From 66453b81d38e250aabfbb82b2433c5f5d504653e Mon Sep 17 00:00:00 2001 From: Eren Golge Date: Tue, 14 May 2019 17:49:08 +0200 Subject: [PATCH] config update --- .compute | 2 +- config_cluster.json | 34 ++++++++++++++++++---------------- 2 files changed, 19 insertions(+), 17 deletions(-) diff --git a/.compute b/.compute index d2d37fc1..8e033754 100644 --- a/.compute +++ b/.compute @@ -9,6 +9,6 @@ pip3 install https://download.pytorch.org/whl/cu100/torch-1.0.1.post2-cp36-cp36m wget https://www.dropbox.com/s/wqn5v3wkktw9lmo/install.sh?dl=0 -O install.sh sudo sh install.sh python3 setup.py develop -python3 distribute.py --config_path config_cluster.json --data_path ${USER_DIR}/MozillaAll2/Mozilla/ --restore_path ${USER_DIR}/checkpoint_123000_4761.pth.tar +python3 distribute.py --config_path config_cluster.json --data_path ${USER_DIR}/MozillaAll2/Mozilla/ # python3 distribute.py --config_path config_cluster.json --data_path ${SHARED_DIR}/data/mozilla/Judy/ # while true; do sleep 1000000; done diff --git a/config_cluster.json b/config_cluster.json index 11c2415f..b04c4df1 100644 --- a/config_cluster.json +++ b/config_cluster.json @@ -1,6 +1,6 @@ { - "run_name": "mozilla-fattn", - "run_description": "Finetune 4761 with BN + Dropout. It is to compare to 4780 and see how dropout behaves with BN.", + "run_name": "mozilla-no-loc", + "run_description": "using Bahdenau attention, with original prenet.", "audio":{ // Audio processing parameters @@ -31,22 +31,24 @@ "reinit_layers": [], - "model": "Tacotron2", // one of the model in models/ - "grad_clip": 1, // upper limit for gradients for clipping. - "epochs": 1000, // total number of epochs to train. - "lr": 0.0001, // Initial learning rate. If Noam decay is active, maximum learning rate. - "lr_decay": false, // if true, Noam learning rate decaying is applied through training. - "warmup_steps": 4000, // Noam decay steps to increase the learning rate from 0 to "lr" - "windowing": false, // Enables attention windowing. Used only in eval mode. - "memory_size": 5, // ONLY TACOTRON - memory queue size used to queue network predictions to feed autoregressive connection. Useful if r < 5. + "model": "Tacotron2", // one of the model in models/ + "grad_clip": 1, // upper limit for gradients for clipping. + "epochs": 1000, // total number of epochs to train. + "lr": 0.0002, // Initial learning rate. If Noam decay is active, maximum learning rate. + "lr_decay": false, // if true, Noam learning rate decaying is applied through training. + "warmup_steps": 4000, // Noam decay steps to increase the learning rate from 0 to "lr" + "windowing": false, // Enables attention windowing. Used only in eval mode. + "memory_size": 5, // ONLY TACOTRON - memory queue size used to queue network predictions to feed autoregressive connection. Useful if r < 5. "attention_norm": "softmax", // softmax or sigmoid. Suggested to use softmax for Tacotron2 and sigmoid for Tacotron. - "prenet_type": "bn", // ONLY TACOTRON2 - "original" or "bn". - "prenet_dropout": true, // ONLY TACOTRON2 - enable/disable dropout at prenet. - "use_forward_attn": true, // ONLY TACOTRON2 - if it uses forward attention. In general, it aligns faster. - "transition_agent": false, // ONLY TACOTRON2 - enable/disable transition agent of forward attention. - "location_attn": false, // ONLY TACOTRON2 - enable_disable location sensitive attention. It is enabled for TACOTRON by default. - "loss_masking": false, // enable / disable loss masking against the sequence padding. + "prenet_type": "original", // ONLY TACOTRON2 - "original" or "bn". + "prenet_dropout": true, // ONLY TACOTRON2 - enable/disable dropout at prenet. + "use_forward_attn": false, // ONLY TACOTRON2 - if it uses forward attention. In general, it aligns faster. + "transition_agent": false, // ONLY TACOTRON2 - enable/disable transition agent of forward attention. + "location_attn": false, // ONLY TACOTRON2 - enable_disable location sensitive attention. It is enabled for TACOTRON by default. + "loss_masking": false, // enable / disable loss masking against the sequence padding. "enable_eos_bos_chars": false, // enable/disable beginning of sentence and end of sentence chars. + "stopnet": false, // Train stopnet predicting the end of synthesis. + "separate_stopnet": false, // train stopnet seperately. It prevents stopnet loss to influence the rest of the model. It causes a better model, but it trains SLOWER. "batch_size": 32, // Batch size for training. Lower values than 32 might cause hard to learn attention. "eval_batch_size":16,