Hyperparameters for Fine-Tuning a Model in Generative AI
OCI Generative AI fine-tunes each base model using the following hyperparameters, which are based on the pre-trained base model.
Tip
Start training each model with its default hyperparameter values. After the model is created, in the model's detail page, under Model Performance, check the values for accuracy and loss. If you're not happy with the results, create another model with either a larger dataset or different hyperparameters until the performance improves.
Start training each model with its default hyperparameter values. After the model is created, in the model's detail page, under Model Performance, check the values for accuracy and loss. If you're not happy with the results, create another model with either a larger dataset or different hyperparameters until the performance improves.
meta.llama-3.3-70b-instruct
The following table outlines the hyperparameters OCI
Generative AI uses to train a meta.llama-3.3-70b-instruct base model with the LoRA method.
| Hyperparameter | Description | Valid Range | Default Value |
|---|---|---|---|
| Total training epochs | The number of times the training iterates through the entire training dataset. For example, 1 epoch means that the model is trained by using the entire training dataset one time. |
1 or a higher integer |
3 |
| Learning rate | The speed at which the model weights are updated against the error gradient. | A number between 0 and 1.0 | 0.0002 |
| Training batch size | The number of samples in a mini batch to go through before updating the model's parameters. | An integer between 8 and 16 | 8 |
| Early stopping patience | Defines the number of grace periods to continue the evaluation cycle, after the early stopping threshold is triggered. Training stops if the loss metric doesn't improve beyond the early stopping threshold for this many times of evaluation. | 0 to disable and 1 or a higher integer to add a grace period | 15 |
| Early stopping threshold | Loss improves when it decreases in the next training cycle. If loss doesn't improve enough, you can stop the training. Define the minimum evaluation loss improvement that should trigger the early stopping counter. If loss doesn't improve beyond the minimum value during the patience period, training stops. Otherwise, training continues and the counter resets. | 0 or a positive number | 0.0001 |
| Log model metrics interval in steps | The number of steps per logging. Model metrics such as training loss and learning rate are logged. If the training loss is not decreasing as expected, review the training data or training rate. | Preset to 10 | 10 |
LoRA r (for LoRA method only) |
The attention dimension (rank) of the update matrices. A lower rank results in smaller update matrices with fewer trainable parameters. | An integer between 1 and 64 | 8 |
LoRA alpha (for LoRA method only) |
The alpha parameter for LoRA scaling. The LoRA weight matrices are scaled by dividing LoRA alpha by LoRA r. The alpha parameter defines the LoRA weights, which are a smaller number of new weights and are the only weights that are trained in the model. |
An integer between 1 and 128 | 8 |
LoRA dropout (For LoRA method only) |
The dropout probability for neurons in the LoRA layers. The dropout method prevents overfitting by randomly ignoring (dropping out) neurons within a layer. A 10% dropout means that each neuron has a 10% chance of being dropped. |
A decimal number less than 1 for percentage, such as 0.1 for 10% | 0.1 |
The following equation shows how the model calculates the
totalTrainingSteps parameter.totalTrainingSteps = (totalTrainingEpochs * size(trainingDataset)) / trainingBatchSizemeta.llama-3.1-70b-instruct
The following table outlines the hyperparameters OCI
Generative AI uses to train a meta.llama-3.1-70b-instruct base model with the LoRA method.
