Download Model #63
-
|
How to download the models and load it offline |
Beta Was this translation helpful? Give feedback.
Replies: 12 comments 28 replies
-
|
Running the script the first time for a model will download that specific model; it stores (on windows) the model at Once downloaded, the model doesn't need to be downloaded again. Considering that the medium model alone is ~1.5 GB, that seems to be the best solution. |
Beta Was this translation helpful? Give feedback.
-
|
Can I change the download directory or the dir the model is loaded from? I dont know python... |
Beta Was this translation helpful? Give feedback.
-
|
@Theo1996 If you're using the command line, you can specify |
Beta Was this translation helpful? Give feedback.
-
|
@jongwook --model_dirdoesnt work |
Beta Was this translation helpful? Give feedback.
-
|
@jongwook I needed the latest commit, now it works. |
Beta Was this translation helpful? Give feedback.
-
|
Can also use the command line option --model_dir to save models to a specific folder |
Beta Was this translation helpful? Give feedback.
-
|
Download links are in init.py @ lines 17-27 For windows location to download see the above comment, on WSL/Linux they go in Note if you are experimenting with both WSL and Windows Native versions you need to put the models in BOTH locations. As of today those links are: |
Beta Was this translation helpful? Give feedback.
-
Hum, ok ! But more strange enough, on my side i got better results with the multilingual pack for transcribe English than the dedicate English one ... |
Beta Was this translation helpful? Give feedback.
-
Hi sir, i've downloaded this but i get this error: |
Beta Was this translation helpful? Give feedback.
-
|
i download large-v2 model,but i use this model. |
Beta Was this translation helpful? Give feedback.
-
|
@navakelvin I used the link from |
Beta Was this translation helpful? Give feedback.
-
|
hello i download the model small.en.pt but it didn't calculate the wer metric is there any idea to add it ? |
Beta Was this translation helpful? Give feedback.
-
|
to find the URL examine MODELS: site-packages/whisper/init.py use that URL in the cache dir: .cache/whisper % curl -L -O -C - … |
Beta Was this translation helpful? Give feedback.
-
|
It would be nice to have the option to download the model like Spacy with |
Beta Was this translation helpful? Give feedback.
-
|
Some code is required, but it can be done: import sys
from whisper import _download, _MODELS
_download(_MODELS["base"], "/models/", False)This will also make sure the models are downloaded correctly or already exist in the given model directory in the correct version. |
Beta Was this translation helpful? Give feedback.
-
|
To download all models at once: import sys
from whisper import _download, _MODELS
models = ["tiny.en", "tiny", "base.en", "base", "small.en", "small", "medium.en", "medium", "large"]
for model in models:
_download(_MODELS[model], "~/.cache/whisper", False) |
Beta Was this translation helpful? Give feedback.
-
|
Tweaked @mayeaux's useful code so that it gets the entire list direct from whisper (more future-proof in case list changes) import sys
from whisper import _download, _MODELS
models = whisper.available_models()
for model in models:
_download(_MODELS[model], "~/.cache/whisper", False) |
Beta Was this translation helpful? Give feedback.
-
|
How do I load this downloaded model? |
Beta Was this translation helpful? Give feedback.
-
|
i wonder why the tiny downlaoded model is 75.6 MB in size and not 151 MB |
Beta Was this translation helpful? Give feedback.
-
|
For all those wondering for Mac You can see your 😛 I used this opportunity to clean out some other stuff too from my |
Beta Was this translation helpful? Give feedback.
-
|
How can I add a model to a docker image? Do I add a new folder with the model, like this; |
Beta Was this translation helpful? Give feedback.
-
|
Also wondering this. I'd like to download my models at build time of the docker container, then reference them in my app.py file from the download directory. |
Beta Was this translation helpful? Give feedback.
-
|
anyone figure this out ? |
Beta Was this translation helpful? Give feedback.
-
|
Came here for the same reason. |
Beta Was this translation helpful? Give feedback.
-
|
You can put the model to wherever you want and specify that path with "--model_dir PATH" for both the .exe and python versions. |
Beta Was this translation helpful? Give feedback.
-
|
Just import whisper and download a model during buildtime: |
Beta Was this translation helpful? Give feedback.
-
|
I have a question! It says "Once downloaded, the model doesn't need to be downloaded again.". But i have to download the model everytime i transcribe an audio.. is it because i have to run the program as administrator? |
Beta Was this translation helpful? Give feedback.
-
|
So you can change the download directory to some other directory |
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
|
You must change the init.py in "whisper-main\whisper" Change these segments in code os.makedirs(os.path.join(root, "your Path"), exist_ok=True) in the folowing code is not "your Path" it is "models" " import hashlib import torch from .audio import load_audio, log_mel_spectrogram, pad_or_trim _MODELS = { base85-encoded (n_layers, n_heads) boolean arrays indicating the cross-attention heads that arehighly correlated to the word-level timing, i.e. the alignment between audio and text tokens.ALIGNMENT_HEADS = { def _download(url: str, root: str, in_memory: bool) -> Union[bytes, str]: def available_models() -> List[str]: def load_model( " |
Beta Was this translation helpful? Give feedback.
