Gusthema comments

Results 13 comments of


                                            Gusthema

Preprocessor model can't be saved and loaded if we have only one input_segment

Hi Ian, sorry for the delay I've just tried saving a preprocess model with only one feature and it worked fine. Maybe I'm missing some part of your test. What...

'subprocess-exited-with-error' when installing gemma

Did the last suggestion help solve the problem or it still persists?

A web runtime supported version of gemma is really needed and high value

you can run Gemma models with MediaPipe in the web: https://ai.google.dev/edge/mediapipe/solutions/genai/llm_inference also, for quick testing, you can try the model inside Google AI Studio: https://aistudio.google.com/

Output with higher max_length is repetition of base text

it is expected these repetitions on PT models.it would be better to fine tune them or use the IT models

Colabs don't seem to work

@selamw1 I liked your colab! that would be a nice recipe for the Gemma cookbook: [goo.gle/gemma-cookbook](https://goo.gle/gemma-cookbook) For future users, we have some more tutorials using JAX + Gemma here: [ai.google.dev/gemma](https://ai.google.dev/gemma)

how can run, input a sentence, return results

this may be able to help: https://ai.google.dev/gemma/docs/jax_inference

After deplyed google/gemma-7b-it, there always is error response.

did the solution work?

How can I set temperatur of model when I reload Keras fine-tuned Gemma?

you can set the sampler for that, more information at the bottom in this tutorial: https://ai.google.dev/gemma/docs/keras_inference

Inconsistent 'query_pre_attn_scalar' Setting Between 9B and 27B Models

This change is to align the model better with the official internal implementation and these new values should be the correct one following the technical report [link to the technical...

Sliding Window Attention

We are working to release the Gemma 2 related code soon, sorry for the delay