i need help , v0.2.8 ,Frequently occurs, A process in the process pool was terminated abruptly while the future was running or pending.

[mushuang]

2025-08-30 00:51:38,229 - ERROR - Error processing result from iteration 323: A process in the process pool was terminated abruptly while the future was running or pending.
2025-08-30 00:51:38,237 - ERROR - Error submitting iteration 353: A child process terminated abruptly, the process pool is not usable anymore
2025-08-30 00:51:38,237 - ERROR - Error processing result from iteration 337: A process in the process pool was terminated abruptly while the future was running or pending.
2025-08-30 00:51:38,246 - ERROR - Error submitting iteration 353: A child process terminated abruptly, the process pool is not usable anymore
2025-08-30 00:51:38,254 - ERROR - Error submitting iteration 353: A child process terminated abruptly, the process pool is not usable anymore
2025-08-30 00:51:38,254 - ERROR - Error processing result from iteration 339: A process in the process pool was terminated abruptly while the future was running or pending.
2025-08-30 00:51:38,258 - ERROR - Error submitting iteration 353: A child process terminated abruptly, the process pool is not usable anymore
2025-08-30 00:51:38,264 - ERROR - Error submitting iteration 353: A child process terminated abruptly, the process pool is not usable anymore
2025-08-30 00:51:38,270 - ERROR - Error submitting iteration 353: A child process terminated abruptly, the process pool is not usable anymore
2025-08-30 00:51:38,270 - ERROR - Error processing result from iteration 340: A process in the process pool was terminated abruptly while the future was running or pending.

[codelion]

what is the config.yaml you were using? Is it an existing example?

[mushuang]

what is the config.yaml you were using? Is it an existing example?

hi，this is my config：

`# Configuration for function minimization example
max_iterations: 500
checkpoint_interval: 1
log_level: "INFO"

Evolution settings

diff_based_evolution: true # Use diff-based evolution (true) or full rewrites (false)

max_code_length: 30000 # Maximum allowed code length in characters

LLM configuration

llm:

Models for evolution

models:
# List of available models with their weights
- name: "deepseek-reasoner"
weight: 0.8
- name: "deepseek-reasoner"
weight: 0.2

Models for LLM feedback

evaluator_models:
# List of available models with their weights
- name: "deepseek-reasoner"
weight: 0.8
- name: "deepseek-reasoner"
weight: 0.2

api_base: "https://api.deepseek.com/v1"
api_key: null

temperature: 0.7
top_p: 0.95
max_tokens: 4096

Request parameters

timeout: 1200 # Timeout for API requests in seconds
retries: 5 # Number of retries for failed requests
retry_delay: 5 # Delay between retries in seconds

Prompt configuration

prompt:
system_message: |
One, Identity
You are a highly experienced programmer specializing in optimization algorithms, a top-tier Ph.D. in algorithms.
You are also a senior quantitative trading strategy expert proficient in all trading techniques throughout human history,
with exceptional analytical and innovative capabilities.
......... more .........

num_top_programs: 3
use_template_stochasticity: true

Database configuration

database:
population_size: 70
archive_size: 20
num_islands: 8
elite_selection_ratio: 0.2
exploitation_ratio: 0.7

You can mix built-in features with custom metrics from your evaluator:

feature_dimensions: # Dimensions for MAP-Elites feature map (for diversity, NOT fitness)
- "complexity" # Code length (built-in)
- "diversity" # Code diversity (built-in)

Evaluator configuration

evaluator:

General settings

timeout: 600 # Maximum evaluation time in seconds
max_retries: 5 # Maximum number of retries for evaluation

Evaluation strategies

cascade_evaluation: false # Use cascade evaluation to filter bad solutions early
cascade_thresholds: # Thresholds for advancing to next evaluation stage
- 0.5 # First stage threshold
- 0.75 # Second stage threshold
- 0.9 # Third stage threshold
parallel_evaluations: 16
use_llm_feedback: false

Evolution settings

diff_based_evolution: false
allow_full_rewrites: false
`

[codelion]

I will try and add a fix by setting a timeout in future.result(timeout=timeout_seconds)

[mushuang]

@codelion hi,
at v0.2.11 , occurs, only change " parallel_evaluations: 4" -> parallel_evaluations: 16

