feat(watsonx): add an optional params for min_new_tokens, repetition_…

…penalty & stop_sequences (watson-developer-cloud#178) * feat(watsonx): add an optional param for min_new_tokens * fix: default values * fix: session vars * feat: add new param repetition_penalty * feat: add another param Stopping Criteria * fix: nmber validation and ask 3 times * readme * fix: update openAPI spec to 1.0.1 & action fixes --------- Co-authored-by: Anirudha Bapat <[email protected]>
wanghaotokyo · Jul 21, 2023 · f95c1b7 · f95c1b7
1 parent ae9b77e
commit f95c1b7
Show file tree

Hide file tree

Showing 3 changed files with 246 additions and 24 deletions.
diff --git a/integrations/extensions/starter-kits/language-model-watsonx/README.md b/integrations/extensions/starter-kits/language-model-watsonx/README.md
@@ -67,12 +67,19 @@ These are the session variables used in this example.
 
 - `model_id`: The id of the watsonx model that you select for this action. Defaults to `google/flan-ul2`.
 - `model_input`: The input to the model.
-- `model_parameters_max_new_tokens` : The maximum number of new tokens to be generated. Defaults to 20.
-- `model_parameters_temperature` : The value used to control the next token probabilities. The range is from 0.05 to 1.00; 0.05 makes it _mostly_ deterministic.
+- `model_parameters_max_new_tokens` : The maximum number of new tokens to be generated. Defaults to 300.
+- `model_parameters_min_new_tokens`: The minimum number of the new tokens to be generated. Defaults to 1.
+- `model_parameters_temperature` : The value used to control the next token probabilities. The range is from 0.05 to 1.00; 0 makes it _mostly_ deterministic.
+- `model_parameters_repetition_penalty`: Represents the penalty for penalizing tokens that have already been generated or belong to the context. The range is 1.0 to 2.0 and defaults to 1.1.
+- `model_parameters_stop_sequences`: Stop sequences are one or more strings which will cause the text generation to stop if/when they are produced as part of the output. Stop sequences encountered prior to the minimum number of tokens being generated will be ignored. The list may contain up to 6 strings. Defaults to ["\n\n"]
 - `model_response`: The text generated by the model in response to the user input.
 - `watsonx_api_version` - watsonx api date version. It currently defaults to `2023-05-29`.
 - `watsonx_project_id`: You **MUST** set this value to be [a project ID value from watsonx](https://dataplatform.cloud.ibm.com/docs/content/wsj/manage-data/manage-projects.html). By default, this is a [sandbox project id](https://dataplatform.cloud.ibm.com/docs/content/wsj/manage-data/sandbox.html) that is automatically created for you when you sign up for watsonx.ai.
 
 Here is an example of how to use the `Test model` action:
 
