Skip to content

Commit

Permalink
feat: update oasst model (#411)
Browse files Browse the repository at this point in the history 
* feat: update oasst model

* fix: model descriptions (#410)

* fix: sync dream_persona with demos (#413)

* Feat/llm-based and ranking-based dm (#405)

* feat: create files for llm based response selection

* feat: tests

* feat: algorithm

* fix: filter bad or toxic uttrs

* feat: ranking based response selector

* feat: model cards

* feat: ranking and refactoring

* feat: component files

* feat: component description

* feat: component description

* feat: use gpt-j resp selector

* feat: context

* feat: utilize hypotheses correctly

* fix: configs

* fix: configs

* fix: mapping

* fix: utilize ranker based selector

* fix: decay dummy

* fix: decay dummy

* fix: response

* fix: remove old selector

* fix: logs

* fix: getting annotations

* fix: code style

* fix: deepy uses resp selector

* fix: no language

* fix: ru usage

* fix: usage of formatter

* fix: codestyle

* fix: logs

* fix: after review

* fix: sync

* fix: add proxy usage (#416)

* feat: new proxy address (#417)

* prompt asks chatbot to be concise (otherwise problems with too long answers) (#418)

* added gptj to proxy file (#424)

* added gptj to proxy file

* return newline

* fix:dummy score decrease (#423)

* Feat/api keys from attributes (#415)

* fix: add attributes to formatter

* fix: try to get env vars from user utterance

* fix: none var values

* fix: logs correct

* fix: send lowercased keys

* prompt to ask fewer questions (#429)

* feat: universal dist requires api keys in request (#425)

* feat: template_template (#414)

* feat: template_template

* feat: readme

* fix: add chatgpt

* fix: tempalte does not use secret env

* fixed cutoff for AI utterance (#426)

* fixed cutoff for AI utterance

* fix for cases with extra ROBOT: etc

* style

* fix for newline

* feat: list of public dists (#433)

* fix: remove badlists from prompted dists (#431)

* added info about payments (#432)

* added info about payments

* better descs + reworked fashion prompt

* table update

* typo fix

* typo fix

* typo

* also increase fashion timeout in yml files

* forgot one file

* fix: no beauty in table

---------

Co-authored-by: dilyararimovna <dilyara.rimovna@gmail.com>

* timeout 30 for fairytales (#427)

* longer generation and timeout to reduce cutoffs (#420)

* longer generation and timeout to reduce cutoffs

* also updated in comp and pipe files

* change prompts for ai faq and nutrition (#430)

* reworked prompts to perform well with oasst

* solved conflict

* fix: tempalte dist and description (#435)

* Feat/multiskill assistant (#434)

* feat: multiskill assistant

* fix: prompt for meeting notes

* fix: waiting for

* fix: formatters

* fix: dot

* fix: secrets

* fix: add dream persona skill

* fix: add dream persona skill

* fix: add to public dists

* fix: folder name

* fix: description

* fix: component cards

* fix: component cards

* feat: thresholds for toxic_cls (#437)

* feat: thresholds for toxic_cls

* fix: codestyle

* fix: update pip (#439)

* fix: urllib neuralcoref

* fix: update pip

* fix: update pip in spacy annotator

* feat: smaller context for prompt selector (#438)

* feat: smaller context for prompt selector

* fix: index error

* Add compose definitions to component cards (#384)

* Fix requirements.txt (#84)