-
|
hello thank you for your response, now i need a model.pt can do
transcription and calculate wer but all models i download can not give the
wer do you have any idea or the links of models you sent can do that ?
thank you so much for your help.
…On Thu, Jan 4, 2024 at 12:00 PM Aikon404 ***@***.***> wrote:
You must change the *init*.py in "whisper-main\whisper"
Change these segments in code
os.makedirs(os.path.join(root, "your Path"), exist_ok=True)
download_target = os.path.join(root, "your Path", file_name)
default = os.path.join(os.path.dirname(os.path.abspath(__file__)), "your Path")
in the folowing code is not "your Path" it is "models"
"
import hashlib
import io
import os
import urllib
import warnings
from typing import List, Optional, Union
import torch
from tqdm import tqdm
from .audio import load_audio, log_mel_spectrogram, pad_or_trim
from .decoding import DecodingOptions, DecodingResult, decode,
detect_language
from .model import ModelDimensions, Whisper
from .transcribe import transcribe
from .version import *version*
_MODELS = {
"tiny.en": "
https://openaipublic.azureedge.net/main/whisper/models/d3dd57d32accea0b295c96e26691aa14d8822fac7d9d27d5dc00b4ca2826dd03/tiny.en.pt
",
"tiny": "
https://openaipublic.azureedge.net/main/whisper/models/65147644a518d12f04e32d6f3b26facc3f8dd46e5390956a9424a650c0ce22b9/tiny.pt
",
"base.en": "
https://openaipublic.azureedge.net/main/whisper/models/25a8566e1d0c1e2231d1c762132cd20e0f96a85d16145c3a00adf5d1ac670ead/base.en.pt
",
"base": "
https://openaipublic.azureedge.net/main/whisper/models/ed3a0b6b1c0edf879ad9b11b1af5a0e6ab5db9205f891f668f8b0e6c6326e34e/base.pt
",
"small.en": "
https://openaipublic.azureedge.net/main/whisper/models/f953ad0fd29cacd07d5a9eda5624af0f6bcf2258be67c92b79389873d91e0872/small.en.pt
",
"small": "
https://openaipublic.azureedge.net/main/whisper/models/9ecf779972d90ba49c06d968637d720dd632c55bbf19d441fb42bf17a411e794/small.pt
",
"medium.en": "
https://openaipublic.azureedge.net/main/whisper/models/d7440d1dc186f76616474e0ff0b3b6b879abc9d1a4926b7adfa41db2d497ab4f/medium.en.pt
",
"medium": "
https://openaipublic.azureedge.net/main/whisper/models/345ae4da62f9b3d59415adc60127b97c714f32e89e936602e85993674d08dcb1/medium.pt
",
"large-v1": "
https://openaipublic.azureedge.net/main/whisper/models/e4b87e7e0bf463eb8e6956e646f1e277e901512310def2c24bf0e11bd3c28e9a/large-v1.pt
",
"large-v2": "
https://openaipublic.azureedge.net/main/whisper/models/81f7c96c852ee8fc832187b0132e569d6c3065a3252ed18e56effd0b6a73e524/large-v2.pt
",
"large-v3": "
https://openaipublic.azureedge.net/main/whisper/models/e5b1a55b89c1367dacf97e3e19bfd829a01529dbfdeefa8caeb59b3f1b81dadb/large-v3.pt
",
"large": "
https://openaipublic.azureedge.net/main/whisper/models/e5b1a55b89c1367dacf97e3e19bfd829a01529dbfdeefa8caeb59b3f1b81dadb/large-v3.pt
",
}
base85-encoded (n_layers, n_heads) boolean arrays indicating the
cross-attention heads that are highly correlated to the word-level
timing, i.e. the alignment between audio and text tokens.
