Implementation:Cohere ai Cohere python FinetuningClient Monitoring

Field	Value
Type	Implementation
Source	Cohere Python SDK
Domain	Fine-tuning MLOps Monitoring
Last Updated	2026-02-15
Implements	Principle:Cohere_ai_Cohere_python_Training_Monitoring

Overview

Concrete methods for monitoring fine-tuning job status, events, and training metrics.

Description

Three monitoring methods on FinetuningClient: get_finetuned_model (returns current status and model details), list_events (returns paginated lifecycle events), list_training_step_metrics (returns paginated per-step training metrics). All take the finetuned_model_id as primary parameter.

Code Reference

• src/cohere/finetuning/client.py Lines L135-167 (get_finetuned_model)
• src/cohere/finetuning/client.py Lines L263-327 (list_events)
• src/cohere/finetuning/client.py Lines L329-379 (list_training_step_metrics)

Signature

def get_finetuned_model(self, id: str, *, request_options=None) -> GetFinetunedModelResponse:

def list_events(self, finetuned_model_id: str, *, page_size=None, page_token=None,
                order_by=None, request_options=None) -> ListEventsResponse:

def list_training_step_metrics(self, finetuned_model_id: str, *, page_size=None,
                               page_token=None, request_options=None) -> ListTrainingStepMetricsResponse:

Import

Access via client.finetuning.get_finetuned_model(), client.finetuning.list_events(), client.finetuning.list_training_step_metrics()

Inputs

Parameter	Type	Required	Description
id / finetuned_model_id	str	Yes	The ID of the fine-tuned model to monitor
page_size	Optional[int]	No	Number of results per page
page_token	Optional[str]	No	Pagination token for next page
order_by	Optional[str]	No	Ordering for events (list_events only)

Outputs

• GetFinetunedModelResponse -- current status and model details
• ListEventsResponse -- event timeline with timestamps and status changes
• ListTrainingStepMetricsResponse -- training loss and metrics per step

Example

import time

# Poll for completion
while True:
    status_response = client.finetuning.get_finetuned_model(model_id)
    status = status_response.finetuned_model.status
    print(f"Status: {status}")
    if status in ["STATUS_READY", "STATUS_FAILED"]:
        break
    time.sleep(60)

# Check training metrics
metrics = client.finetuning.list_training_step_metrics(model_id)
for step in metrics.step_metrics:
    print(f"Step {step.step_number}: loss={step.metrics}")

# View lifecycle events
events = client.finetuning.list_events(model_id)
for event in events.events:
    print(f"{event.created_at}: {event.status}")

Related

• Principle:Cohere_ai_Cohere_python_Training_Monitoring