[wasm] Add more device testing

[lina128]

Added more device testing for WASM backend to nightly.

Tested nightly build in local: https://pantheon.corp.google.com/cloud-build/builds/108f3152-1162-4fbd-94df-e843b58f70f4;step=5?project=learnjs-174218

---

This change is 

[lina128]

Reviewable status: 0 of 1 approvals obtained (waiting on @annxingyuan, @pyu10055, and @tafsiri)

---

tfjs-backend-wasm/src/setup_test.ts, line 227 at r1 (raw file):

  {include: 'split'},
  {include: 'pad ', excludes: ['complex', 'zerosLike']},
  {include: 'clip', excludes: ['gradient', 'propagates NaNs']},

Hi @annxingyuan, this test fails with ios safari, the last value is not NaN, but 50, I excluded it for now. Please investigate, I will ping you the log.

[annxingyuan]

Reviewable status: complete! 1 of 1 approvals obtained (waiting on @annxingyuan, @lina128, @pyu10055, and @tafsiri)

---

tfjs-backend-wasm/karma.conf.js, line 108 at r1 (raw file):

    browserNoActivityTimeout: 3e5,
    browserDisconnectTimeout: 3e5,
    browserDisconnectTolerance: 0,

curious why we want to set it explicitly to 0? I thought the default was 0?

---

tfjs-backend-wasm/scripts/test-ci.sh, line 16 at r1 (raw file):

if [ "$NIGHTLY" = true ]; then
  yarn run-browserstack --browsers=bs_safari_mac

just curious why you split up the runs this way?

---

tfjs-backend-wasm/src/setup_test.ts, line 227 at r1 (raw file):

Previously, lina128 (Na Li) wrote…

Hi @annxingyuan, this test fails with ios safari, the last value is not NaN, but 50, I excluded it for now. Please investigate, I will ping you the log.

sounds good

[lina128]

Reviewable status: complete! 1 of 1 approvals obtained (waiting on @annxingyuan, @lina128, @pyu10055, and @tafsiri)

---

tfjs-backend-wasm/karma.conf.js, line 108 at r1 (raw file):

Previously, annxingyuan (Ann Yuan) wrote…

curious why we want to set it explicitly to 0? I thought the default was 0?

This set of configurations are recommended settings from BrowserStack. I changed disconnect tolerance to 1 once, but changed it back because at that time disconnect happens at timeout, so reconnect doesn't help. But maybe we can try changing tolerance to 1 for wasm, because I've seen disconnect occur very early recently. For now, I prefer to leave the field here so that we can tweak it easily later.

---

tfjs-backend-wasm/scripts/test-ci.sh, line 16 at r1 (raw file):

Previously, annxingyuan (Ann Yuan) wrote…

just curious why you split up the runs this way?

These are split randomly. All we need is to kickoff one browser first, so that it passed the binary to Browserstack, then later browsers will all share the binary. Is there any flag you want to test too?