* fix itsdangerous requirements

* pin itsdangerous requirements for all flask==1.1.1 servers

* add compose definitions to component cards

* add host, port, endpoint to pipeline cards

* fix authors, add missing pipelines, add template

* add prompt-selector template, fix templates

* fix template keys

* remove unused comet-atomic component

* rework service, component cards

* move build args and env vars to environment.yml file

* fix empty proxy values

* fix pipeline confs

* fix component connector.annotations

* fix wait hosts in response_annotator_selectors service

* fix wait hosts in other services, environments

* fix response selectors, pipelines

* fix ports in response selector

* fix generative components

* fix: remove spelling and badlists from dream_persona_prompted

* fix: ignore all env_secret

* fix: new formatters for dff

* fix: universal dist does not utilize env_secret

* fix: multiskill components and timeouts

* fix: remove template_template distribution

* fix: deeppavlov assistant

* fix: formatters in components

* fix: volumes for resp selectors

* fix: correct filenames for multiskill

* fix: rullama7bru distribution

* fix pipelines after merge

* fix sentseg annotator source in pipelines

* fix agent services, components

* fix: sentseg and prompt selector do not wait for spelling

* fix: response_annotator_selectors card

* fix: timeouts

* fix: build args

* create services, components for prompt selector

* fix prompt selectors environment

* fix: flask server

* fix: path to prompt selector

* fix: required groups

* fix: required group skill_selectors

* fix: required elements

* fix: previous services

* fix: correct link to sentseg components card

* fix: correct link to sentseg components card

* remove unused prompt selector component

* remove old configs

* fix: rename files without dashes

---------

Co-authored-by: Andrii.Hura <54397922+AndriiHura@users.noreply.github.com>
Co-authored-by: Dilyara Baymurzina <dilyara.rimovna@gmail.com>

* Fix/remove duplicates (#443)

* fix: remove duplicating files

* fix: remove duplicating files

* Multiskill update (#440)

* long gen, long timeout, more prompts

* prompt update

* timeouts

* added longer context to persona

* longer max_tokens

* fix: increase to 20 sec

* fix: used config

* fix: dream persona openai

* fix:multiskill components

---------

Co-authored-by: dilyararimovna <dilyara.rimovna@gmail.com>

* Fix/increase all timeouts (#444)

* fix: increase all timeouts

* fix: increase all timeouts in pipelines

* fix: 7sec to 20 also

* fix: 5sec to 20 also for dream persona

* fix: rebase and change to new model

* fix: upd model

* fix: display name

---------

Co-authored-by: Nika Smilga <42929200+smilni@users.noreply.github.com>
Co-authored-by: Maxim Talimanchuk <mtalimanchuk@gmail.com>
Co-authored-by: Andrii.Hura <54397922+AndriiHura@users.noreply.github.com>
  • Loading branch information
@dilyararimovna @smilni
@mtalimanchuk @AndriiHura
4 people authored May 12, 2023
1 parent 1f3ff38 commit 77029f9
Show file tree
Hide file tree
Showing 6 changed files with 14 additions and 14 deletions.
18 changes: 9 additions & 9 deletions MODELS.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,12 +2,12 @@

Here you may find a list of models that currently available for use in Generative Assistants.

