Commits · 386fede8de8ca56fb2e794602f271676b086efde · OpenDAS / AutoAWQ

21 Sep, 2023 2 commits
- Fix argument · 7c48033a
  Casper Hansen authored Sep 21, 2023
  
  7c48033a
- Improve model loading · 2dada8f8
  Casper Hansen authored Sep 21, 2023
  
  2dada8f8
20 Sep, 2023 6 commits
- Remove cosine loss · 90bf52c7
  Casper Hansen authored Sep 20, 2023
  
  90bf52c7
- Better comments. Implement cosine similarity · c9d01ac3
  Casper Hansen authored Sep 20, 2023
  
  c9d01ac3
- Move import up · e205548d
  Casper Hansen authored Sep 20, 2023
  
  e205548d
- Simplify saving model · ed618bb0
  Casper Hansen authored Sep 20, 2023
  
  ed618bb0
- Remove old quantization code · d3550fec
  Casper Hansen authored Sep 20, 2023
  
  d3550fec
- Working for OPT · 724bda58
  Casper Hansen authored Sep 20, 2023
  
  724bda58
15 Sep, 2023 2 commits
- Default to empty string for model file · 435b3b4b
  Casper Hansen authored Sep 15, 2023
  
  435b3b4b
- Actually pass args to calib function · a9cef34b
  Casper authored Sep 15, 2023
  
  a9cef34b
14 Sep, 2023 1 commit
- Pass split and text_column arguments to calib_data function · 84e82744
  Casper authored Sep 14, 2023
  
  84e82744
13 Sep, 2023 4 commits
- Unify using variable safetensors · 219ccb33
  Casper Hansen authored Sep 13, 2023
  
  219ccb33
- Fix ignore patterns · 7ba31254
  Casper Hansen authored Sep 13, 2023
  
  7ba31254
- Add safetensors support · db8ba322
  Casper Hansen authored Sep 13, 2023
  
  db8ba322
- Implement saving sharded weights + safetensors · 97d38e29
  Casper Hansen authored Sep 13, 2023
  
  97d38e29
09 Sep, 2023 1 commit
- Remove warnings about GEMM · c58ec73b
  Casper Hansen authored Sep 09, 2023
  
  c58ec73b
08 Sep, 2023 4 commits
- Fix variables and no contiguous memory for GEMV · e197a733
  Casper Hansen authored Sep 08, 2023
  
  e197a733
- Implement GEMM/GEMV in quantize function and fused modules · 5db43a7f
  Casper Hansen authored Sep 08, 2023
  
  5db43a7f
- Add deprecation warning · 9b2946b6
  Casper Hansen authored Sep 08, 2023
  
  9b2946b6
- Refactor modules, create separate GEMM and GEMV · fe314160
  Casper Hansen authored Sep 08, 2023
  
  fe314160
06 Sep, 2023 2 commits
- Remove simple dispatch · 562c0d52
  Casper authored Sep 06, 2023
  
  562c0d52
- Multi-GPU support for quantized models · f8273a0c
  Casper Hansen authored Sep 06, 2023
  
  f8273a0c
05 Sep, 2023 1 commit
- allow user to use custom calibration data for quantization · 1712ce21
  Erwan BOEHM authored Sep 05, 2023
  
  1712ce21
04 Sep, 2023 2 commits
- Pass trust_remote_code to AutoModelForCausalLM · 0090ad81
  Casper Hansen authored Sep 04, 2023
  
  0090ad81
- fuse_layers bug fix · 9106ce29
  qwopqwop200 authored Sep 04, 2023
  
  9106ce29
27 Aug, 2023 1 commit
- Rename to hijacked_inputs · 50d1025f
  Casper authored Aug 27, 2023
  
  50d1025f
25 Aug, 2023 1 commit
- fixed catcher input name · 6b1c96c7
  EC2 Default User authored Aug 25, 2023
  
  6b1c96c7
24 Aug, 2023 1 commit
- Integrate fused modules into AWQ model loading · 63346c34
  Casper Hansen authored Aug 24, 2023
  
  63346c34
22 Aug, 2023 2 commits
- Implement loading correct sequence length based on config + custom max_new_tokens · d9bab50c
  Casper Hansen authored Aug 22, 2023
  
  d9bab50c
- Set full quant as default. Always load unquantized with AutoModel instead of accelerate. · 321d74ff
  Casper Hansen authored Aug 22, 2023
  
  321d74ff
19 Aug, 2023 8 commits
- Fix logic for saving model · dd3010fb
  Casper Hansen authored Aug 19, 2023
  
  dd3010fb
- Support safetensors · 04e177ad
  Casper Hansen authored Aug 19, 2023
  
  04e177ad
- Fix dict access · 06073073
  Casper Hansen authored Aug 19, 2023
  
  06073073
- Save model file with filename from quant_config · dcd58818
  Casper Hansen authored Aug 19, 2023
  
  dcd58818
- Create generate method · cc779bed
  Casper Hansen authored Aug 19, 2023
  
  cc779bed
- Load quantized model with saved quant_config · d68441d0
  Casper Hansen authored Aug 19, 2023
  
  d68441d0
- Make BaseAWQForCausalLM a torch.nn.Module · 84d23089
  Casper Hansen authored Aug 19, 2023
  
  84d23089
- Rename q_config -> quant_config. Include w_bit in quant_config. Save quant_config.json. · d35ade75
  Casper Hansen authored Aug 19, 2023
  
  d35ade75
18 Aug, 2023 2 commits
- Increase robustness of model loading · 471f811b
  Casper Hansen authored Aug 18, 2023
  
  471f811b
- Refactor model class. Add LLaMa support (LLaMa-2, Vicuna, etc). · e2858cef
  Casper Hansen authored Aug 18, 2023
  
  e2858cef