Seems stuck at ░░░░░░░░░░░░░░░░░░░░ [0/5] Fetching packages... for a long time

[dheerajck]

I am not sure if this is a bug or an ui/ux issue

uv pip install -r requirements.txt 
Resolved 426 packages in 164ms
░░░░░░░░░░░░░░░░░░░░ [0/5] Fetching packages...

its stuck here for a long time, this doesnt happen with pip install -r requirements.txt , pip downloads and install packages way faster but uv doesnt so this isn an internet speed issue

[dheerajck]

It would be great if we can have a better progress bar, and if we can make downloads faster(I think this has to do with some uv internals which handles downloads)

[zanieb]

Could you please share output with the -v flag as well as your requirements file?

[zanieb]

What does a long time mean in concrete terms?

[Mateleo]

Same for me, problem is when downloading Pytorch

[dheerajck]

I have similar messages in the logs,

I am even getting timeout error sometimes

error: Failed to download distributions
  Caused by: Failed to fetch wheel: nvidia-cudnn-cu12==8.9.2.26
  Caused by: Failed to extract archive
  Caused by: Failed to download distribution due to network timeout. Try increasing UV_HTTP_TIMEOUT (current value: 300s).

Can I know why do we have a TIMEOUT, does pip have something similar,

if progress bar can show something like this with the progress bar

░░░░░░░░░░░░░░░░░░░░ 10%, 10MBPS

it will be helpful

[zanieb]

We're going to need more information to reproduce this. Please see issues marked as great writeup for example.

[kevinhu]

I was running into this error too, so I tried vanilla pip to see if that could help diagnose the error. It looks like several of the Nvidia packages simply take really long to download and are exceeding the HTTP timeout. I ran the command on a GCP VM that has 500mbps+ bandwidth, but I was capped at 1mbps on a few of the downloads, which ended up taking several minutes to complete.

If uv could show a per-package download speed or process indicator for slow downloads, that would be super helpful for avoiding confusion. Right now, it's hard to tell if the timeout is failing because the download speed is too slow or if there is an error finding the distribution in the first place.

[charliermarsh]

Yeah the Nvidia packages are enormous, as is Torch (hundreds of megabytes). My best guess here is that we're running into issues by making too many parallel requests to the index, and thus being served any given request too slowly to complete under the timeout (or something like that). Can you share more on what you did to hit this? E.g., what was the requirements.txt? Were you using any extra indexes, or just PyPI?

[kevinhu]

I'm only using pypi—I encountered the timeout when installing torch alone:

uv venv
source .venv/bin/activate
uv pip install torch

[zanieb]

For reference, we have some issues related to this:

• #1209
• #1921

[dheerajck]

uv --version
uv 0.1.27

time uv pip install nvidia-cudnn-cu12==8.9.2.26 -v
INFO Found a virtualenv through VIRTUAL_ENV at: ~/.uvvenv
DEBUG Probing interpreter info for: ~/.uvvenv/bin/python
DEBUG Found Python 3.11.6 for: ~/.uvvenv/bin/python
DEBUG Using Python 3.11.6 environment at .uvvenv/bin/python
DEBUG Using registry request timeout of 300s
DEBUG Solving with target Python version 3.11.6
DEBUG Adding direct dependency: nvidia-cudnn-cu12==8.9.2.26
DEBUG No cache entry for: https://pypi.org/simple/nvidia-cudnn-cu12/
DEBUG No credentials found for: https://pypi.org/simple/nvidia-cudnn-cu12/
DEBUG Searching for a compatible version of nvidia-cudnn-cu12 (==8.9.2.26)
DEBUG Selecting: nvidia-cudnn-cu12==8.9.2.26 (nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl)
DEBUG No cache entry for: https://files.pythonhosted.org/packages/ff/74/a2e2be7fb83aaedec84f391f082cf765dfb635e7caa9b49065f73e4835d8/nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl.metadata
DEBUG No credentials found for: https://files.pythonhosted.org/packages/ff/74/a2e2be7fb83aaedec84f391f082cf765dfb635e7caa9b49065f73e4835d8/nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl.metadata
DEBUG Adding transitive dependency: nvidia-cublas-cu12*
DEBUG No cache entry for: https://pypi.org/simple/nvidia-cublas-cu12/
DEBUG No credentials found for already-seen URL: https://pypi.org/simple/nvidia-cublas-cu12/
DEBUG No credentials found for: https://pypi.org/simple/nvidia-cublas-cu12/
DEBUG Searching for a compatible version of nvidia-cublas-cu12 (*)
DEBUG Selecting: nvidia-cublas-cu12==12.4.2.65 (nvidia_cublas_cu12-12.4.2.65-py3-none-manylinux2014_x86_64.whl)
DEBUG No cache entry for: https://files.pythonhosted.org/packages/f2/44/fdfd1a9f54167714d2c3dbc1bb1cb030340d6fb936a1dab9ef60999e3f73/nvidia_cublas_cu12-12.4.2.65-py3-none-manylinux2014_x86_64.whl.metadata
DEBUG No credentials found for already-seen URL: https://files.pythonhosted.org/packages/f2/44/fdfd1a9f54167714d2c3dbc1bb1cb030340d6fb936a1dab9ef60999e3f73/nvidia_cublas_cu12-12.4.2.65-py3-none-manylinux2014_x86_64.whl.metadata
DEBUG No credentials found for: https://files.pythonhosted.org/packages/f2/44/fdfd1a9f54167714d2c3dbc1bb1cb030340d6fb936a1dab9ef60999e3f73/nvidia_cublas_cu12-12.4.2.65-py3-none-manylinux2014_x86_64.whl.metadata
Resolved 2 packages in 860ms
DEBUG Identified uncached requirement: nvidia-cublas-cu12==12.4.2.65
DEBUG Identified uncached requirement: nvidia-cudnn-cu12==8.9.2.26
DEBUG No cache entry for: https://files.pythonhosted.org/packages/ff/74/a2e2be7fb83aaedec84f391f082cf765dfb635e7caa9b49065f73e4835d8/nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl
DEBUG No credentials found for already-seen URL: https://files.pythonhosted.org/packages/ff/74/a2e2be7fb83aaedec84f391f082cf765dfb635e7caa9b49065f73e4835d8/nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl
DEBUG No credentials found for: https://files.pythonhosted.org/packages/ff/74/a2e2be7fb83aaedec84f391f082cf765dfb635e7caa9b49065f73e4835d8/nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl
DEBUG No cache entry for: https://files.pythonhosted.org/packages/f2/44/fdfd1a9f54167714d2c3dbc1bb1cb030340d6fb936a1dab9ef60999e3f73/nvidia_cublas_cu12-12.4.2.65-py3-none-manylinux2014_x86_64.whl
DEBUG No credentials found for already-seen URL: https://files.pythonhosted.org/packages/f2/44/fdfd1a9f54167714d2c3dbc1bb1cb030340d6fb936a1dab9ef60999e3f73/nvidia_cublas_cu12-12.4.2.65-py3-none-manylinux2014_x86_64.whl
DEBUG No credentials found for: https://files.pythonhosted.org/packages/f2/44/fdfd1a9f54167714d2c3dbc1bb1cb030340d6fb936a1dab9ef60999e3f73/nvidia_cublas_cu12-12.4.2.65-py3-none-manylinux2014_x86_64.whl
Downloaded 2 packages in 1m 04s
Installed 2 packages in 1ms
 + nvidia-cublas-cu12==12.4.2.65
 + nvidia-cudnn-cu12==8.9.2.26

