Commits · 3319eb5490103b3639e671da2b629fa6303a294c · chenpangpang / transformers

12 Sep, 2023 1 commit
- Generate: legacy mode is only triggered when `generation_config` is untouched (#25962) · 3319eb54
  Joao Gante authored Sep 12, 2023
  
  3319eb54
23 Aug, 2023 1 commit
- Generate: general test for decoder-only generation from `inputs_embeds` (#25687) · 3c2383b1
  Joao Gante authored Aug 23, 2023
```
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
```
  3c2383b1
16 Aug, 2023 1 commit
- Generate: fix default max length warning (#25539) · 3f9cb335
  Joao Gante authored Aug 16, 2023
  
  3f9cb335
09 Aug, 2023 1 commit

aligned sample_beam output selection with beam_search (#25375) · cb3c821c

hukuda222 authored Aug 10, 2023



* aligned sample_beam specs with beam_search

* pull origin main

* Revert "pull origin main"

This reverts commit 06d356f1137bb52272e120a03636598c44449cf3.

* update test_utils.py

* fix format

* remove comment

---------
Co-authored-by: Shogo Fujita <shogo.fujita@legalontech.jp>

cb3c821c

06 Aug, 2023 1 commit
- add CFG for .generate() (#24654) · d5334651
  Guillaume "Vermeille" Sanchez authored Aug 06, 2023
  
  d5334651
20 Jul, 2023 1 commit
- Contrastive Search peak memory reduction (#24120) · caf5e369
  Benjamin Badger authored Jul 20, 2023
```
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>
```
  caf5e369
27 Jun, 2023 1 commit
- Generate: `group_beam_search` requires `diversity_penalty>0.0` (#24456) · 5f3efdf7
  Joao Gante authored Jun 27, 2023
```
* add exception

* update docs
```
  5f3efdf7
23 Jun, 2023 1 commit

Replace python random with torch.rand to enable dynamo.export (#24434) · a28325e2

Bowen Bao authored Jun 23, 2023

* Replace python random with torch.rand to enable dynamo.export

* revert changes to flax model code

* Remove unused random import

* Fix torch template

* Move torch.manual_seed(0) to right location

a28325e2

07 Jun, 2023 1 commit
- Generate: increase left-padding test atol (#23448) · 612b2a1a
  Joao Gante authored Jun 07, 2023
```
increase atol
```
  612b2a1a
18 May, 2023 2 commits
- Less flaky `test_assisted_decoding_matches_greedy_search` (#23451) · 2406dbdc
  Yih-Dar authored May 18, 2023
```
* fix

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
```
  2406dbdc
- Generate: skip left-padding tests on old models (#23437) · aea7b23b
  Joao Gante authored May 18, 2023
  
  aea7b23b
16 May, 2023 1 commit
- Generate: add test to check KV format (#23403) · 918a06e2
  Joao Gante authored May 16, 2023
```
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>
```
  918a06e2
08 May, 2023 1 commit
- Generate: starcoder 🤜 🤛 assisted generation (#23182) · bbfb9fc2
  Joao Gante authored May 08, 2023
```
* starcoder has joined the chat

* indexing that works for all
```
  bbfb9fc2
03 May, 2023 1 commit
- Generate: slow assisted generation test (#23125) · ce31e3c8
  Joao Gante authored May 03, 2023
  
  ce31e3c8
29 Apr, 2023 1 commit
- Generate: prepare assisted generation for release (#23052) · 849367cc
  Joao Gante authored Apr 29, 2023
  
  849367cc
24 Apr, 2023 1 commit
- Generate: assisted generation with sample (take 2) (#22949) · e4a97f82
  Joao Gante authored Apr 24, 2023
```
* temperature controls speed
```
  e4a97f82
18 Apr, 2023 2 commits

Generate: Add assisted generation (#22211) · 78cda46f

Joao Gante authored Apr 18, 2023

* working mvp

* remove breakpoint

* fix commit

* standardize outputs

* tmp commit

* tests almost ready

* tmp commit

* skip a few models

* Add streaming; Docs and examples

* document limitations

* PR commits

* Amy PR comments

78cda46f

Fix `test_eos_token_id_int_and_list_top_k_top_sampling` (#22826) · 90247d3e
Yih-Dar authored Apr 18, 2023
```
* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
```
90247d3e

13 Apr, 2023 1 commit
- Generate: handle text conditioning with multimodal encoder-decoder models (#22748) · 9dfd6a4b
  Joao Gante authored Apr 13, 2023
  
