Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
update
游雁fcbbe8af9f22
diff --git a/examples/aishell/branchformer/conf/branchformer_12e_6d_2048_256.yaml b/examples/aishell/branchformer/conf/branchformer_12e_6d_2048_256.yamlindex d86e6281c..61f815ed5 100644--- a/examples/aishell/branchformer/conf/branchformer_12e_6d_2048_256.yaml+++ b/examples/aishell/branchformer/conf/branchformer_12e_6d_2048_256.yaml@@ -94,7 +94,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: RankFullLocalShuffleBatchSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 32 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/examples/aishell/conformer/conf/conformer_12e_6d_2048_256.yaml b/examples/aishell/conformer/conf/conformer_12e_6d_2048_256.yamlindex 4238d8462..12eeb3e19 100644--- a/examples/aishell/conformer/conf/conformer_12e_6d_2048_256.yaml+++ b/examples/aishell/conformer/conf/conformer_12e_6d_2048_256.yaml@@ -94,7 +94,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: RankFullLocalShuffleBatchSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 32 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/examples/aishell/e_branchformer/conf/e_branchformer_12e_6d_2048_256.yaml b/examples/aishell/e_branchformer/conf/e_branchformer_12e_6d_2048_256.yamlindex 27734599b..df13e67ab 100644--- a/examples/aishell/e_branchformer/conf/e_branchformer_12e_6d_2048_256.yaml+++ b/examples/aishell/e_branchformer/conf/e_branchformer_12e_6d_2048_256.yaml@@ -94,7 +94,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: RankFullLocalShuffleBatchSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 32 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/examples/aishell/paraformer/conf/paraformer_conformer_12e_6d_2048_256.yaml b/examples/aishell/paraformer/conf/paraformer_conformer_12e_6d_2048_256.yamlindex c100f0da9..5625623cb 100644--- a/examples/aishell/paraformer/conf/paraformer_conformer_12e_6d_2048_256.yaml+++ b/examples/aishell/paraformer/conf/paraformer_conformer_12e_6d_2048_256.yaml@@ -93,7 +93,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: RankFullLocalShuffleBatchSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 32 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/examples/aishell/transformer/conf/transformer_12e_6d_2048_256.yaml b/examples/aishell/transformer/conf/transformer_12e_6d_2048_256.yamlindex 674fc94c2..51936c1b7 100644--- a/examples/aishell/transformer/conf/transformer_12e_6d_2048_256.yaml+++ b/examples/aishell/transformer/conf/transformer_12e_6d_2048_256.yaml@@ -88,7 +88,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: RankFullLocalShuffleBatchSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 32 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/examples/industrial_data_pretraining/llm_asr/conf/template.yaml b/examples/industrial_data_pretraining/llm_asr/conf/template.yamlindex 3c51ff423..c64c88663 100644--- a/examples/industrial_data_pretraining/llm_asr/conf/template.yaml+++ b/examples/industrial_data_pretraining/llm_asr/conf/template.yaml@@ -73,7 +73,7 @@ scheduler_conf:dataset: AudioLLMDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: RankFullLocalShuffleBatchSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 8 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/bin/compute_audio_cmvn.py b/funasr/bin/compute_audio_cmvn.pyindex ffad65272..6282e707c 100644--- a/funasr/bin/compute_audio_cmvn.py+++ b/funasr/bin/compute_audio_cmvn.py@@ -28,7 +28,7 @@ def main_hydra(kwargs: DictConfig):def main(**kwargs):print(kwargs)# set random seed- tables.print()+ # tables.print()set_all_random_seed(kwargs.get("seed", 0))torch.backends.cudnn.enabled = kwargs.get("cudnn_enabled", torch.backends.cudnn.enabled)torch.backends.cudnn.benchmark = kwargs.get("cudnn_benchmark", torch.backends.cudnn.benchmark)@@ -54,21 +54,14 @@ def main(**kwargs):dataset_train = dataset_class(kwargs.get("train_data_set_list"), frontend=frontend, tokenizer=None, is_training=False, **kwargs.get("dataset_conf"))# dataloader- batch_sampler = kwargs["dataset_conf"].get("batch_sampler", "DynamicBatchLocalShuffleSampler")- batch_sampler_train = None- if batch_sampler is not None:- batch_sampler_class = tables.batch_sampler_classes.get(batch_sampler)- dataset_conf = kwargs.get("dataset_conf")- dataset_conf["batch_type"] = "example"- dataset_conf["batch_size"] = 1- batch_sampler_train = batch_sampler_class(dataset_train, is_training=False, **dataset_conf)--- dataloader_train = torch.utils.data.DataLoader(dataset_train,- collate_fn=dataset_train.collator,- batch_sampler=batch_sampler_train,- num_workers=int(kwargs.get("dataset_conf").get("num_workers", 4)),- pin_memory=True)+ batch_sampler = kwargs["dataset_conf"].get("batch_sampler", "BatchSampler")+ batch_sampler_class = tables.batch_sampler_classes.get(batch_sampler)+ dataset_conf = kwargs.get("dataset_conf")+ dataset_conf["batch_type"] = "example"+ dataset_conf["batch_size"] = 1+ batch_sampler_train = batch_sampler_class(dataset_train, is_training=False, **dataset_conf)++ dataloader_train = torch.utils.data.DataLoader(dataset_train, collate_fn=dataset_train.collator, **batch_sampler_train)iter_stop = int(kwargs.get("scale", 1.0)*len(dataloader_train))diff --git a/funasr/bin/train.py b/funasr/bin/train.pyindex 3f97f9efc..1db37dd46 100644--- a/funasr/bin/train.py+++ b/funasr/bin/train.py@@ -148,7 +148,7 @@ def main(**kwargs):# datasetlogging.info("Build dataloader")- dataloader_class = tables.dataloader_classes.get( kwargs["dataset_conf"].get("dataloader", "DataloaderMapStyle"))+ dataloader_class = tables.dataloader_classes.get(kwargs["dataset_conf"].get("dataloader", "DataloaderMapStyle"))dataloader_tr, dataloader_val = dataloader_class(**kwargs)trainer = Trainer(local_rank=local_rank,diff --git a/funasr/datasets/audio_datasets/jsonl2scp.py b/funasr/datasets/audio_datasets/jsonl2scp.pyindex 9a2b023a3..b2d1703d0 100644--- a/funasr/datasets/audio_datasets/jsonl2scp.py+++ b/funasr/datasets/audio_datasets/jsonl2scp.py@@ -40,6 +40,7 @@ def gen_scp_from_jsonl(jsonl_file, data_type_list, wav_scp_file, text_file):def main_hydra(cfg: DictConfig):kwargs = OmegaConf.to_container(cfg, resolve=True)+ print(kwargs)scp_file_list = kwargs.get("scp_file_list", ("/Users/zhifu/funasr1.0/test_local/wav.scp", "/Users/zhifu/funasr1.0/test_local/text.txt"))if isinstance(scp_file_list, str):diff --git a/funasr/datasets/audio_datasets/scp2jsonl.py b/funasr/datasets/audio_datasets/scp2jsonl.pyindex 00754ddb6..a53712786 100644--- a/funasr/datasets/audio_datasets/scp2jsonl.py+++ b/funasr/datasets/audio_datasets/scp2jsonl.py@@ -79,6 +79,7 @@ def parse_context_length(data_list: list, data_type: str):def main_hydra(cfg: DictConfig):kwargs = OmegaConf.to_container(cfg, resolve=True)+ print(kwargs)scp_file_list = kwargs.get("scp_file_list", ("/Users/zhifu/funasr1.0/test_local/wav.scp", "/Users/zhifu/funasr1.0/test_local/text.txt"))if isinstance(scp_file_list, str):diff --git a/funasr/datasets/dataloader_entry.py b/funasr/datasets/dataloader_entry.pyindex 21e38341d..a1e4da2fb 100644--- a/funasr/datasets/dataloader_entry.py+++ b/funasr/datasets/dataloader_entry.py@@ -13,7 +13,7 @@ def DataloaderMapStyle(frontend=None, tokenizer=None, **kwargs):dataset_val = dataset_class(kwargs.get("valid_data_set_list"), frontend=frontend, tokenizer=tokenizer, is_training=False, **kwargs.get("dataset_conf"))# dataloader- batch_sampler = kwargs["dataset_conf"].get("batch_sampler", "DynamicBatchLocalShuffleSampler")+ batch_sampler = kwargs["dataset_conf"].get("batch_sampler", "BatchSampler")batch_sampler_val = Noneif