| Hyperparameter | Description | Valid Range | Default Value |
|---|---|---|---|
| Total training epochs | The number of times the training iterates through the entire training dataset. For example, 1 epoch means that the model is trained by using the entire training dataset one time. |
1 or a higher integer |
3 |
| Learning rate | The speed at which the model weights are updated against the error gradient. | A number between 0 and 1.0 | 0.0002 |
| Training batch size | The number of samples in a mini batch to go through before updating the model's parameters. | An integer between 8 and 16 | 8 |
| Early stopping patience | Defines the number of grace periods to continue the evaluation cycle, after the early stopping threshold is triggered. Training stops if the loss metric doesn't improve beyond the early stopping threshold for this many times of evaluation. | 0 to disable and 1 or a higher integer to add a grace period | 15 |
| Early stopping threshold | Loss improves when it decreases in the next training cycle. If loss doesn't improve enough, you can stop the training. Define the minimum evaluation loss improvement that should trigger the early stopping counter. If loss doesn't improve beyond the minimum value during the patience period, training stops. Otherwise, training continues and the counter resets. | 0 or a positive number | 0.0001 |
| Log model metrics interval in steps | The number of steps per logging. Model metrics such as training loss and learning rate are logged. If the training loss is not decreasing as expected, review the training data or training rate. | Preset to 10 | 10 |
LoRA r (for LoRA method only) |
The attention dimension (rank) of the update matrices. A lower rank results in smaller update matrices with fewer trainable parameters. | An integer between 1 and 64 | 8 |
LoRA alpha (for LoRA method only) |
The alpha parameter for LoRA scaling. The LoRA weight matrices are scaled by dividing LoRA alpha by LoRA r. The alpha parameter defines the LoRA weights, which are a smaller number of new weights and are the only weights that are trained in the model. |
An integer between 1 and 128 | 8 |
LoRA dropout (For LoRA method only) |
The dropout probability for neurons in the LoRA layers. The dropout method prevents overfitting by randomly ignoring (dropping out) neurons within a layer. A 10% dropout means that each neuron has a 10% chance of being dropped. |
A decimal number less than 1 for percentage, such as 0.1 for 10% | 0.1 |
The following equation shows how the model calculates the
totalTrainingSteps parameter.totalTrainingSteps = (totalTrainingEpochs * size(trainingDataset)) / trainingBatchSizemeta.llama-3-70b-instruct
The following table outlines the hyperparameters OCI
Generative AI uses to train a meta.llama-3-70b-instruct (deprecated) base model with the LoRA method.
| Hyperparameter | Description | Valid Range | Default Value |
|---|---|---|---|
| Total training epochs | The number of times the training iterates through the entire training dataset. For example, 1 epoch means that the model is trained by using the entire training dataset one time. |
1 or a higher integer |
3 |
| Learning rate | The speed at which the model weights are updated against the error gradient. | A number between 0 and 1.0 | 0.0002 |
| Training batch size | The number of samples in a mini batch to go through before updating the model's parameters. | An integer between 8 and 16 | 8 |
| Early stopping patience | Defines the number of grace periods to continue the evaluation cycle, after the early stopping threshold is triggered. Training stops if the loss metric doesn't improve beyond the early stopping threshold for this many times of evaluation. | 0 to disable and 1 or a higher integer to add a grace period | 15 |
| Early stopping threshold | Loss improves when it decreases in the next training cycle. If loss doesn't improve enough, you can stop the training. Define the minimum evaluation loss improvement that should trigger the early stopping counter. If loss doesn't improve beyond the minimum value during the patience period, training stops. Otherwise, training continues and the counter resets. | 0 or a positive number | 0.0001 |
| Log model metrics interval in steps | The number of steps per logging. Model metrics such as training loss and learning rate are logged. If the training loss is not decreasing as expected, review the training data or training rate. | Preset to 10 | 10 |
LoRA r (for LoRA method only) |
The attention dimension (rank) of the update matrices. A lower rank results in smaller update matrices with fewer trainable parameters. | An integer between 1 and 64 | 8 |
LoRA alpha (for LoRA method only) |
The alpha parameter for LoRA scaling. The LoRA weight matrices are scaled by dividing LoRA alpha by LoRA r. The alpha parameter defines the LoRA weights, which are a smaller number of new weights and are the only weights that are trained in the model. |
An integer between 1 and 128 | 8 |
LoRA dropout (For LoRA method only) |
The dropout probability for neurons in the LoRA layers. The dropout method prevents overfitting by randomly ignoring (dropping out) neurons within a layer. A 10% dropout means that each neuron has a 10% chance of being dropped. |
A decimal number less than 1 for percentage, such as 0.1 for 10% | 0.1 |
The following equation shows how the model calculates the
totalTrainingSteps parameter.totalTrainingSteps = (totalTrainingEpochs * size(trainingDataset)) / trainingBatchSizecohere.command-r-16k
(deprecated)
The following table outlines the hyperparameters OCI
Generative AI uses to train a cohere.command-r-16k (deprecated) base model with the T-Few method.