 <img src="./assets/sample.png" width="300"/>
+
+### Limitations
+
+Adding large values for max_new_tokens & min_new_tokens can result in timeouts, if you see any timeouts please adjust the values of those tokens.
diff --git a/integrations/extensions/starter-kits/language-model-watsonx/watsonx-actions.json b/integrations/extensions/starter-kits/language-model-watsonx/watsonx-actions.json
@@ -3,10 +3,10 @@
  "type": "action",
  "valid": true,
  "status": "Available",
- "created": "2023-07-18T13:36:37.767Z",
- "updated": "2023-07-18T21:22:48.224Z",
+ "created": "2023-07-21T23:27:18.637Z",
+ "updated": "2023-07-21T23:31:17.833Z",
  "language": "en",
- "skill_id": "9c3433c1-f60d-4f0a-9648-9b43a23100de",
+ "skill_id": "096d12a9-3b20-43ac-bc18-9d2e4c79ec1a",
  "workspace": {
  "actions": [
  {
@@ -112,7 +112,8 @@
  }
  ],
  "response_type": "text",
- "selection_policy": "sequential"
+ "selection_policy": "sequential",
+ "repeat_on_reprompt": true
  }
  ]
  },
@@ -128,10 +129,17 @@
  {
  "values": [
  {
- "text": "I didn't catch that. Enter a number."
+ "text_expression": {
+ "concat": [
+ {
+ "scalar": "I didn't catch that. Enter a number between 1 to 1024"
+ }
+ ]
+ }
  }
  ],
- "response_type": "text"
+ "response_type": "text",
+ "selection_policy": "incremental"
  }
  ]
  },
@@ -150,7 +158,7 @@
  {
  "name": "maximum",
  "value": {
- "scalar": 4000
+ "scalar": 1024
  }
  }
  ]
@@ -163,10 +171,17 @@
  {
  "values": [
  {
- "text": "I didn't catch that. Enter a number."
+ "text_expression": {
+ "concat": [
+ {
+ "scalar": "I didn't catch that. Enter a number between 1 to 1024"
+ }
+ ]
+ }
  }
  ],
- "response_type": "text"
+ "response_type": "text",
+ "selection_policy": "incremental"
  }
  ]
  },
@@ -193,6 +208,118 @@
  "type": "continue"
  },
  "variable": "step_333",
+ "next_step": "step_577"
+ },
+ {
+ "step": "step_577",
+ "output": {
+ "generic": [
+ {
+ "values": [
+ {
+ "text_expression": {
+ "concat": [
+ {
+ "scalar": "How many minimum tokens do you want to generate?"
+ }
+ ]
+ }
+ }
+ ],
+ "response_type": "text",
+ "selection_policy": "sequential",
+ "repeat_on_reprompt": false
+ }
+ ]
+ },
+ "handlers": [
+ {
+ "type": "input_validation",
+ "title": "input_validation_handler_1",
+ "output": {
+ "generic": [
+ {
+ "values": [
+ {
+ "text_expression": {
+ "concat": [
+ {
+ "scalar": "I didn't catch that. Enter a number between 0 to 1024"
+ }
+ ]
+ }
+ }
+ ],
+ "response_type": "text",
+ "selection_policy": "incremental"
+ }
+ ]
+ },
+ "handler": "input_validation_handler_1",
+ "resolver": {
+ "type": "prompt_again"
+ },
+ "next_handler": "validation_not_found_handler",
+ "input_constraints": [
+ {
+ "name": "maximum",
+ "value": {
+ "scalar": 1024
+ }
+ },
+ {
+ "name": "minimum",
+ "value": {
+ "scalar": 0
+ }
+ }
+ ]
+ },
+ {
+ "type": "not_found",
+ "title": "validation_not_found_handler",
+ "output": {
+ "generic": [
+ {
+ "values": [
+ {
+ "text_expression": {
+ "concat": [
+ {
+ "scalar": "I didn't catch that. Enter a number between 0 to 1024"
+ }
+ ]
+ }
+ }
+ ],
+ "response_type": "text",
+ "selection_policy": "incremental"
+ }
+ ]
+ },
+ "handler": "validation_not_found_handler",
+ "resolver": {
+ "type": "prompt_again"
+ },
+ "next_handler": "validation_not_found_max_tries_handler"
+ },
+ {
+ "type": "not_found_max_tries",
+ "title": "validation_not_found_max_tries_handler",
+ "handler": "validation_not_found_max_tries_handler",
+ "resolver": {
+ "type": "fallback"
+ }
+ }
+ ],
+ "question": {
+ "entity": "sys-number",
+ "max_tries": 3
+ },
+ "resolver": {
+ "type": "continue"
+ },
+ "variable": "step_577",
  "next_step": "step_605"
  },
  {
@@ -212,7 +339,8 @@
  }
  ],
  "response_type": "text",
- "selection_policy": "sequential"
+ "selection_policy": "sequential",
+ "repeat_on_reprompt": false
  }
  ]
  },
@@ -225,10 +353,17 @@
  {
  "values": [
  {
- "text": "I didn't catch that. Enter a number."