*ALIGNMENT_HEADS = { "tiny.en": ***@***.***
<https://github.com/0>{>%R00Bk>$p{7v037oCl~+#00", "tiny":
***@***.***=7ytkO", "base.en":
b"ABzY8;40c<0{>%RzzG;p*o+Vo09|#PsxSZm00", "base":
b"ABzY8KQ!870{>%RzyTQH3Q^yNP!>##QT-<FaQ7m", "small.en": b"ABzY8>?*
)10{>%RpeA61k&I|OI3I$65C{;;pbCHh0B{qLQ;+}v00",
"small": b"ABzY8DmU6=0{>%Rpa?J`kvJ6qF(V^F86#Xh7JUGMK}P<N0000",
"medium.en": b"ABzY8usPae0{>%R7<zz_OvQ{)4kMa0BMw6u5rT}kRKX;$NfYBv00
***@***.***", "medium": b"ABzY8B0Jh+0{>%R7}kK1fFL7w6%<-Pf*t^=N)Qr&0RR9",
"large-v1": b"ABzY8r9j$a0{>%R7#4sLmoOs{s)o384-RPdcFk!JR<kSfC2yj",
"large-v2":
b"ABzY8zd+h!0{>%R7=D0pU<_bnWW*tkYAhobTNnu$jnkEkXqp)j;w1Tzk)UH3X%SZd&fFZ2fC2yj",
"large-v3": b"ABzY8gWO1E0{>%R7(9S+Kn!D%ngiGaR?*L!iJG9p-nab0JQ=-{D1-g00",
"large": b"ABzY8gWO1E0{>%R7(9S+Kn!D~%ngiGaR?*L!iJG9p-nab0JQ=-{D1-g00",
}
def _download(url: str, root: str, in_memory: bool) -> Union[bytes, str]:
os.makedirs(os.path.join(root, "models"), exist_ok=True)
expected_sha256 = url.split("/")[-2]
file_name = os.path.basename(url)
download_target = os.path.join(root, "models", file_name)
if os.path.exists(download_target) and not os.path.isfile(download_target):
raise RuntimeError(f"{download_target} exists and is not a regular file")
if os.path.isfile(download_target):
with open(download_target, "rb") as f:
model_bytes = f.read()
if hashlib.sha256(model_bytes).hexdigest() == expected_sha256:
return model_bytes if in_memory else download_target
else:
warnings.warn(
f"{download_target} exists, but the SHA256 checksum does not match; re-downloading the file"
)
with urllib.request.urlopen(url) as source, open(download_target, "wb") as output:
with tqdm(
total=int(source.info().get("Content-Length")),
ncols=80,
unit="iB",
unit_scale=True,
unit_divisor=1024,
) as loop:
while True:
buffer = source.read(8192)
if not buffer:
break
output.write(buffer)
loop.update(len(buffer))
model_bytes = open(download_target, "rb").read()
if hashlib.sha256(model_bytes).hexdigest() != expected_sha256:
raise RuntimeError(
"Model has been downloaded but the SHA256 checksum does not match. Please retry loading the model."
)
return model_bytes if in_memory else download_target
def available_models() -> List[str]:
"""Returns the names of available models"""
return list(_MODELS.keys())
def load_model(
name: str,
device: Optional[Union[str, torch.device]] = None,
download_root: str = None,
in_memory: bool = False,
) -> Whisper:
"""
Load a Whisper ASR model
Parameters
----------
name : str
one of the official model names listed by `whisper.available_models()`, or
path to a model checkpoint containing the model dimensions and the model state_dict.
device : Union[str, torch.device]
the PyTorch device to put the model into
download_root: str
path to download the model files; by default, it uses "~/.cache/whisper"
in_memory: bool
whether to preload the model weights into host memory
Returns
-------
model : Whisper
The Whisper ASR model instance
"""
if device is None:
device = "cuda" if torch.cuda.is_available() else "cpu"
if download_root is None:
default = os.path.join(os.path.dirname(os.path.abspath(__file__)), "models")
download_root = os.path.join(default, "whisper")
if name in _MODELS:
checkpoint_file = _download(_MODELS[name], download_root, in_memory)
alignment_heads = _ALIGNMENT_HEADS[name]
elif os.path.isfile(name):
checkpoint_file = open(name, "rb").read() if in_memory else name
alignment_heads = None
else:
raise RuntimeError(
f"Model {name} not found; available models = {available_models()}"
)
with (
io.BytesIO(checkpoint_file) if in_memory else open(checkpoint_file, "rb")
) as fp:
checkpoint = torch.load(fp, map_location=device)
del checkpoint_file
dims = ModelDimensions(**checkpoint["dims"])
model = Whisper(dims)
model.load_state_dict(checkpoint["model_state_dict"])
if alignment_heads is not None:
model.set_alignment_heads(alignment_heads)
return model.to(device)
"
—
Reply to this email directly, view it on GitHub
<#63 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AT3UJLFUILJ2QF5WW2AQVVDYM2DWTAVCNFSM6AAAAAAQTZ4VH2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4DAMJRGU3TM>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
no sorry, I only have the models that we got from openai |
Beta Was this translation helpful? Give feedback.
-
|
Ok, thanks anyway.
…On Thu, Jan 4, 2024 at 2:47 PM Aikon404 ***@***.***> wrote:
no sorry, I only have the models that we got from openai
—
Reply to this email directly, view it on GitHub
<#63 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AT3UJLBRLWGK34SUCMTNC6TYM2XHRAVCNFSM6AAAAAAQTZ4VH2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4DAMJSHE2DE>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
checkpoint = torch.load(fp, map_location=device) |
Beta Was this translation helpful? Give feedback.
and others
Running the script the first time for a model will download that specific model; it stores (on windows) the model at
C:\Users\<username>\.cache\whisper\<model>.Once downloaded, the model doesn't need to be downloaded again. Considering that the medium model alone is ~1.5 GB, that seems to be the best solution.