v0.3.0
In this release we moved all generation options to a new %Bumblebee.Text.GenerationConfig{} struct, which needs to be explicitly loaded and configured. A number of generation options is model-specific and they used to be a part of model specification, but encapsulating everything in a single struct improves the transparency of options origin and reconfiguration. The text generation servings (generation, speech-to-text and conversation) need to be adjusted as follows:
{:ok, model_info} = Bumblebee.load_model({:hf, "gpt2"})
{:ok, tokenizer} = Bumblebee.load_tokenizer({:hf, "gpt2"})
+{:ok, generation_config} = Bumblebee.load_generation_config({:hf, "gpt2"})
+generation_config = Bumblebee.configure(generation_config, max_new_tokens: 100)
+serving = Bumblebee.Text.generation(model_info, tokenizer, generation_config)
-serving = Bumblebee.Text.generation(model_info, tokenizer, max_new_tokens: 100)Added
- Word-based aggregations for token classification (#174)
- BLIP model (#181)
- Text-to-image serving (#181)
- Generation option to avoid repeated n-grams (#182)
- Blenderbot model (#177)
- Option to load models from cache without outgoing traffic (#183)
- Whisper Phoenix demo (#184)
- Image channels normalization in featurizers (#189)
- T5 encoder model (#190)
- Contrastive search for sequence generation (#192)
- Multinomial sampling for sequence generation (#161)
- Support for loading sharded params checkpoints (#200)