Benchmark.prepare_backend() #204
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@recursix that's a bit hacky but it should do for now |
recursix
reviewed
Oct 24, 2024
browsergym/experiments/src/browsergym/experiments/benchmark/base.py
Outdated
Show resolved
Hide resolved
recursix
reviewed
Oct 24, 2024
| @@ -55,6 +55,7 @@ class Benchmark(DataClassJsonMixin): | |||
| high_level_action_set_args: HighLevelActionSetArgs | |||
| is_multi_tab: bool | |||
| env_args_list: list[EnvArgs] | |||
| full_reset_script: Optional[str] | |||
There was a problem hiding this comment.
hmmm this is a symptom that we should have went with a class hieararchy instead of a single class for all benchmarks
recursix
reviewed
Oct 24, 2024
recursix
reviewed
Oct 24, 2024
| @@ -50,6 +53,29 @@ def __init__( | |||
|
|
|||
| self.credentials = ACCOUNTS | |||
|
|
|||
| def full_reset(self): | |||
There was a problem hiding this comment.
helper function instead of code duplicate?
There was a problem hiding this comment.
What do you mean code duplicate? This code is new
There was a problem hiding this comment.
it's 90% the same code on the side of visualwebarena
recursix
previously approved these changes
Oct 24, 2024
gasse
changed the title
full_reset_script
recursix
previously approved these changes
Oct 24, 2024
qipeng
pushed a commit
to orby-ai-engineering/BrowserGym
that referenced
this pull request
Nov 20, 2024
qipeng
added a commit
to orby-ai-engineering/BrowserGym
that referenced
this pull request
Jan 18, 2025
* Patch VWA task IDs * Add BLIP2 evaluator; patch timeout * Actually add the captioning_fn into evaluator constructor * downgrading ubuntu version for github tests (ServiceNow#179) * making webarena tests not run on PRs (ServiceNow#181) * making webarena tests not run on PRs * making visualwebarena tests not run on PRs * SoM bugfix (ServiceNow#185) * version bump v0.8.1 * workflow image downgrade: ubuntu-latest -> ubuntu-22.04 * support custom observation * add user data dir * Benchmarks (ServiceNow#173) * new ControlOrMeta key modifier (ServiceNow#187) * Multi-tab fix (ServiceNow#188) * Global demo_mode flag (ServiceNow#177) * HighLevelActionSetArgs default value (ServiceNow#191) * version bump v0.9.0 * Reverting workarena_l1 benchmark to original seed sampling (ServiceNow#198) * Benchmarks update (ServiceNow#197) * Miniwob number of seeds 10 -> 5 * remove most benchmark variants --------- Co-authored-by: Maxime Gasse <[email protected]> * New benchmark AssistantBench (ServiceNow#186) --------- Co-authored-by: Maxime Gasse <[email protected]> * Default `browsergym_split` metadata for every benchmark (ServiceNow#190) --------- Co-authored-by: Xing Han Lu <[email protected]> Co-authored-by: ljang0 <[email protected]> Co-authored-by: Megh Thakkar <[email protected]> * Fixing logging with multiple jobs (ServiceNow#182) * Benchmark updates (ServiceNow#199) --------- Co-authored-by: Maxime Gasse <[email protected]> * version bump 0.10.0 * README update (ServiceNow#200) * Train / test splits for workarena-l2/l3 (ServiceNow#203) * Fine-grained benchmark action sets (ServiceNow#202) * version bump v0.10.1 * Update README.md * Update README.md * Benchmark.prepare_backend() (ServiceNow#204) * save_step_info bugfix (obs=None) (ServiceNow#207) * version bump v0.10.2 * full_reset fixes (ServiceNow#209) * Hide all bids from obs (ServiceNow#212) * Adding weblinx config to DEFAULT_BENCHMARKS (ServiceNow#208) --------- Co-authored-by: Maxime Gasse <[email protected]> * Leaner Unicode() gym space (ServiceNow#218) * a method to get the status of an experiment (ServiceNow#219) * version bump v0.11.0 * Rename benchmark after subset_from_split() (ServiceNow#221) * exp_dir sanitization (ServiceNow#222) * get_step_info() bugfix (ServiceNow#223) * Set webarena / visualwebarena max steps to 30 (ServiceNow#214) * Benchmark dependencies (ServiceNow#220) * Include nltk.download() in benchmark.prepare_backend() for webarena / visualwebarena (ServiceNow#224) * version bump v0.11.1 * ExpResult.status minor fix (ServiceNow#225) * version bump 0.11.2 * Fix duplicate depends_on in webarena metadata (ServiceNow#228) * Duplicate webarena dependencies fix (ServiceNow#229) * nltk.download() during import for webarena and visualwebarena (ServiceNow#227) * Refactor full_reset() for webarena / visualwebarena (ServiceNow#230) * webarena_tiny (ServiceNow#232) * Set ExpArgs.exp_id at post-init time (ServiceNow#231) * Remove ARIA extraction warnings (ServiceNow#233) * Update README.md * Update README.md * Update README.md * version bump v0.11.3 * ci tests fix (ServiceNow#234) * Benchmark update for weblinx (ServiceNow#235) * Refactor ExpArgs.exp_id generation (ServiceNow#236) * VisualWebArena task dependencies (ServiceNow#237) * VWA dependencies fix (ServiceNow#239) * VWA evaluator fix, missing captioning_fn (ServiceNow#240) * version bump v0.12.0 * Update README.md * VWA hide huggingface progress bar (ServiceNow#241) * WebLINX pre-download data in prepare_backend() (ServiceNow#226) * AssistantBench + WebLINX fixes (ServiceNow#244) * Increase assistantbench max_steps to 30 * Setting AssistantBench locale and timezone * Dedicated AssistantBench action set * small fix * missing change * Lenient frame marking in last retry (ServiceNow#245) * WA / VWA default action set update (ServiceNow#247) * version bump v0.13.0 * visualwebarena massage (ServiceNow#248) * Minor fix (ServiceNow#250) * Remove gym warnings "obs not within observation space" (ServiceNow#251) * Lower trace level info -> debug (ServiceNow#252) * Make env.close() usable after failure (finally block) (ServiceNow#253) * add init script support * VWA / WA updates (ServiceNow#254) * Minor refactors (ServiceNow#255) * Optional method AbstractBrowserTask.teardown() * browsergym registration refactor * Deal with problematic frame unmarking (ServiceNow#256) * Add missing property exception to _get_obs() retry (ServiceNow#258) * Bump libwebarena / libvisualwebarena dependencies (ServiceNow#257) * Massage WebArena instance (ServiceNow#259) * Refactor AssistantBench output directories (ServiceNow#242) --------- Co-authored-by: Maxime Gasse <[email protected]> * version bump v0.13.1 * Update README.md * Update README.md * Update README.md * Update README.md * Authors update (ServiceNow#260) * TapeAgents export for experiment results (ServiceNow#238) * Update README.md * Cleanup * Add weblinx_browsergym as a dependency (ServiceNow#261) * Typo fix (ServiceNow#264) * Update requirements.txt to latest libvisualwebarena package that includes local hosting (ServiceNow#165) * adding AgentInfo to __init__ for convenience (ServiceNow#166) * libvisualwebarena==0.0.14 (ServiceNow#171) fixed the jsons file! * Leaner traces (ServiceNow#169) * images aren't saved in pkl files anymore, and are stuffed back in at load time * added kwargs to control img/som saving * saving as png, adding screenshots back into obs * retrocompatibility for image loading * making get_screenshots work for png and jpg * fixing image types and closing files * Goal refactor to allow for local image files (ServiceNow#110) --------- Co-authored-by: Thibault Le Sellier de Chezelles <[email protected]> Co-authored-by: Maxime Gasse <[email protected]> * version bump 0.8.0 * Integrate AgentLab tests (ServiceNow#176) * downgrading ubuntu version for github tests (ServiceNow#179) * making webarena tests not run on PRs (ServiceNow#181) * making webarena tests not run on PRs * making visualwebarena tests not run on PRs * SoM bugfix (ServiceNow#185) * version bump v0.8.1 * workflow image downgrade: ubuntu-latest -> ubuntu-22.04 * Benchmarks (ServiceNow#173) * new ControlOrMeta key modifier (ServiceNow#187) * Multi-tab fix (ServiceNow#188) * Global demo_mode flag (ServiceNow#177) * HighLevelActionSetArgs default value (ServiceNow#191) * version bump v0.9.0 * Reverting workarena_l1 benchmark to original seed sampling (ServiceNow#198) * Benchmarks update (ServiceNow#197) * Miniwob number of seeds 10 -> 5 * remove most benchmark variants --------- Co-authored-by: Maxime Gasse <[email protected]> * New benchmark AssistantBench (ServiceNow#186) --------- Co-authored-by: Maxime Gasse <[email protected]> * Default `browsergym_split` metadata for every benchmark (ServiceNow#190) --------- Co-authored-by: Xing Han Lu <[email protected]> Co-authored-by: ljang0 <[email protected]> Co-authored-by: Megh Thakkar <[email protected]> * Fixing logging with multiple jobs (ServiceNow#182) * Benchmark updates (ServiceNow#199) --------- Co-authored-by: Maxime Gasse <[email protected]> * version bump 0.10.0 * README update (ServiceNow#200) * Train / test splits for