  9dfd6a4b
23 Mar, 2023 1 commit
- Generate: add test for left-padding support (#22322) · 502fec77
  Joao Gante authored Mar 23, 2023
  
  502fec77
22 Mar, 2023 1 commit
- Beef up Llama tests (#22314) · fd3eb3e3
  Joao Gante authored Mar 22, 2023
```
* tmp commit

* beef up llama tests
```
  fd3eb3e3
16 Mar, 2023 1 commit

(#22204) · 5110e574

Yih-Dar authored Mar 16, 2023



* py38 + torch 2

* increment cache versions

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

5110e574

22 Feb, 2023 1 commit
- Apply ruff flake8-comprehensions (#21694) · 5e8c8eb5
  Aaron Gokaslan authored Feb 22, 2023
  
  5e8c8eb5
14 Feb, 2023 1 commit
- Generate: filter encoder inputs when its signature does not accept wildcards (#21603) · 13e03e61
  Joao Gante authored Feb 14, 2023
  
  13e03e61
13 Feb, 2023 3 commits
- Generate: correct default model input creation for decoder-only models (#21580) · fa4bdb0a
  Joao Gante authored Feb 13, 2023
  
  fa4bdb0a
- Generate: Fix flaky indexing error in `test_constrained_beam_search_generate_dict_output` (#21561) · 24273268
  Joao Gante authored Feb 13, 2023
  
  24273268
- Generate: TF supports multiple eos tokens (#21571) · eb6c59bc
  Joao Gante authored Feb 13, 2023
  
  eb6c59bc
09 Feb, 2023 1 commit
- Generate: make TF `.generate()` signature == PT `.generate()` signature (#21525) · e69f9715
  Joao Gante authored Feb 09, 2023
  
  e69f9715
08 Feb, 2023 2 commits

Fix multiple `eos_token_id`s in model.generate(...) (#21461) · 9960506c

Motoki Wu authored Feb 08, 2023

* add tests with multiple eos_token_ids

* make math.prod instead of sum

* make fixup

* fix long and also use np.prod since math.prod does not exist <python 3.8

* make fixup

* add prod util

* use prod util instead of np.prod

* make fixup

* previous .long location

* use tensor ops

* remove prod

* remove prod

* update device

* make fixup

* fix none

9960506c

Generate: TF `compute_transition_scores` (#21341) · 1d9c26a4
Joao Gante authored Feb 08, 2023

1d9c26a4

07 Feb, 2023 1 commit
- Generate: TF can now generate from embeddings in encoder-decoder models (#21475) · 1e4cf8bb
  Joao Gante authored Feb 07, 2023
  
  1e4cf8bb
06 Feb, 2023 2 commits

Update quality tooling for formatting (#21480) · 6f79d264

Sylvain Gugger authored Feb 06, 2023

* Result of black 23.1

* Update target to Python 3.7

* Switch flake8 to ruff

* Configure isort

* Configure isort

* Apply isort with line limit

* Put the right black version

* adapt black in check copies

* Fix copies

6f79d264

Generate: TF can now accept custom logits processors (#21454) · 49433310
Joao Gante authored Feb 06, 2023

49433310

03 Feb, 2023 2 commits
- Avoid flaky generation sampling tests (#21445) · 59d5edef
  Yih-Dar authored Feb 03, 2023
```
* fix

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
```
  59d5edef
- 🚨🚨 Generate: standardize beam search behavior across frameworks (#21368) · f21af262
  Joao Gante authored Feb 03, 2023
  
  f21af262
01 Feb, 2023 1 commit
- Generate: decoder-only models can generate with `inputs_embeds` (#21405) · 92ce53aa
  Joao Gante authored Feb 01, 2023
  
  92ce53aa
31 Jan, 2023 1 commit
- Template for framework-agnostic tests (#21348) · 623346ab
  Joao Gante authored Jan 31, 2023
  
  623346ab
30 Jan, 2023 1 commit
- Generate: Relaxed `max_length` and `max_new_tokens` coexistence (#21347) · 42b60f8b
  Joao Gante authored Jan 30, 2023
```
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
```
  42b60f8b
20 Jan, 2023 1 commit
- Generate: documented function to compute the transition scores (#21191) · af37d183
  Joao Gante authored Jan 20, 2023
```
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
```
  af37d183
04 Jan, 2023 1 commit
- Generate: Fix CI related to #20727 (#21003) · b9104896
  Joao Gante authored Jan 04, 2023
  
  b9104896