________________________________________________________
Executed in   64.90 secs    fish           external
   usr time    6.83 secs  271.00 micros    6.83 secs
   sys time    4.25 secs   77.00 micros    4.25 secs

python -m pip --version
pip 23.2

time pip install nvidia-cudnn-cu12==8.9.2.26 -v
Using pip 23.2 from ~/.venv/lib/python3.11/site-packages/pip (python 3.11)
Collecting nvidia-cudnn-cu12==8.9.2.26
  Obtaining dependency information for nvidia-cudnn-cu12==8.9.2.26 from https://files.pythonhosted.org/packages/ff/74/a2e2be7fb83aaedec84f391f082cf765dfb635e7caa9b49065f73e4835d8/nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl.metadata
  Downloading nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl.metadata (1.6 kB)
Collecting nvidia-cublas-cu12 (from nvidia-cudnn-cu12==8.9.2.26)
  Obtaining dependency information for nvidia-cublas-cu12 from https://files.pythonhosted.org/packages/f2/44/fdfd1a9f54167714d2c3dbc1bb1cb030340d6fb936a1dab9ef60999e3f73/nvidia_cublas_cu12-12.4.2.65-py3-none-manylinux2014_x86_64.whl.metadata
  Downloading nvidia_cublas_cu12-12.4.2.65-py3-none-manylinux2014_x86_64.whl.metadata (1.5 kB)
Downloading nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl (731.7 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 731.7/731.7 MB 4.7 MB/s eta 0:00:00
Downloading nvidia_cublas_cu12-12.4.2.65-py3-none-manylinux2014_x86_64.whl (363.0 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 363.0/363.0 MB 7.2 MB/s eta 0:00:00
Installing collected packages: nvidia-cublas-cu12, nvidia-cudnn-cu12
Successfully installed nvidia-cublas-cu12-12.4.2.65 nvidia-cudnn-cu12-8.9.2.26

________________________________________________________
Executed in   61.87 secs    fish           external
   usr time    9.64 secs  148.00 micros    9.64 secs
   sys time    7.22 secs  138.00 micros    7.22 secs

[dheerajck]

Can you explain this log
DEBUG No credentials found for: https://pypi.org/simple/nvidia-cudnn-cu12/

[dheerajck]

Also as asked above, adding a proper progress bar and download speed like this 10%, 10MBPS or like this 10MB/100MB, 10MBPS will be really helpful when downloading large packages

and I personally never want timeout just because download is slow, because that doesnt help, unless this somehow make download faster next time(this means download was slow because of a temp issue with uv download), timeout when connection is idle for x seconds makes sense to me

[zanieb]

DEBUG No credentials found for: https://pypi.org/simple/nvidia-cudnn-cu12/

This means we did not find any credentials for that URL and consequently did not populate them. It doesn't matter here since that's a public index.

adding a proper progress bar and download speed

Yep this is tracked in #1209

I personally never want timeout just because download is slow

The timeout should not occur during download, just during connection problems. See #1921 for more discussion.

I'm going to close this in favor of #1209 which is more actionable.