| Hyperparameter | Description | Valid Range | Default Value |
|---|---|---|---|
| Total training epochs | The number of times the training iterates through the entire training dataset.
For example, 1 epoch means that the model is trained by using the
entire training dataset one time. |
An integer between 1 and 10 |
1 |
| Learning rate | The speed at which the model weights are updated against the error gradient. | A number between 0.000005 and 0.1 | 0.01 |
| Training batch size | The number of samples in a mini batch to go through before updating the model's parameters. | An integer between 8 and 32 | 16 |
| Early stopping patience | Defines the number of grace periods to continue the evaluation cycle, after the early stopping threshold is triggered. Training stops if the loss metric doesn't improve beyond the early stopping threshold for this many times of evaluation. | 0 to disable and an integer between 1 and 16 to add a grace period | 10 |
| Early stopping threshold | Loss improves when it decreases in the next training cycle. If loss doesn't improve enough, you can stop the training. Define the minimum evaluation loss improvement that should trigger the early stopping counter. If loss doesn't improve beyond the minimum value during the patience period, training stops. Otherwise, training continues and the counter resets. | A number between 0.001 and 0.1 | 0.001 |
| Log model metrics interval in steps | The number of steps per logging. Model metrics such as training loss and learning rate are logged. If the training loss is not decreasing as expected, review the training data or training rate. | Can't be tuned and is set to 1. | 1 |
The following equation shows how the model calculates the
totalTrainingSteps
parameter.totalTrainingSteps = (totalTrainingEpochs * size(trainingDataset)) / trainingBatchSizecohere.command-r-08-2024
The following table outlines the hyperparameters OCI
Generative AI uses to train a cohere.command-r-16k base model with the T-Few method.
| Hyperparameter | Description | Valid Range | Default Value |
|---|---|---|---|
| Total training epochs | The number of times the training iterates through the entire training dataset. For example, 1 epoch means that the model is trained by using the entire training dataset one time. |
An integer between 1 and 10 |
1 |
| Learning rate | The speed at which the model weights are updated against the error gradient. | A number between 0.000005 and 0.1 | 0.01 |
| Training batch size | The number of samples in a mini batch to go through before updating the model's parameters. | An integer between 8 and 32 | 16 |
| Early stopping patience | Defines the number of grace periods to continue the evaluation cycle, after the early stopping threshold is triggered. Training stops if the loss metric doesn't improve beyond the early stopping threshold for this many times of evaluation. | 0 to disable and an integer between 1 and 16 to add a grace period | 10 |
| Early stopping threshold | Loss improves when it decreases in the next training cycle. If loss doesn't improve enough, you can stop the training. Define the minimum evaluation loss improvement that should trigger the early stopping counter. If loss doesn't improve beyond the minimum value during the patience period, training stops. Otherwise, training continues and the counter resets. | A number between 0.001 and 0.1 | 0.001 |
| Log model metrics interval in steps | The number of steps per logging. Model metrics such as training loss and learning rate are logged. If the training loss is not decreasing as expected, review the training data or training rate. | Can't be tuned and is set to 1. | 1 |
The following equation shows how the model calculates the
totalTrainingSteps parameter.totalTrainingSteps = (totalTrainingEpochs * size(trainingDataset)) / trainingBatchSizecohere.command
(deprecated)
The following table describes the hyperparameters that OCI
Generative AI uses to train the cohere.command (deprecated) base model and displays the default values for the T-Few and the Vanilla methods.