workarena-l2/l3 (ServiceNow#203) * Fine-grained benchmark action sets (ServiceNow#202) * version bump v0.10.1 * Update README.md * Update README.md * Benchmark.prepare_backend() (ServiceNow#204) * save_step_info bugfix (obs=None) (ServiceNow#207) * version bump v0.10.2 * full_reset fixes (ServiceNow#209) * Hide all bids from obs (ServiceNow#212) * Adding weblinx config to DEFAULT_BENCHMARKS (ServiceNow#208) --------- Co-authored-by: Maxime Gasse <[email protected]> * Leaner Unicode() gym space (ServiceNow#218) * a method to get the status of an experiment (ServiceNow#219) * version bump v0.11.0 * Rename benchmark after subset_from_split() (ServiceNow#221) * exp_dir sanitization (ServiceNow#222) * get_step_info() bugfix (ServiceNow#223) * Set webarena / visualwebarena max steps to 30 (ServiceNow#214) * Benchmark dependencies (ServiceNow#220) * Include nltk.download() in benchmark.prepare_backend() for webarena / visualwebarena (ServiceNow#224) * version bump v0.11.1 * ExpResult.status minor fix (ServiceNow#225) * version bump 0.11.2 * Fix duplicate depends_on in webarena metadata (ServiceNow#228) * Duplicate webarena dependencies fix (ServiceNow#229) * nltk.download() during import for webarena and visualwebarena (ServiceNow#227) * Refactor full_reset() for webarena / visualwebarena (ServiceNow#230) * webarena_tiny (ServiceNow#232) * Set ExpArgs.exp_id at post-init time (ServiceNow#231) * Remove ARIA extraction warnings (ServiceNow#233) * Update README.md * Update README.md * Update README.md * version bump v0.11.3 * ci tests fix (ServiceNow#234) * Benchmark update for weblinx (ServiceNow#235) * Refactor ExpArgs.exp_id generation (ServiceNow#236) * VisualWebArena task dependencies (ServiceNow#237) * VWA dependencies fix (ServiceNow#239) * VWA evaluator fix, missing captioning_fn (ServiceNow#240) * version bump v0.12.0 * Update README.md * VWA hide huggingface progress bar (ServiceNow#241) * WebLINX pre-download data in prepare_backend() (ServiceNow#226) * AssistantBench + WebLINX fixes (ServiceNow#244) * Increase assistantbench max_steps to 30 * Setting AssistantBench locale and timezone * Dedicated AssistantBench action set * small fix * missing change * Lenient frame marking in last retry (ServiceNow#245) * WA / VWA default action set update (ServiceNow#247) * version bump v0.13.0 * visualwebarena massage (ServiceNow#248) * Minor fix (ServiceNow#250) * Remove gym warnings "obs not within observation space" (ServiceNow#251) * Lower trace level info -> debug (ServiceNow#252) * Make env.close() usable after failure (finally block) (ServiceNow#253) * VWA / WA updates (ServiceNow#254) * Minor refactors (ServiceNow#255) * Optional method AbstractBrowserTask.teardown() * browsergym registration refactor * Deal with problematic frame unmarking (ServiceNow#256) * Add missing property exception to _get_obs() retry (ServiceNow#258) * Bump libwebarena / libvisualwebarena dependencies (ServiceNow#257) * Massage WebArena instance (ServiceNow#259) * Refactor AssistantBench output directories (ServiceNow#242) --------- Co-authored-by: Maxime Gasse <[email protected]> * version bump v0.13.1 * Fix broken links * Update README.md * fix merging issues * Update README.md (ServiceNow#270) * Update README.md * README update * More permissive WA/VWA instance reset (ServiceNow#272) * New debug benchmark visualwebarena_tiny (ServiceNow#271) * Version bump v0.13.2 * Update README.md * Metadata column fix (ServiceNow#278) * Update README.md * Update README.md * Update README.md * Update README.md * Shunt WA / VWA unit tests * README update * Minor fixes (ServiceNow#281) * version bump v0.13.3 * remove unused fluff * revert more unintended changes --------- Co-authored-by: Peng Qi <[email protected]> Co-authored-by: Thibault LSDC <[email protected]> Co-authored-by: Maxime Gasse <[email protected]> Co-authored-by: Yanan Xie <[email protected]> Co-authored-by: Alexandre Lacoste <[email protected]> Co-authored-by: oriyor <[email protected]> Co-authored-by: Xing Han Lu <[email protected]> Co-authored-by: ljang0 <[email protected]> Co-authored-by: Megh Thakkar <[email protected]> Co-authored-by: Imene Kerboua <[email protected]> Co-authored-by: Oleh Shliazhko <[email protected]> Co-authored-by: Thibault Le Sellier de Chezelles <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.