updated custom_model now managed by session, fixed various bugs

[ROBERT-MCDOWELL]

Check if the issue you had from your last test is solved with this PR

[DrewThomasson]

Kk gimmi a sec...

[DrewThomasson]

yeah it works again but the hashes are identical for the two models again tho

the David attenburo and the std xtts model

as you can see here :/

Last login: Sat Dec 21 19:04:40 on ttys005
(base) drew@wmughal-CN4D09397T ~ % cmp -b /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--xtts_v2/model.pth /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--drewThomasson/fineTunedTTSModels/xtts-v2/eng/DavidAttenborough/model.pth

(base) drew@wmughal-CN4D09397T ~ % cmp -l /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--xtts_v2/model.pth /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--drewThomasson/fineTunedTTSModels/xtts-v2/eng/DavidAttenborough/model.pth

(base) drew@wmughal-CN4D09397T ~ % ls -l /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--xtts_v2/model.pth /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--drewThomasson/fineTunedTTSModels/xtts-v2/eng/DavidAttenborough/model.pth

-rw-r--r--  1 drew  staff  1867929118 Dec 21 19:11 /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--drewThomasson/fineTunedTTSModels/xtts-v2/eng/DavidAttenborough/model.pth
-rw-r--r--  1 drew  staff  1867929118 Dec 21 19:07 /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--xtts_v2/model.pth
(base) drew@wmughal-CN4D09397T ~ % shasum -a 256 /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--xtts_v2/model.pth /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--drewThomasson/fineTunedTTSModels/xtts-v2/eng/DavidAttenborough/model.pth

c7ea20001c6a0a841c77e252d8409f6a74fb423e79b3206a0771ba5989776187  /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--xtts_v2/model.pth
c7ea20001c6a0a841c77e252d8409f6a74fb423e79b3206a0771ba5989776187  /Users/drew/ebook2audiobook/models/tts/tts_models--multilingual--multi-dataset--drewThomasson/fineTunedTTSModels/xtts-v2/eng/DavidAttenborough/model.pth
(base) drew@wmughal-CN4D09397T ~ % 

[DrewThomasson]

I think were just going to have to download the model files like this,

from huggingface_hub import hf_hub_download

# Specify the repo ID and subdirectory
repo_id = "username/repo-name"
subdir = "subfolder_name"
files = ["model.pth", "vocab.json", "config.json"]

# Download files from the subdirectory
for file_name in files:
    file_path = f"{subdir}/{file_name}"
    local_file_path = hf_hub_download(repo_id=repo_id, filename=file_path)
    print(f"Downloaded {file_name} to {local_file_path}")

[DrewThomasson]

seems easier and still flexible as you can have all the data about every model in the config.py file?

:/ :)

[DrewThomasson]

Example for David Attenburo as well as the standard model

from huggingface_hub import hf_hub_download

# Specify the repo ID and subdirectory
repo_id = "drewThomasson/fineTunedTTSModels"
subdir = "xtts-v2/eng/DavidAttenborough"
files = ["model.pth", "vocab.json", "config.json"]

# Download files from the subdirectory
for file_name in files:
    file_path = f"{subdir}/{file_name}"
    local_file_path = hf_hub_download(repo_id=repo_id, filename=file_path)
    print(f"Downloaded {file_name} to {local_file_path}")

[DrewThomasson]

I can try to implement it on my end if you want

[DrewThomasson]

Once we get this working I'll be able to merge it cause the readme is pretty much set along with the other winning features of v2.0
:)

[agentxan]

When I try to use the --headless option I get the following error.

A subdirectory or file C:\Windows\Temp\Calibre already exists.
processed file: C:\Windows\Temp\Calibre
Successfully processed 1 files; Failed processing 0 files
Running in native mode
Package requirements ok
Dictionary ok
convert_ebook() Exception: 'Namespace' object does not support item assignment
Conversion failed: 'Namespace' object does not support item assignment

[ROBERT-MCDOWELL]

1. issue: from gui/headless? are you using David from fine-tuned or custom model?

[agentxan]

if I use gui it works.
ebook2audiobook.cmd --headless --device gpu --ebook D:\ebooktest\test_eng.azw3

I look and it seems to be where the args are passed to convert_ebook
if convert_ebook(args) is changed to convert_ebook(vars(args)) it seems to work but then get

convert_ebook() Exception: The fine tuned model does not exist.
Conversion failed: The fine tuned model does not exist.

It looks like these two options are reversed

parser.add_argument(options[16], type=str, default=default_fine_tuned,
                    help='Name of the fine tuned model. Optional, uses the standard model according to the TTS engine and language.')
parser.add_argument(options[17], action='store_true',
                    help='Enable splitting text into sentences. Default to False.')

If those are swapped then the headless command works

[ROBERT-MCDOWELL]

ok I fix it