if it's like six five like jacked as out like he's not jacked but I love him and he loves me and we communicate very well so our sex is pretty good he was actually my tutor oh that's how you guys met yeah so like his smartness attracted you kind of we met last year because I needed a chemistry tutor and then he was a year older than me and someone was like hey like if you're struggling with chemistry he can help you yo he smooth he just slide in as a tutor smooth over there yo I need to start tutoring people start tutoring Yeah what's good yo yo what's up what's your ethnicity bruh I'm love Laos that's cool uh do you live in the Us yeah I do so like uh immigrants nah no my parents yeah so they were the immigrants how's the immigration life uh I mean that's alright I mean to be honest I wouldn't know I was born and raised here but like do you get girls though like bro like I picked up so many girls at the club like these past couple weekends oh going to the club is the easiest way to pick up why why do you think it is why do you think it's easy for you because I live in Texas in the town that I'm at there's like no Asians I'm like the only Asian guy at the club is that like an advantage or like a disadvantage I think it's an advantage because like you're different you're exotic I see that but like you know Texas is like uh kind of a conservative state I just saw like uh white girls might not be into like you know people with color I mean that's like everywhere though like white girls are like hard to get but not hard for you maybe in some places California it might be a little bit easier Texas is gonna be a little bit harder you have no problem getting girls though no because like all the Hispanic girls here in Texas they'll be into some type of Asian I see I see so it's like a Asian Hispanic match yeah any tips for like a fellow Asian boy who's trying to game white woman it don't matter what race she is at the club if you're drunk just go up and talk to her I was super nervous at first uh huh you just go up say hey like you're really pretty do you wanna dance they'll start dancing with you right away for real is that easy that's how easy it is at the club Oh shit in person it's not that easy damn that's why people go to the clubs but like sometimes like I don't even get into the club I just straight up get rejected wait how old are you a sixteen sixteen Yeah that's why you can't get to the club there's one club here in Texas that allows sixteen year olds to go are you serious Yeah yeah there's a lot in the United States that like sixteen plus eighteen plus no do they drink there they can't drink unless you're twenty one but you can get in you can drink before and go oh shit as long as like security don't know can you imagine though like a bunch of sixteen years old just like hanging around the club hell yeah that's what I'm gonna do w my brother when I go back home he's not twenty one so I'm gonna have him drink before uh shit I'ma take him to the eighteen plus club so he can pick up some girls crazy where are you from I'm from China Yeah no so you guys can drink there right Yeah yeah so like all you gotta do is just find some friends here get some drinks before you go to the club just don't act stupid when you're going in if I see like a really attractive uh black woman cause that's what I'm into I just go up to her I just be like yo what's up Shaniqua uh you really pretty you wanna dance Yeah that's it that's it that's all you gotta do she'll either say yes or no if she says no okay cool just leave it and then go find someone else yo you know like low key is kind of a numbers game you know what I'm saying yeah it's like if you try enough times like one girl is gonna say yes yeah I got you girl like this is my second time at the club yo what yo how do you find the courage to do so man like holy shit it was the first time I've ever picked up a girl so it was really nerve wracking but my friends they kept on telling me to pick up a girl and I got annoyed so I was like alright fine whatever and I went up to a girl I said yo you're really pretty would you like to dance? She's like yeah sure let's dance and I told her I was like yo like you're looking really pretty and I wanna kiss you and she's like let's kiss then yo my man smooth you just gotta say that's all you gotta do but when you're drunk it's easier to say it true cause you don't care as much Yeah exact that's why people drink I guess here I'll give you my Instagram if you ever hear in Texas I'll take you out yo let's go let's do it if you ever hear in Texas I'll buy you drinks you can drink them before and I'll take you to the club yo I got my homie in Texas let's go alright man good luck with uh pulling more girls and uh yeah you have kill one man got you peace you have a good night"] [NeMo I 2026-09-19 02:31:04 speech_to_text_cache_aware_streaming_infer:308] Found 0 differences in the outputs of the model in streaming mode vs offline mode. Error executing job with overrides: ['model_path=/home/nemotron-speech-streaming-en-0.6b.nemo', 'audio_file=/home/alan_podcast2_convert_last5m.wav', 'compare_vs_offline=true', 'att_context_size=[70,13]', 'amp=false', 'debug_mode=false', 'rnnt_decoding.greedy.use_cuda_graph_decoder=false'] Traceback (most recent call last): File "/home/speech_to_text_cache_aware_streaming_infer.py", line 493, in main if cfg.compare_vs_offline and len(all_refs_text) == len(all_offline_tran): ^^^^^^^^^^^^^ UnboundLocalError: cannot access local variable 'all_refs_text' where it is not associated with a value Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.