Note
The
The
cohere.command is available only in the US Midwest
(Chicago) region.| Hyperparameter | Description | Valid Range | Default Value for T-Few | Default Value for Vanilla |
|---|---|---|---|---|
| Total training epochs | The number of times the training iterates through the entire training dataset.
For example, 1 epoch means that the model is trained by using the
entire training dataset one time. |
1 or a higher integer |
3 | 3 |
| Learning rate | The speed at which the model weights are updated against the error gradient. | A number between 0 and 1.0 | 0.01 | 0.0000006 (6e-7) |
| Training batch size | The number of samples in a mini batch to go through before updating the model's parameters. | 8 | 8 | 8 |
| Early stopping patience | Defines the number of grace periods to continue the evaluation cycle, after the early stopping threshold is triggered. Training stops if the loss metric doesn't improve beyond the early stopping threshold for this many times of evaluation. | 0 to disable and 1 or a higher integer to add a grace period | 6 | 6 |
| Early stopping threshold | Loss improves when it decreases in the next training cycle. If loss doesn't improve enough, you can stop the training. Define the minimum evaluation loss improvement that should trigger the early stopping counter. If loss doesn't improve beyond the minimum value during the patience period, training stops. Otherwise, training continues and the counter resets. | 0 or a positive number | 0.01 | 0.01 |
| Log model metrics interval in steps | The number of steps per logging. Model metrics such as training loss and learning rate are logged. If the training loss is not decreasing as expected, review the training data or training rate. | 0 to disable and an integer between 1 and the total training steps to log. | 10 | 10 |
Number of last layers (for Vanilla method only) |
The number of last layers to be fine-tuned in the Vanilla
method. |
An integer between 1 and 15 | not applicable | 15 |
The following equation shows how the model calculates the
totalTrainingSteps
parameter.totalTrainingSteps = (totalTrainingEpochs * size(trainingDataset)) / trainingBatchSizecohere.command-light
(deprecated)
The following table describes the hyperparameters that OCI
Generative AI uses to train the cohere.command-light (deprecated) base model and displays the default values for the T-Few and the Vanilla methods.
Note
The
The
cohere.command-light is available only in the US Midwest
(Chicago) region.| Hyperparameter | Description | Valid Range | Default Value for T-Few | Default Value for Vanilla |
|---|---|---|---|---|
| Total training epochs | The number of times the training iterates through the entire training dataset.
For example, 1 epoch means that the model is trained by using the
entire training dataset one time. |
1 or a higher integer |
3 | 3 |
| Learning rate | The speed at which the model weights are updated against the error gradient. | A number between 0 and 1.0 | 0.01 | 0.00001 |
| Training batch size | The number of samples in a mini batch to go through before updating the model's parameters. | An integer between 8 and 16 | 16 | 16 |
| Early stopping patience | Defines the number of grace periods to continue the evaluation cycle, after the early stopping threshold is triggered. Training stops if the loss metric doesn't improve beyond the early stopping threshold for this many times of evaluation. | 0 to disable and 1 or a higher integer to add a grace period | 6 | 6 |
| Early stopping threshold | Loss improves when it decreases in the next training cycle. If loss doesn't improve enough, you can stop the training. Define the minimum evaluation loss improvement that should trigger the early stopping counter. If loss doesn't improve beyond the minimum value during the patience period, training stops. Otherwise, training continues and the counter resets. | 0 or a positive number | 0.01 | 0.01 |
| Log model metrics interval in steps | The number of steps per logging. Model metrics such as training loss and learning rate are logged. If the training loss is not decreasing as expected, review the training data or training rate. | 0 to disable and an integer between 1 and the total training steps to log. | 10 | 10 |
Number of last layers (for Vanilla method only) |
The number of last layers to be fine-tuned in the Vanilla
method. |
An integer between 1 and 14 | not applicable | 14 |
The following equation shows how the model calculates the
totalTrainingSteps
parameter.totalTrainingSteps = (totalTrainingEpochs * size(trainingDataset)) / trainingBatchSize