Commit Graph
54 Commits
Author SHA1 Message Date
Morgan Funtowicz b98ff88544 Added pipelines quick tour in README 2019-12-20 15:52:50 +01:00
Morgan Funtowicz 7f74084528 Fix leading axis added when saving through the command run 2019-12-20 14:47:04 +01:00
Morgan Funtowicz 61d9ee45e3 All tests are green. 2019-12-20 11:47:56 +01:00
Morgan Funtowicz e516a34a15 Use BasicTokenizer to split over whitespaces. 2019-12-20 09:38:08 +01:00
Morgan Funtowicz 9d0d1cd339 Filter out entity for NER task. 2019-12-20 09:30:37 +01:00
Morgan Funtowicz a305067f2d Removed __main__ 2019-12-19 19:41:48 +01:00
Morgan Funtowicz 3492a6ec17 Addressing Thom's comments. 2019-12-19 19:06:44 +01:00
Morgan Funtowicz 81a911cce5 Doc, doc, ... doc. 2019-12-19 15:12:06 +01:00
Morgan Funtowicz faef6f6191 Fix logic order for USE_TF/USE_TORCH 2019-12-19 12:28:17 +01:00
Morgan Funtowicz 5664327c24 Hide train command for now. 2019-12-19 12:27:54 +01:00
Morgan Funtowicz 3b29322d4c Expose all the pipeline argument on serve command. 2019-12-19 12:24:17 +01:00
Morgan Funtowicz fc624716aa Renaming framework env variables flags from NO_ to USE_ 2019-12-19 11:49:06 +01:00
Morgan Funtowicz f516cf3956 Allow pipeline to write output in binary format 2019-12-19 11:42:33 +01:00
Morgan Funtowicz d72fa2a0f6 Fix inputs_for_model call in QuestionAnsweringPipeline accessing __dict__ on list. 2019-12-19 10:54:10 +01:00
Morgan Funtowicz bcc99fd92e Fix wrong automatic config allocation through AutoConfig 2019-12-19 10:32:21 +01:00
Morgan Funtowicz ec5d6c6a70 Adressing issue with NER task omitting first and last word. 2019-12-19 00:12:10 +01:00
Morgan Funtowicz d0724d0794 Add PipedPipelineDataFormat 2019-12-18 23:27:26 +01:00
Morgan Funtowicz 7711403bbd Expose config through the cli arguments 2019-12-18 22:59:51 +01:00
Morgan Funtowicz 8bb166db5d Expose more information in the output of TextClassificationPipeline 2019-12-18 22:53:19 +01:00
Morgan Funtowicz 04b602f96f Put module import on top of the module. 2019-12-18 18:28:39 +01:00
Morgan Funtowicz 0c88c856d5 Unnest QuestionAnsweringArgumentHandler 2019-12-18 18:18:16 +01:00
Morgan Funtowicz e347725d8c More fine-grained control over pipeline creation with config argument. 2019-12-18 10:41:24 +01:00
Morgan Funtowicz 2fde5a2489 Initial bunch of documentation. 2019-12-17 12:16:07 +01:00
Morgan Funtowicz d7c62661a3 Provide serving dependencies for tensorflow and pytorch (serving-tf, serving-torch) 2019-12-17 11:23:39 +01:00
Morgan Funtowicz a096e2a88b WIP serving through HTTP internally using pipelines. 2019-12-16 16:38:02 +01:00
Morgan Funtowicz 43a4e1bbe4 Adressing issue in varargs handling for question answering. 2019-12-16 16:00:41 +01:00
Morgan Funtowicz 46ccbb42fc Make CLI run command use integer mapping for device argument. 2019-12-16 15:49:41 +01:00
Morgan Funtowicz bbc707cf39 Fix non-keyworded varargs handling in DefaultArgumentHandler for pipeline. 2019-12-16 15:49:09 +01:00
Morgan Funtowicz 9c391277cc Allow tensors placement on specific device through CLI and pipeline. 2019-12-16 15:19:13 +01:00
Morgan Funtowicz 955d7ecb57 Refactored Pipeline with dedicated argument handler. 2019-12-16 14:34:54 +01:00
Morgan Funtowicz 8e3b1c860f Added FeatureExtraction pipeline. 2019-12-15 01:37:52 +01:00
Morgan Funtowicz f1971bf303 Binding pipelines to the cli. 2019-12-15 01:37:16 +01:00
Morgan Funtowicz 0b51532ce9 Reintroducing the batch_encode_plus method 2019-12-13 16:22:50 +01:00
Morgan Funtowicz 8938b546bf Removed from_config 2019-12-13 14:27:04 +01:00
Morgan Funtowicz 1ca52567a4 Allow model conversion in the pipeline allocator. 2019-12-13 14:13:14 +01:00
Morgan Funtowicz 28e64ad5a4 Raise an exception if the pipeline allocator can't determine the tokenizer from the model. 2019-12-13 14:12:54 +01:00
Morgan Funtowicz be5bf7b81b Added NER pipeline. 2019-12-13 14:12:17 +01:00
Morgan Funtowicz 80eacb8f16 Adding labels mapping for classification models in their respective config. 2019-12-13 14:10:22 +01:00
Morgan Funtowicz f69dbecc38 Expose classification labels mapping (and reverse) in model config. 2019-12-12 10:25:36 +01:00
Morgan Funtowicz c28273793e Add missing DistilBert and Roberta to AutoModelForTokenClassification 2019-12-11 15:31:45 +01:00
Morgan Funtowicz b040bff6df Added supported model to AutoModelTokenClassification 2019-12-11 14:13:58 +01:00
Morgan Funtowicz 9a24e0cf76 Refactored qa pipeline argument handling + unittests 2019-12-11 00:33:25 +01:00
Morgan Funtowicz 63e36007ee Make sure padding, cls and another non-context tokens cannot appear in the answer. 2019-12-10 16:47:35 +01:00
Morgan Funtowicz 40a39ab650 Reuse recent SQuAD refactored data structure inside QA pipelines. 2019-12-10 15:59:38 +01:00
Morgan Funtowicz aae74065df Added QuestionAnsweringPipeline unit tests. 2019-12-10 13:37:20 +01:00
Morgan Funtowicz a7d3794a29 Remove token_type_ids for compatibility with DistilBert 2019-12-10 13:37:20 +01:00
Morgan Funtowicz fe0f552e00 Use attention_mask everywhere. 2019-12-10 13:37:20 +01:00
Morgan Funtowicz 348e19aa21 Expose attention_masks and input_lengths arguments to batch_encode_plus 2019-12-10 13:37:18 +01:00
Morgan Funtowicz c2407fdd88 Enable the Tensorflow backend. 2019-12-10 13:37:14 +01:00
Morgan Funtowicz f116cf599c Allow hidding frameworks through environment variables (NO_TF, NO_TORCH). 2019-12-10 13:37:07 +01:00
Morgan Funtowicz 6e61e06051 batch_encode_plus generates the encoder_attention_mask to avoid attending over padded values. 2019-12-10 13:37:07 +01:00
Morgan Funtowicz 02110485b0 Added batching, topk, chars index and scores. 2019-12-10 13:36:55 +01:00
Morgan Funtowicz e1d89cb24d Added QuestionAnsweringPipeline with batch support. 2019-12-10 13:36:55 +01:00
Morgan Funtowicz 81babb227e Added download command through the cli.
It allows to predownload models and tokenizers.
2019-12-10 12:18:59 +01:00