| model name | container name | model link | open-source? | size (billion parameters) | GPU usage | max tokens (prompt + response) | description |
|---------------------------|--------------------------|----------------------------------------------------------------------|--------------------------|---------------------------|---------------------------|--------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| BLOOMZ 7B | transformers-lm-bloomz7b | [link](https://huggingface.co/bigscience/bloomz-7b1) | yes | 7.1B | 33GB | 2,048 tokens | An open-source multilingual instruction-based large language model (46 languages). NB: free of charge. This model is up and running on our servers and can be used for free. |
| GPT-J 6B | transformers-lm-gptj | [link](https://huggingface.co/EleutherAI/gpt-j-6b) | yes | 6B | 25GB | 2,048 tokens | An open-source English-only large language model which is NOT fine-tuned for instruction following and NOT capable of code generation. NB: free of charge. This model is up and running on our servers and can be used for free. |
| GPT-3.5 | openai-api-davinci3 | [link](https://platform.openai.com/docs/models/gpt-3-5) | no (paid access via API) | supposedly, 175B | - (cannot be run locally) | 4,097 tokens | A multulingual instruction-based large language model which is capable of code generation. Unlike ChatGPT, not optimised for chat. NB: paid. You must provide your OpenAI API key to use the model. Your OpenAI account will be charged according to your usage. |
| ChatGPT | openai-api-chatgpt | [link](https://platform.openai.com/docs/models/gpt-3-5) | no (paid access via API) | supposedly, 175B | - (cannot be run locally) | 4,096 tokens | Based on gpt-3.5-turbo -- the most capable of the entire GPT-3/GPT-3.5 models family. Optimized for chat. Able to understand and generate code. NB: paid. You must provide your OpenAI API key to use the model. Your OpenAI account will be charged according to your usage. |
| Open-Assistant SFT-1 12B | transformers-lm-oasst12b | [link](https://huggingface.co/OpenAssistant/oasst-sft-1-pythia-12b) | yes | 12B | 26GB (half-precision) | 5,120 tokens | An open-source English-only instruction-based large language model which is NOT good at answering math and coding questions. NB: free of charge. This model is up and running on our servers and can be used for free. |
| GPT-4 | openai-api-gpt4 | [link](https://platform.openai.com/docs/models/gpt-4) | no (paid access via API) | supposedly, 175B | - (cannot be run locally) | 8,192 tokens | A multilingual instruction-based large language model which is capable of code generation and other complex tasks. More capable than any GPT-3.5 model, able to do more complex tasks, and optimized for chat. NB: paid. You must provide your OpenAI API key to use the model. Your OpenAI account will be charged according to your usage. |
| GPT-4 32K | openai-api-gpt4-32k | [link](https://platform.openai.com/docs/models/gpt-4) | no (paid access via API) | supposedly, 175B | - (cannot be run locally) | 32,768 tokens | A multilingual instruction-based large language model which is capable of code generation and other complex tasks. Same capabilities as the base gpt-4 mode but with 4x the context length. NB: paid. You must provide your OpenAI API key to use the model. Your OpenAI account will be charged according to your usage. |
| model name | container name | model link | open-source? | size (billion parameters) | GPU usage | max tokens (prompt + response) | description |
|---------------------------|--------------------------|-------------------------------------------------------------------------|--------------------------|---------------------------|---------------------------|--------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| BLOOMZ 7B | transformers-lm-bloomz7b | [link](https://huggingface.co/bigscience/bloomz-7b1) | yes | 7.1B | 33GB | 2,048 tokens | An open-source multilingual instruction-based large language model (46 languages). NB: free of charge. This model is up and running on our servers and can be used for free. |
| GPT-J 6B | transformers-lm-gptj | [link](https://huggingface.co/EleutherAI/gpt-j-6b) | yes | 6B | 25GB | 2,048 tokens | An open-source English-only large language model which is NOT fine-tuned for instruction following and NOT capable of code generation. NB: free of charge. This model is up and running on our servers and can be used for free. |
