Commits · df909e830893694e213d25ba260bced9c0f72e28 · OpenDAS / AutoAWQ

11 Nov, 2023 1 commit
- Reset cache on new generation (#178) · df909e83
  Casper authored Nov 11, 2023
  
  df909e83
10 Nov, 2023 3 commits
- New logic for passing past_key_value (#177) · 5db86ec5
  Younes Belkada authored Nov 10, 2023
  
  5db86ec5
- Fix CUDA error: invalid argument (#179) · 63e3fd83
  Casper authored Nov 10, 2023
  
  63e3fd83
- Make sure to delete dummy model (#180) · 1f6f5155
  Casper authored Nov 10, 2023
  
  1f6f5155
06 Nov, 2023 1 commit
- Yi support (#167) · 37c63440
  Casper authored Nov 06, 2023
  
  37c63440
05 Nov, 2023 4 commits
- Fix typo in benchmark command (#161) · 781dc944
  Casper authored Nov 05, 2023
  
  781dc944
- New benchmarks in README (#160) · 5c7e3682
  Casper authored Nov 05, 2023
  
  5c7e3682
- Exclude download of CUDA wheels (#159) · 9f2c383f
  Casper authored Nov 05, 2023
  
  9f2c383f
- Build older cuda wheels (#158) · 9f9ce41f
  Casper authored Nov 05, 2023
  
  9f9ce41f
04 Nov, 2023 2 commits
- Create fused LlamaLikeModel (#152) · 8110e028
  Casper authored Nov 04, 2023
  
  8110e028
- Default to safetensors for quantized models (#151) · 84a26861
  Casper authored Nov 04, 2023
  
  84a26861
03 Nov, 2023 3 commits
- Fix condition when rolling cache (#150) · 958678df
  Casper authored Nov 04, 2023
  
  958678df
- [`core` / `attention`] Fix fused attention generation with newest transformers version (#146) · 92a403b2
  Younes Belkada authored Nov 04, 2023
```
Co-authored-by: Casper <casperbh.96@gmail.com>
```
  92a403b2
- Fix performance regression (#148) · 3eda6562
  Casper authored Nov 03, 2023
  
  3eda6562
02 Nov, 2023 6 commits
- Revert to previous version (#141) · bf64abd8
  Casper authored Nov 02, 2023
  
  bf64abd8
- CUDA 12 release (#140) · f3f05c75
  Casper authored Nov 02, 2023
  
  f3f05c75
- Bump to v0.1.6 (#139) · abf44cce
  Casper authored Nov 02, 2023
  
  abf44cce
- Benchmark info (#138) · 2bfa234f
  Casper authored Nov 02, 2023
  
  2bfa234f
- Update readme (#137) · 0f3214b1
  Casper authored Nov 02, 2023
  
  0f3214b1
- Fix init quant (#136) · 89f3feae
  Casper authored Nov 02, 2023
  
  89f3feae
31 Oct, 2023 1 commit
- AwqConfig class (#132) · eea08aa6
  Casper authored Oct 31, 2023
  
  eea08aa6
30 Oct, 2023 1 commit
- CUDA 11.8.0 and 12.1.1 build (#128) · a7d87540
  Casper authored Oct 30, 2023
  
  a7d87540
29 Oct, 2023 1 commit
- Pseudo dequantize function (#127) · f1a2e56e
  Casper authored Oct 29, 2023
  
  f1a2e56e
28 Oct, 2023 6 commits
- Bump to v0.1.5 · b9ed6641
  Casper authored Oct 28, 2023
  
  b9ed6641
- Add benchmark of Aquila2 34B AWQ in README.md. (#126) · d6ac3be3
  ldwang authored Oct 29, 2023
```
Signed-off-by: ldwang <ftgreat@gmail.com>
Co-authored-by: ldwang <ftgreat@gmail.com>
```
  d6ac3be3
- Support Aquila models. (#123) · dc13f0b8
  ldwang authored Oct 28, 2023
```
Signed-off-by: ldwang <ftgreat@gmail.com>
Co-authored-by: ldwang <ftgreat@gmail.com>
Co-authored-by: Casper Hansen <casperbh.96@gmail.com>
```
  dc13f0b8
- Fixing starcoder based models with 15B (#118) · 6c05f669
  Sebastian Bodza authored Oct 28, 2023
  
  6c05f669
- Fix potential overflow (#102) · 0cadd65f
  Casper authored Oct 28, 2023
  
  0cadd65f
- [`core`] Support fp32 / bf16 inference (#121) · 18712d00
  Younes Belkada authored Oct 28, 2023
  
  18712d00
27 Oct, 2023 1 commit
- 400 -> 600 models available · 26e94d2e
  Casper authored Oct 27, 2023
  
  26e94d2e
25 Oct, 2023 1 commit
- Update symbol to downloads/month · e539be9a
  Casper authored Oct 25, 2023
  
  e539be9a
20 Oct, 2023 1 commit
- Merge pull request #113 from twaka/gpt_neox · be395987
  Casper authored Oct 20, 2023
```
add gpt_neox support
```
  be395987
19 Oct, 2023 1 commit
- add comments · bcfdeb38
  twaka authored Oct 17, 2023
  
  bcfdeb38
17 Oct, 2023 1 commit
- gpt_neox · ad45716f
  twaka authored Oct 17, 2023
  
  ad45716f
07 Oct, 2023 2 commits
- Merge pull request #96 from casper-hansen/fix_attention_mask · 1b54b9f9
  Casper authored Oct 07, 2023
```
Only apply attention mask if seqlen is greater than 1
```
  1b54b9f9
- Only apply attention mask if seqlen is greater than 1 · e94b7f40
  Casper Hansen authored Oct 07, 2023
  
  e94b7f40
06 Oct, 2023 4 commits
- Bump to v0.1.4 · 0baf5e18
  Casper authored Oct 06, 2023
  
  0baf5e18
- Merge pull request #95 from casper-hansen/cache_refactor · 1b68975a
  Casper authored Oct 06, 2023
```
Refactor cache and embedding modules
```
  1b68975a
- Remove attention sinks · b13e2a85
  Casper Hansen authored Oct 06, 2023
  
  b13e2a85
- Create RoPE module · 69733d2c
  Casper Hansen authored Oct 06, 2023
  
  69733d2c