having issues with read.py - RuntimeError: stack expects a non-empty TensorList

[embanot]

when I try to run the read.py script to generate longer voice samples from a text file, I get this error message: "RuntimeError: stack expects a non-empty TensorList"

I tried googling this but couldn't find a viable solution. Any idea what's going on? It works fine when just generating short samples with the normal command.

Here is the full output from Anaconda when running the read.py script: Traceback (most recent call last): File "tortoise/read.py", line 62, in gen = tts.tts_with_preset(text, voice_samples=voice_samples, conditioning_latents=conditioning_latents, File "E:\tortoise-tts\tortoise\api.py", line 328, in tts_with_preset return self.tts(text, **settings) File "E:\tortoise-tts\tortoise\api.py", line 392, in tts auto_conditioning, diffusion_conditioning, auto_conds, _ = self.get_conditioning_latents(voice_samples, return_mels=True) File "E:\tortoise-tts\tortoise\api.py", line 274, in get_conditioning_latents auto_conds = torch.stack(auto_conds, dim=1) RuntimeError: stack expects a non-empty TensorList

[plincter10]

I found that it will accept an address in place of the actual file, C:\Users\ricke\Modelpics\Hayley\speech\speech.txt. without the <>. One page of story took 3hours on my 3060 12 gig and outputted a bunch of single wav files instead of 3 when it reads a line. In fact, it will read 3 lines before it chokes using the line script. I'm wondering where all the commands are other than the couple in the readme. It references something about "These settings are not available in the normal scripts packaged with Tortoise. They are available, however, in the API. See api.tts for a full list." Where is this file and list?