batch_sampler is not None:batch_sampler_class = tables.batch_sampler_classes.get(batch_sampler)diff --git a/funasr/models/bicif_paraformer/template.yaml b/funasr/models/bicif_paraformer/template.yamlindex d2b0e0a48..db7ce5555 100644--- a/funasr/models/bicif_paraformer/template.yaml+++ b/funasr/models/bicif_paraformer/template.yaml@@ -112,7 +112,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/branchformer/template.yaml b/funasr/models/branchformer/template.yamlindex 09b1e0f4f..0fec4976a 100644--- a/funasr/models/branchformer/template.yaml+++ b/funasr/models/branchformer/template.yaml@@ -94,7 +94,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/conformer/template.yaml b/funasr/models/conformer/template.yamlindex f646acc9d..5a8687ecd 100644--- a/funasr/models/conformer/template.yaml+++ b/funasr/models/conformer/template.yaml@@ -95,7 +95,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/contextual_paraformer/template.yaml b/funasr/models/contextual_paraformer/template.yamlindex fbd15cea9..22052500c 100644--- a/funasr/models/contextual_paraformer/template.yaml+++ b/funasr/models/contextual_paraformer/template.yaml@@ -108,7 +108,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/e_branchformer/template.yaml b/funasr/models/e_branchformer/template.yamlindex dea580e52..7b0b27c76 100644--- a/funasr/models/e_branchformer/template.yaml+++ b/funasr/models/e_branchformer/template.yaml@@ -94,7 +94,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/monotonic_aligner/template.yaml b/funasr/models/monotonic_aligner/template.yamlindex b1379de78..f8d5ded7d 100644--- a/funasr/models/monotonic_aligner/template.yaml+++ b/funasr/models/monotonic_aligner/template.yaml@@ -93,7 +93,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/paraformer/template.yaml b/funasr/models/paraformer/template.yamlindex bccf63871..7809457df 100644--- a/funasr/models/paraformer/template.yaml+++ b/funasr/models/paraformer/template.yaml@@ -100,7 +100,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/paraformer_streaming/template.yaml b/funasr/models/paraformer_streaming/template.yamlindex d1300ac79..889971ad1 100644--- a/funasr/models/paraformer_streaming/template.yaml+++ b/funasr/models/paraformer_streaming/template.yaml@@ -121,7 +121,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/sanm/template.yaml b/funasr/models/sanm/template.yamlindex 156926f2c..a7f7b12d7 100644--- a/funasr/models/sanm/template.yaml+++ b/funasr/models/sanm/template.yaml@@ -98,7 +98,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/scama/template.yaml b/funasr/models/scama/template.yamlindex f647a9222..214046e2d 100644--- a/funasr/models/scama/template.yaml+++ b/funasr/models/scama/template.yaml@@ -104,7 +104,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/seaco_paraformer/template.yaml b/funasr/models/seaco_paraformer/template.yamlindex ab2301a32..fcaf5243d 100644--- a/funasr/models/seaco_paraformer/template.yaml+++ b/funasr/models/seaco_paraformer/template.yaml@@ -134,7 +134,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/transformer/template.yaml b/funasr/models/transformer/template.yamlindex 87814dc3b..eb597f9ea 100644--- a/funasr/models/transformer/template.yaml+++ b/funasr/models/transformer/template.yaml@@ -88,7 +88,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,diff --git a/funasr/models/uniasr/template.yaml b/funasr/models/uniasr/template.yamlindex 35c6b2eb2..e72a2d527 100644--- a/funasr/models/uniasr/template.yaml+++ b/funasr/models/uniasr/template.yaml@@ -184,7 +184,7 @@ scheduler_conf:dataset: AudioDatasetdataset_conf:index_ds: IndexDSJsonl- batch_sampler: DynamicBatchLocalShuffleSampler+ batch_sampler: BatchSamplerbatch_type: example # example or lengthbatch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,