+ "text_expression": {
+ "concat": [
+ {
+ "scalar": "I didn't catch that. Enter a number."
+ }
+ ]
+ }
  }
  ],
- "response_type": "text"
+ "response_type": "text",
+ "selection_policy": "incremental"
  }
  ]
  },
@@ -260,10 +395,17 @@
  {
  "values": [
  {
- "text": "I didn't catch that. Enter a number."
+ "text_expression": {
+ "concat": [
+ {
+ "scalar": "I didn't catch that. Enter a number."
+ }
+ ]
+ }
  }
  ],
- "response_type": "text"
+ "response_type": "text",
+ "selection_policy": "incremental"
  }
  ]
  },
@@ -346,6 +488,12 @@
  },
  "skill_variable": "model_parameters_max_new_tokens"
  },
+ {
+ "value": {
+ "variable": "step_577"
+ },
+ "skill_variable": "model_parameters_min_new_tokens"
+ },
  {
  "value": {
  "variable": "step_605"
@@ -365,6 +513,7 @@
  "type": "invoke_another_action",
  "invoke_action": {
  "action": "action_3200-2",
+ "policy": "default",
  "parameters": null,
  "result_variable": "step_817_result_1"
  }
@@ -429,6 +578,11 @@
  "variable": "step_333",
  "data_type": "number"
  },
+ {
+ "title": "How many minimum tokens do you want to generate?",
+ "variable": "step_577",
+ "data_type": "number"
+ },
  {
  "title": "What temperature do you want to use?",
  "variable": "step_605",
@@ -475,8 +629,8 @@
  "type": "integration_interaction",
  "method": "POST",
  "internal": {
- "spec_hash_id": "f22a973be9c24f524794f9f86d114b5fe9e4e1168ae47ea950e4d7889e972125",
- "catalog_item_id": "3a1fabde-51ff-42cc-ad27-23b965370021"
+ "spec_hash_id": "efd13c68dff079388ffa62bfc3aff447e49efd7024450961cc74d6de89106cce",
+ "catalog_item_id": "f53a6acc-4505-45a5-a785-27fe357b96da"
  },
  "request_mapping": {
  "body": [
@@ -509,6 +663,24 @@
  "skill_variable": "model_parameters_max_new_tokens"
  },
  "parameter": "parameters.max_new_tokens"
+ },
+ {
+ "value": {
+ "skill_variable": "model_parameters_min_new_tokens"
+ },
+ "parameter": "parameters.min_new_tokens"
+ },
+ {
+ "value": {
+ "skill_variable": "model_parameters_stop_sequences"
+ },
+ "parameter": "parameters.stop_sequences"
+ },
+ {
+ "value": {
+ "skill_variable": "model_parameters_repetition_penalty"
+ },
+ "parameter": "parameters.repetition_penalty"
  }
  ],
  "query": [
@@ -678,6 +850,7 @@
  "type": "invoke_another_action_and_end",
  "invoke_action": {
  "action": "fallback",
+ "policy": "default",
  "parameters": null,
  "result_variable": "step_002_result_1"
  }
@@ -1241,7 +1414,34 @@
  "data_type": "number",
  "description": "The maximum number of new tokens to be generated.",
  "initial_value": {
- "scalar": 20
+ "scalar": 300
+ }
+ },
+ {
+ "title": "model_parameters_min_new_tokens",
+ "variable": "model_parameters_min_new_tokens",
+ "data_type": "number",
+ "description": "",
+ "initial_value": {
+ "scalar": 1
+ }
+ },
+ {
+ "title": "model_parameters_repetition_penalty",
+ "variable": "model_parameters_repetition_penalty",
+ "data_type": "number",
+ "description": "",
+ "initial_value": {
+ "scalar": 1.1
+ }
+ },
+ {
+ "title": "model_parameters_stop_sequences",
+ "variable": "model_parameters_stop_sequences",
+ "data_type": "any",
+ "description": "",
+ "initial_value": {
+ "expression": "[\"\\n\\n\"]"
  }
  },
  {
@@ -1250,7 +1450,7 @@
  "data_type": "number",
  "description": "The value used to module the next token probabilities. The range is 0.05 to 1.00, a value set to 0.05 would make it mostly deterministic.",
  "initial_value": {
- "scalar": 0.05
+ "scalar": 0
  }
  },
  {
@@ -1380,8 +1580,8 @@
  "learning_opt_out": true
  },
  "description": "created for assistant 05c10d7d-336f-4d33-8cb3-5c53520d61ce",
- "assistant_id": "3b229721-63a8-4e2e-8aa1-380654a5cdb0",
- "workspace_id": "9c3433c1-f60d-4f0a-9648-9b43a23100de",
+ "assistant_id": "69f5847f-17c6-4cb0-bafe-e1c3e5f7cfd2",
+ "workspace_id": "096d12a9-3b20-43ac-bc18-9d2e4c79ec1a",
  "dialog_settings": {},
  "next_snapshot_version": "1"
 }