inseq-team inseq issues

inseq-team / inseq

Interpretability for sequence generation models 🐛 🔍

https://inseq.org

Apache License 2.0

378 stars 36 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Problem with skip_special_token in attribute method

#293 rafikg opened 2 weeks ago
0
Fix `MultiDimensional` attribution aggregation

#292 gsarti closed 2 weeks ago
0
Issue with zeroing_value attribution and SubwordAggregator()

#291 rafikg closed 1 week ago
8
Add `StringSplitAggregator`

#290 gsarti closed 2 weeks ago
0
Fix LLaMA 3.2, add `clean_special_chars`

#289 gsarti closed 1 month ago
0
How to set `max_length` in generation?

#288 Ningshiqi closed 1 month ago
1
Model LLaMA 3.2 not suported

#287 Ningshiqi closed 1 month ago
4
CUDA OOM with mistralai/Mistral-7B-Instruct-v0.3

#286 Ningshiqi closed 1 month ago
3
Does Inseq support attributions with example granularity as in Captum's few-shot?

#285 Ningshiqi closed 1 week ago
9
`treescope` visualization for `attribute_context`

#284 gsarti closed 3 months ago
0
`treescope` support and new visualizations

#283 gsarti closed 3 months ago
2
Improve attribution viz and `SliceAggregator`

#282 gsarti closed 4 months ago
0
Show only part of the input when calling show() method

#281 RibinMTC closed 4 months ago
2
Fix multi-gpu inference and duplicate BOS issue for decoder-only

#280 gsarti closed 4 months ago
0
CUDA out of memory issue on long context

#279 acDante closed 1 week ago
4
Forced generations with decoder-only models must start with input texts

#278 acDante closed 4 months ago
1
Generating explanations

#277 SVC04 closed 5 months ago
0
Expected all tensors to be on the same device multi-gpu

#276 saxenarohit closed 4 months ago
7
Add possibility to skip special tokens during attribution

#275 gsarti closed 6 months ago
0
`attribute_context` Error when output contains emoji

#274 Betswish opened 6 months ago
0
Save tensors in lower precision

#273 LuukSuurmeijer closed 4 months ago
3
Can't run inseq with batch_size>1 for methods that use gradients (e.g. integrated gradients)

#272 issam9 closed 6 months ago
2
Question about start token?

#271 FYYFU closed 6 months ago
7
Aggregate the attribution for each input sentence with LLama2 model

#270 acDante closed 6 months ago
4
Fix nans in attention, add pad in tokenizer if missing

#269 gsarti closed 7 months ago
0
Why sometimes the target attribution can be 'nan'?

#268 frankdarkluo closed 7 months ago
7
Support Value Zeroing for non-eager attention types

#267 gsarti closed 7 months ago
0
Issue with the implementation of Value Zeroing

#266 hmohebbi closed 7 months ago
2
Does inseq work with Mistral-7B or Llama-2-7B models?

#265 hadiasghari closed 7 months ago
3
Add support for `device_map` use

#264 gsarti closed 7 months ago
0
Can I load the inseq model onto multiple GPUs?

#263 frankdarkluo closed 7 months ago
5
Update lime.py

#262 yuzhaouoe opened 7 months ago
1
Which aggregation function is better?

#261 MarshtompCS closed 7 months ago
1
Long input document is truncated when calculating attribution scores

#260 acDante closed 8 months ago
3
Fix URL to arXiv

#259 bbjoverbeek closed 8 months ago
0
Various fixes to `attribute-context`

#258 gsarti closed 8 months ago
0
Update attribute_context.py

#257 Betswish closed 8 months ago
1
Sanitize strings from markup upon Rich printing

#256 gsarti closed 9 months ago
0
rich.errors.MarkupError for chat template

#255 mohsenfayyaz closed 9 months ago
2
Rollout aggregation function

#254 gsarti opened 9 months ago
0
Fix casting step outputs to device

#253 gsarti closed 9 months ago
3
Potential fix for failed `test_attribute_slice_seq2seq` on CUDA-enabled platforms

#252 xuan25 closed 9 months ago
0
Test failed on CUDA enabled platform

#251 xuan25 closed 9 months ago
0
Feature/reagent branch for ReAGent

#250 casszhao closed 7 months ago
7
Dev QOL updates

#249 gsarti closed 9 months ago
0
[Summary] Add step functions

#248 gsarti opened 9 months ago
0
Fix `aggregate_contiguous`

#247 gsarti closed 10 months ago
0
Producing heatmap visualizations with SubwordAggregator and SequenceAttributionAggregator

#246 nfelnlp closed 10 months ago
1
`attribute` speed and memory optimizations

#245 gsarti closed 10 months ago
1
Slow attribution possibly due to FeatureAttributionSequenceOutput.from_step_attributions

#243 saxenarohit closed 10 months ago
7