ability to generate in seconds

README.md

@@ -95,14 +95,15 @@ loss.backward()
 
 # and now you can generate state-of-the-art speech
 
-generated_audio = model.generate(1024, batch_size = 2)  # generated audio is also a raw wave if soundstream is present
+generated_audio = model.generate(seconds = 30, batch_size = 2)  # generate 30 seconds of audio (it will calculate the length in seconds based off the sampling frequency and cumulative downsamples in the soundstream passed in above)
 ```
 
 ## Todo
 
 - [x] integrate soundstream
+- [x] when generating, and length can be defined in seconds (takes into sampling freq etc)
 
-- [ ] when generating, make sure it can return audio file, and length can be defined in seconds (takes into sampling freq etc)
+- [ ] option to return list of audio files when generating
 - [ ] turn it into a command line tool
 - [ ] add cross attention and adaptive layernorm conditioning (just copy paste in the entire conformer repository, if conditioning adds too much cruft to the other repo)
 - [ ] make sure grouped rvq is supported. concat embeddings rather than sum across group dimension

setup.py

@@ -3,7 +3,7 @@
 setup(
   name = 'soundstorm-pytorch',
   packages = find_packages(exclude=[]),
-  version = '0.0.6',
+  version = '0.0.7',
   license='MIT',
   description = 'SoundStorm - Efficient Parallel Audio Generation from Google Deepmind, in Pytorch',
   author = 'Phil Wang',

soundstorm_pytorch/soundstorm.py

@@ -276,13 +276,21 @@ def __init__(
     @torch.no_grad()
     def generate(
         self,
-        max_seq_len,
+        seq_len = None,
+        *,
+        seconds = None,
         batch_size = None,
         start_temperature = 1.,
         filter_thres = 0.7,
         noise_level_scale = 1.,
         **kwargs
     ):
+        assert exists(seq_len) ^ exists(seconds)
+
+        if not exists(seq_len):
+            assert exists(self.soundstream), 'soundstream must be passed in to generate in seconds'
+            seq_len = (seconds * self.soundstream.target_sample_hz) //  self.soundstream.seq_len_multiple_of
+
         sample_one = not exists(batch_size)
         batch_size = default(batch_size, 1)
 
@@ -295,14 +303,14 @@ def generate(
 
         # sequence starts off as all masked
 
-        shape = (batch_size, max_seq_len)
+        shape = (batch_size, seq_len)
 
         seq = torch.full(shape, self.mask_id, device = device)
         mask = torch.full(shape, True, device = device)
 
         # slowly demask
 
-        all_mask_num_tokens = (self.schedule_fn(times[1:]) * max_seq_len).long()
+        all_mask_num_tokens = (self.schedule_fn(times[1:]) * seq_len).long()
 
         # self conditioning