Not possible to install on Ubuntu 22.04

[smhanov]

I can't install it on Ubuntu 22.04.2 LTS:

smhanov@localhost:~/llama$ npm install dalai
npm ERR! code 1
npm ERR! path /home/smhanov/llama/node_modules/node-pty
npm ERR! command failed 
npm ERR! command sh -c node scripts/install.js

npm ERR! A complete log of this run can be found in:
npm ERR!     /home/smhanov/.npm/_logs/2023-04-12T16_31_28_771Z-debug-0.log

node version is 18.5.0 python version is 3.10.6

I also tried using the npx method:

smhanov@localhost:~/llama$ npx dalai llama install 7B
Need to install the following packages:
      [email protected]
Ok to proceed? (y) 

The command takes about 10 seconds.

smhanov@localhost:~/llama$ npx dalai serve
Need to install the following packages:
  [email protected]
Ok to proceed? (y) 
smhanov@localhost:~/llama$ 

After this nothing seems to happen. There is no server running on port 3000.

The error in the logs mentioned above is: 1649 timing build:link:node_modules/semver Completed in 15ms 1650 timing build:link:node_modules/sha.js Completed in 15ms 1651 timing build:link:node_modules/utp-native Completed in 15ms 1652 timing build:link Completed in 21ms 1653 info run [email protected] install node_modules/bufferutil node-gyp-build 1654 info run [email protected] install { code: 0, signal: null } 1655 timing build:run:install:node_modules/bufferutil Completed in 675ms 1656 info run [email protected] install node_modules/node-pty node scripts/install.js 1657 info run [email protected] install { code: 1, signal: null } 1658 timing reify:rollback:createSparse Completed in 938ms 1659 timing reify:rollback:retireShallow Completed in 0ms 1660 timing command:install Completed in 19961ms 1661 verbose stack Error: command failed 1661 verbose stack at ChildProcess. (/snap/node/7268/lib/node_modules/npm/node_modules/@npmcli/promise-spaw n/lib/index.js:53:27) 1661 verbose stack at ChildProcess.emit (node:events:513:28) 1661 verbose stack at maybeClose (node:internal/child_process:1091:16) 1661 verbose stack at ChildProcess._handle.onexit (node:internal/child_process:302:5) 1662 verbose pkgid [email protected] 1663 verbose cwd /home/smhanov/llama 1664 verbose Linux 5.15.0-69-generic 1665 verbose node v18.15.0 1666 verbose npm v9.5.0 1667 error code 1 1668 error path /home/smhanov/llama/node_modules/node-pty 1669 error command failed 1670 error command sh -c node scripts/install.js 1671 verbose exit 1 1672 timing npm Completed in 20089ms 1673 verbose unfinished npm timer reify 1681317088901 1674 verbose unfinished npm timer reify:build 1681317106583 1675 verbose unfinished npm timer build 1681317106585 1676 verbose unfinished npm timer build:deps 1681317106585 1677 verbose unfinished npm timer build:run:install 1681317106641 1678 verbose unfinished npm timer build:run:install:node_modules/node-pty 1681317107318 1679 verbose code 1 1680 error A complete log of this run can be found in: 1680 error /home/smhanov/.npm/_logs/2023-04-12T16_31_28_771Z-debug-0.log

In addition, trying to install using the docker compose method fails at the docker compose step, due to a python venv issue.

#0 363.2 npm ERR! /root/dalai/venv/bin/pip install --user torch torchvision torchaudio sentencepiece numpy
#0 363.2 npm ERR! exit
#0 363.2 npm ERR! root@buildkitsandbox:~/dalai/node_modules/dalai# /root/dalai/venv/bin/pip install --user torch 
torchvision torchaudio sentencepiece numpy
#0 363.2 npm ERR! ERROR: Can not perform a '--user' install. User site-packages are not visible in this virtualenv.
#0 363.2 npm ERR! root@buildkitsandbox:~/dalai/node_modules/dalai# exit
#0 363.2 npm ERR! exit
#0 363.2 npm ERR! Error Error: dependency installation failed   
#0 363.2 npm ERR!     at Dalai.setup (/root/dalai/node_modules/dalai/index.js:510:15)
#0 363.2
#0 363.2 npm ERR! A complete log of this run can be found in:
#0 363.2 npm ERR!     /root/.npm/_logs/2023-04-12T16_42_55_249Z-debug-0.log
------
failed to solve: executor failed running [/bin/sh -c npm install [email protected]]: exit code: 1

[gwerneckp]

Having the same problem here.

[debsdebsenson]

+1

[MayurVirkar]

+1 But on Win 11

[gkhngyk]

+1

[smhanov]

I ended up using llama.cpp instead. It is very easy to get up and running and also supports alpaca and vicuna models. But it only runs on CPU, not GPU.