julius-speech / julius

Open-Source Large Vocabulary Continuous Speech Recognition Engine
BSD 3-Clause "New" or "Revised" License
1.85k stars 300 forks source link

WARNING: 00 _default: no tail silence word survived on the last frame, search failed #111

Open Jake-s16 opened 5 years ago

Jake-s16 commented 5 years ago

Just installed following the instructions. I get to point 4. test mozilla.wav. I see lots of "WARNING: 00 _default: no tail silence word survived on the last frame, search failed" when I try to run ~/Software/julius-master/julius/./julius -C julius.jconf -dnnconf dnn.jconf in my ~/Software/ENVR-v5.4.Dnn.Bin folder

Estalhun commented 5 years ago

I think the wave is too long. Try to cut into pieces shorter than 20 sec.

Jake-s16 notifications@github.com ezt írta (időpont: 2019. júl. 11., Cs 15:56):

Just installed following the instructions. I get to point 4. test mozilla.wav. I see lots of "WARNING: 00 _default: no tail silence word survived on the last frame, search failed" when I try to run ~/Software/julius-master/julius/./julius -C julius.jconf -dnnconf dnn.jconf in my ~/Software/ENVR-v5.4.Dnn.Bin folder

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/julius-speech/julius/issues/111?email_source=notifications&email_token=AHNHTWUYJXJSKCMOVG5HNBLP643Y5A5CNFSM4IBHEMDKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4G6UPCWQ, or mute the thread https://github.com/notifications/unsubscribe-auth/AHNHTWQXFXZBIQ37NESBEM3P643Y5ANCNFSM4IBHEMDA .

Jake-s16 commented 5 years ago

I agree but that really doesn't make sense. Why would someone package that wav file to be used if it was too long?

kingbeowulf commented 5 years ago

Same error here trying to test release 4.5 for distro packaging (BTW, needing <=20s speech segments would be very silly).

Following the Quick Start and README.md in ENVR-v5.4.Dnn.Bin.zip, I see a whole string of WARNING: 00 _default: no tail silence word survived on the last frame, search failed

errors. I compiled with "configure --enable-words-int --enable-sp-segment" I've attached the output. [test.txt](https://github.com/julius-speech/julius/files/3737529/test.txt)
kingbeowulf commented 5 years ago

Same error here trying to test release 4.5 for distro packaging (BTW, needing <=20s speech segments would be very silly).

Following the Quick Start and README.md in ENVR-v5.4.Dnn.Bin.zip, I see a whole string of WARNING: 00 _default: no tail silence word survived on the last frame, search failed

errors.

Turns out ENVR-v5.4.Gmm.Bin.zip is a better choice to test and use release 4.5.

programmin1 commented 5 years ago

I just got the same thing, when i set up in a new directory referencing the download on which I hadn't made the change to dnn.jconf:

feature_type MFCC_E_D_A_Z
feature_options -htkconf wav_config -cvn -cmnload ENVR-v5.3.norm -cvnstatic
num_threads 1
feature_len 48
context_len 11
input_nodes 528
output_nodes 7461
hidden_nodes 1536
hidden_layers 5
W1 ENVR-v5.3.layer2_weight.npy
W2 ENVR-v5.3.layer3_weight.npy
W3 ENVR-v5.3.layer4_weight.npy
W4 ENVR-v5.3.layer5_weight.npy
W5 ENVR-v5.3.layer6_weight.npy
B1 ENVR-v5.3.layer2_bias.npy
B2 ENVR-v5.3.layer3_bias.npy
B3 ENVR-v5.3.layer4_bias.npy
B4 ENVR-v5.3.layer5_bias.npy
B5 ENVR-v5.3.layer6_bias.npy
output_W ENVR-v5.3.layerout_weight.npy
output_B ENVR-v5.3.layerout_bias.npy
state_prior_factor 1.0
state_prior ENVR-v5.3.prior
state_prior_log10nize false