Adamliu1 SNLP_GCW issues

Adamliu1 / SNLP_GCW

3 stars 0 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Added max sample length filter to unlearning and retaining datasets

#156 TheRootOf3 opened 2 months ago
0
Format Python code with psf/black push

#155 github-actions[bot] closed 2 months ago
0
Format Python code with psf/black push

#154 github-actions[bot] closed 2 months ago
0
Fix submodule link

#153 Adamliu1 closed 2 months ago
0
[dashbord] Add dropdowns to further customise the dashbord x axis and further filters

#152 TheRootOf3 closed 2 months ago
1
[experiment infra] Change the directory structure of experiment results

#151 TheRootOf3 opened 2 months ago
1
Format Python code with psf/black push

#150 github-actions[bot] closed 2 months ago
0
Add special case for saving model when running with ZERO-3 optimisation.

#149 Willmish opened 2 months ago
1
Format Python code with psf/black push

#148 github-actions[bot] closed 2 months ago
0
Format Python code with psf/black push

#147 github-actions[bot] closed 2 months ago
0
Added aggregation script and dashboard v1.1

#146 TheRootOf3 closed 2 months ago
1
[dashboard] Develop a simple visualisation dashboard for experiment results

#145 TheRootOf3 closed 2 months ago
0
Format Python code with psf/black push

#144 github-actions[bot] closed 2 months ago
0
Format Python code with psf/black push

#143 github-actions[bot] closed 2 months ago
0
Add precomputing of normal responses (only 1 model loaded at a time).

#142 Willmish closed 2 months ago
0
Figure out a visualisation of sequential unlearning results when compared to baseline/between each other.

#141 TheRootOf3 opened 2 months ago
0
[experiments infra] Write a script for aggregating sequential unlearning results and creating plots

#140 TheRootOf3 closed 2 months ago
0
Schedule the next round of experiments with other epoch count values + 1e-5 LR

#139 TheRootOf3 opened 2 months ago
2
Check model runs on 2 GPUs with normal answers pre-computed without offloading

#138 TheRootOf3 closed 2 months ago
2
Rewrite code to pre-generate normal responses for normal loss computation

#137 TheRootOf3 closed 2 months ago
1
Added multiple response generation and majority voting for harmfulness eval.

#136 TheRootOf3 closed 2 months ago
0
Added run_experiment.sh script for truthfulqa eval.

#135 TheRootOf3 closed 3 months ago
0
Unifying the eval_framework configurations.

#134 TheRootOf3 closed 3 months ago
0
Establish a common version of eval framework running script

#133 TheRootOf3 closed 3 months ago
0
Added model answer generation and evaluation for TruthfulQA.

#132 TheRootOf3 closed 3 months ago
0
Modified the generations config for beavertails harmfulness eval.

#131 TheRootOf3 closed 3 months ago
0
Changed generations config for eval harmfulness.

#130 TheRootOf3 closed 3 months ago
0
[cleaning] Move all the docs markdown files to `docs/` dir

#129 TheRootOf3 opened 3 months ago
0
Added datasets.md docs file. Moved cluster_notes.md to docs/ dir.

#128 TheRootOf3 closed 3 months ago
0
Format Python code with psf/black push

#127 github-actions[bot] closed 3 months ago
0
[Docs] what is the random guess performance for every benchmark

#126 Willmish closed 3 months ago
0
[EXPERIMENTS] Sequential unlearning llama 3 8B

#125 TheRootOf3 opened 3 months ago
0
[CODE OPTIMISATION] Unlearning a 8B model without offloading parameters (2 GPUs?)

#124 TheRootOf3 closed 2 months ago
2
Check length of responses+whitespace ratio+repeating question for Retain and Unlearn sets

#123 Willmish closed 3 months ago
0
[RQ2] Design and conduct a set of experiments for loss components ablation study.

#122 TheRootOf3 opened 3 months ago
0
Verify unlearning with opt1.3B but different retain set (Squad vs TruthfulQA)

#121 Willmish opened 3 months ago
1
Format Python code with psf/black push

#120 github-actions[bot] closed 4 months ago
0
Format Python code with psf/black push

#119 github-actions[bot] closed 4 months ago
0
Refactor the dataset related code.

#118 Davidyz closed 4 months ago
0
Integrate Accelerate, Deepspeed

#117 Adamliu1 closed 3 months ago
0
Format Python code with psf/black push

#116 github-actions[bot] closed 4 months ago
0
Integrate Accelerate, Deepspeed and fix start_loc compute for Microsfot Phi

#115 Adamliu1 closed 4 months ago
0
Quick fix for determine loc_start for Microsoft Phi

#114 Adamliu1 closed 3 months ago
1
DeepSpeed integration for <10B models on strangeporks.

#113 TheRootOf3 opened 5 months ago
2
Implement mathqa dataloader

#112 Davidyz closed 4 months ago
1
FInd Lr For Gemma 2B, bf16, Schedule unlearn, write script

#111 TheRootOf3 closed 5 months ago
1
Format Python code with psf/black push

#110 github-actions[bot] closed 5 months ago
0
[WIP] Eval scripts, harmfulness results, fix to eval_harmfulness, readme

#109 Adamliu1 opened 5 months ago
1
Adam/eval beavertails llama3

#108 Adamliu1 closed 5 months ago
0
Andrzej/bfloat16 unlearning

#107 TheRootOf3 closed 5 months ago
0