ndif-team/nnsight

Issues

• Calling an Envoy outside of a tracing context should work

  #156 opened 2 months ago by JadenFiotto-Kaufman
  1

• Llama-70-B causes GPU memory leak without explicit garbage collection

  #169 opened 3 months ago by arunasank
  4

• Create tutorial for remote execution

  #132 opened 3 months ago by JadenFiotto-Kaufman
  0

• Create tutorial for future lens

  #130 opened 3 months ago by JadenFiotto-Kaufman
  0

• Create tutorial for extending nnsight

  #129 opened 3 months ago by JadenFiotto-Kaufman
  0

• Create tutorial for "performance tips"

  #128 opened 3 months ago by JadenFiotto-Kaufman
  0

• Add a __version__attribute to nnsight

  #228 opened 4 months ago by ericwtodd
  0

• In-place editor calls repeat when the modified model is run

  #210 opened 4 months ago by mitroitskii
  5

• Unrecognized configuration class when using nnsight to load new LLMs

  #211 opened 4 months ago by ruizheliUOA
  4

• Cannot specify custom tokenizer with UnifiedTransformer

  #202 opened 4 months ago by Butanium
  0

• [Bug] Grad Setting Malfunctioning

  #141 opened 4 months ago by AdamBelfki3
  3

• Make tensor creation functions like torch.zeros traceable via nnsight

  #140 opened 4 months ago by JadenFiotto-Kaufman
  3

• Empty body of tracing context blows up

  #142 opened 4 months ago by arjunguha
  1

• Llama remote execution breaks on transformers 4.40.2

  #134 opened 4 months ago by sheikheddy
  2

• GPT-J rotary embeddings cause logits to be returned as FakeTensors when scan=True

  #162 opened 4 months ago by ericwtodd
  1

• Make it easier to see what default inputs can go in `LanguageModel`

  #173 opened 4 months ago by loftusa
  2

• Remote Execution Exception: Unprocessable Entity

  #191 opened 4 months ago by ShichangZh
  1

• nnsight does not work with pytorch 2.4.0 when llama model is not dispatched

  #190 opened 4 months ago by Butanium
  2

• Issue with eos_token_id in LanguageModel.generate() when using local models

  #177 opened 5 months ago by zzn-nzz
  2

• Issues with gemma-2b

  #185 opened 5 months ago by uSaiPrashanth
  2

• Expand pytest tests

  #133 opened 8 months ago by JadenFiotto-Kaufman
  1

• IndexError when flat inputs are concatenated during trace

  #176 opened 5 months ago by austinleedavis
  0

• ellipses does not work in the FakeTensor system

  #174 opened 5 months ago by loftusa
  1

• Make positional args optional

  #172 opened 5 months ago by ericwtodd
  0

• Support Llama-3

  #170 opened 6 months ago by jacklanda
  0

• Remote Execution handle_request should not hang when raising an exception

  #161 opened 6 months ago by JadenFiotto-Kaufman
  0

• Add Ellipses as valid type for remote execution

  #139 opened 7 months ago by JadenFiotto-Kaufman
  2

• AttributeError: 'LlamaForCausalLM' object has no attribute 'transformer' when using deepseek

  #154 opened 7 months ago by g-w1
  2

• TypeError when using CogVLM and pass scan=False

  #153 opened 7 months ago by Zoeyyao27
  4

• [Bug] RecursionError in NNsightModel initialization

  #147 opened 7 months ago by AdamBelfki3
  1

• Separate logging for remote execution

  #127 opened 7 months ago by JadenFiotto-Kaufman
  0

• Including model config in initialization call for LanguageModel causes errors when calling model.trace

  #115 opened 7 months ago by arunasank
  1

• kwargs are not passed to the tokenizer when initializing LanguageModel

  #116 opened 7 months ago by Butanium
  0

• Condition where NNSight silently fails when it should perhaps throw

  #145 opened 7 months ago by arjunguha
  0

• [Feature Request] Make token indexing work for batched input and UnifiedTransformer

  #138 opened 7 months ago by Butanium
  2

• NNsight should fail when an Unset proxy from a previous trace / future computation is used in a patching experiment

  #143 opened 7 months ago by Butanium
  0

• Error when using stopping_criteria in .generate if remote=True

  #137 opened 7 months ago by Butanium
  5

• Create/update tutorial for dictionary learning/SAEs

  #131 opened 8 months ago by JadenFiotto-Kaufman
  0

• Dispatch Error When Using Quantisation

  #106 opened 9 months ago by Xmaster6y
  1

• Cannot use cache if I don't use .value

  #118 opened 8 months ago by Butanium
  0

• Cannot load transformer lens model on CPU

  #121 opened 8 months ago by Butanium
  2

• AttributeError: '_thread._local' object has no attribute 'device'

  #122 opened 8 months ago by atlaie
  2

• `model.named_modules` returns PyTorch modules not NNSight wrapped modules

  #114 opened 8 months ago by neelnanda-io
  4

• Is it possible to apply layer modules out-of-order?

  #109 opened 8 months ago by arjunguha
  2

• Cannot use proxy values from previous runs with `remote=True`

  #104 opened 9 months ago by tvhong
  1

• Trouble using the Llama2 models

  #92 opened 9 months ago by MattBortoletto
  6

• Memory leak after trace?

  #102 opened 9 months ago by tvhong
  3

• 'WrapperModule' object has no attribute 'output'

  #97 opened 9 months ago by davidreber
  1

• Add links to source code in documentation

  #90 opened 9 months ago by ericwtodd
  1

• NNsight bug with quantized models

  #86 opened 9 months ago by HuFY-dev
  3