| GPT-3.5 | openai-api-davinci3 | [link](https://platform.openai.com/docs/models/gpt-3-5) | no (paid access via API) | supposedly, 175B | - (cannot be run locally) | 4,097 tokens | A multulingual instruction-based large language model which is capable of code generation. Unlike ChatGPT, not optimised for chat. NB: paid. You must provide your OpenAI API key to use the model. Your OpenAI account will be charged according to your usage. |
| ChatGPT | openai-api-chatgpt | [link](https://platform.openai.com/docs/models/gpt-3-5) | no (paid access via API) | supposedly, 175B | - (cannot be run locally) | 4,096 tokens | Based on gpt-3.5-turbo -- the most capable of the entire GPT-3/GPT-3.5 models family. Optimized for chat. Able to understand and generate code. NB: paid. You must provide your OpenAI API key to use the model. Your OpenAI account will be charged according to your usage. |
| Open-Assistant Pythia 12B | transformers-lm-oasst12b | [link](https://huggingface.co/OpenAssistant/pythia-12b-sft-v8-7k-steps) | yes | 12B | 26GB (half-precision) | 5,120 tokens | An open-source English-only instruction-based large language model which is NOT good at answering math and coding questions. NB: free of charge. This model is up and running on our servers and can be used for free. |
| GPT-4 | openai-api-gpt4 | [link](https://platform.openai.com/docs/models/gpt-4) | no (paid access via API) | supposedly, 175B | - (cannot be run locally) | 8,192 tokens | A multilingual instruction-based large language model which is capable of code generation and other complex tasks. More capable than any GPT-3.5 model, able to do more complex tasks, and optimized for chat. NB: paid. You must provide your OpenAI API key to use the model. Your OpenAI account will be charged according to your usage. |
| GPT-4 32K | openai-api-gpt4-32k | [link](https://platform.openai.com/docs/models/gpt-4) | no (paid access via API) | supposedly, 175B | - (cannot be run locally) | 32,768 tokens | A multilingual instruction-based large language model which is capable of code generation and other complex tasks. Same capabilities as the base gpt-4 mode but with 4x the context length. NB: paid. You must provide your OpenAI API key to use the model. Your OpenAI account will be charged according to your usage. |
2 changes: 1 addition & 1 deletion assistant_dists/ai_faq_assistant/docker-compose.override.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -107,7 +107,7 @@ services:
args:
SERVICE_PORT: 8158
SERVICE_NAME: transformers_lm_oasst12b
PRETRAINED_MODEL_NAME_OR_PATH: OpenAssistant/oasst-sft-1-pythia-12b
PRETRAINED_MODEL_NAME_OR_PATH: OpenAssistant/pythia-12b-sft-v8-7k-steps
HALF_PRECISION: 1
context: .
dockerfile: ./services/transformers_lm/Dockerfile
Expand Down
2 changes: 1 addition & 1 deletion assistant_dists/fashion_stylist_assistant/docker-compose.override.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -107,7 +107,7 @@ services:
args:
SERVICE_PORT: 8158
SERVICE_NAME: transformers_lm_oasst12b
PRETRAINED_MODEL_NAME_OR_PATH: OpenAssistant/oasst-sft-1-pythia-12b
PRETRAINED_MODEL_NAME_OR_PATH: OpenAssistant/pythia-12b-sft-v8-7k-steps
HALF_PRECISION: 1
context: .
dockerfile: ./services/transformers_lm/Dockerfile
Expand Down
2 changes: 1 addition & 1 deletion assistant_dists/nutrition_assistant/docker-compose.override.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -107,7 +107,7 @@ services:
args:
SERVICE_PORT: 8158
SERVICE_NAME: transformers_lm_oasst12b
PRETRAINED_MODEL_NAME_OR_PATH: OpenAssistant/oasst-sft-1-pythia-12b
PRETRAINED_MODEL_NAME_OR_PATH: OpenAssistant/pythia-12b-sft-v8-7k-steps
HALF_PRECISION: 1
context: .
dockerfile: ./services/transformers_lm/Dockerfile
Expand Down
2 changes: 1 addition & 1 deletion assistant_dists/universal_prompted_assistant/docker-compose.override.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -110,7 +110,7 @@ services:
args:
SERVICE_PORT: 8158
SERVICE_NAME: transformers_lm_oasst12b
PRETRAINED_MODEL_NAME_OR_PATH: OpenAssistant/oasst-sft-1-pythia-12b
PRETRAINED_MODEL_NAME_OR_PATH: OpenAssistant/pythia-12b-sft-v8-7k-steps
HALF_PRECISION: 1
context: .
dockerfile: ./services/transformers_lm/Dockerfile
Expand Down
2 changes: 1 addition & 1 deletion components/sdkajfhsidhf8wfjh2ornfkle.yml
View file
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
name: transformers_lm_oasst12b
display_name: Open-Assistant SFT-1 12B
display_name: Open-Assistant Pythia 12B
component_type: Generative
model_type: NN-based
is_customizable: false
Expand Down

0 comments on commit 77029f9

Please sign in to comment.