025-09-02 03:47:17,490 - INFO - Saved checkpoint at iteration 322 to examples/quantitative_trading\openevolve_output\checkpoints\checkpoint_322
2025-09-02 03:47:22,901 - ERROR - Error processing result from iteration 327: A process in the process pool was terminated abruptly while the future was running or pending.
2025-09-02 03:47:22,907 - ERROR - Error submitting iteration 361: A child process terminated abruptly, the process pool is not usable anymore
2025-09-02 03:47:22,907 - ERROR - Error processing result from iteration 329: A process in the process pool was terminated abruptly while the future was running or pending.
2025-09-02 03:47:22,913 - ERROR - Error submitting iteration 361: A child process terminated abruptly, the process pool is not usable anymore
2025-09-02 03:47:22,919 - ERROR - Error submitting iteration 361: A child process terminated abruptly, the process pool is not usable anymore
2025-09-02 03:47:22,920 - ERROR - Error processing result from iteration 331: A process in the process pool was terminated abruptly while the future was running or pending.

[codelion]

Do you also set the number of islands to 16 when using parallel_evaluations: 16?

[mushuang]

Do you also set the number of islands to 16 when using parallel_evaluations: 16?

well，i set to 8

database:
population_size: 70
archive_size: 20
num_islands: 8

[mushuang]

Do you also set the number of islands to 16 when using parallel_evaluations: 16?

this is my current config：

Configuration for function minimization example

max_iterations: 500
checkpoint_interval: 1
log_level: "INFO"

Evolution settings

diff_based_evolution: true # Use diff-based evolution (true) or full rewrites (false)

max_code_length: 30000 # Maximum allowed code length in characters

LLM configuration

llm:

Models for evolution

models:
# List of available models with their weights
- name: "deepseek-reasoner"
weight: 0.8
- name: "deepseek-reasoner"
weight: 0.2

Models for LLM feedback

evaluator_models:
# List of available models with their weights
- name: "deepseek-reasoner"
weight: 0.8
- name: "deepseek-reasoner"
weight: 0.2

api_base: "https://api.deepseek.com/v1"
api_key: null

temperature: 0.7
top_p: 0.95
max_tokens: 4096

Request parameters

timeout: 1200 # Timeout for API requests in seconds
retries: 5 # Number of retries for failed requests
retry_delay: 5 # Delay between retries in seconds

Prompt configuration

prompt:
system_message: |
One, Identity
You are a highly experienced programmer specializing in optimization algorithms, a top-tier Ph.D. in algorithms.
You are also a senior quantitative trading strategy expert proficient in all trading techniques throughout human history,
with exceptional analytical and innovative capabilities.
（....more）

num_top_programs: 3
use_template_stochasticity: true

Database configuration

database:
population_size: 70
archive_size: 20
num_islands: 8
elite_selection_ratio: 0.2
exploitation_ratio: 0.7

You can mix built-in features with custom metrics from your evaluator:

feature_dimensions: # Dimensions for MAP-Elites feature map (for diversity, NOT fitness)
- "complexity" # Code length (built-in)
- "diversity" # Code diversity (built-in)

Evaluator configuration

evaluator:

General settings

timeout: 600 # Maximum evaluation time in seconds
max_retries: 5 # Maximum number of retries for evaluation

Evaluation strategies

cascade_evaluation: false # Use cascade evaluation to filter bad solutions early
cascade_thresholds: # Thresholds for advancing to next evaluation stage
- 0.5 # First stage threshold
- 0.75 # Second stage threshold
- 0.9 # Third stage threshold
parallel_evaluations: 16
use_llm_feedback: false

Evolution settings

diff_based_evolution: false
allow_full_rewrites: false

[codelion]

Ok, I will take a look at it.

Can you try with 16 islands?

The challenge is when we have parallel_evalution how do we decide from which island to sample the programs from, if the number of parallel evaluators is more than the number of islands we will end up sampling from the same island. We try to pin the evaluter to a particular island to maintain isolation between the islands. There may be some race condition that is causing this failure. Running with 16 should show in logs that each island gets a single worker and may avoid it.

[mushuang]

Ok, I will take a look at it.

Can you try with 16 islands?

The challenge is when we have parallel_evalution how do we decide from which island to sample the programs from, if the number of parallel evaluators is more than the number of islands we will end up sampling from the same island. We try to pin the evaluter to a particular island to maintain isolation between the islands. There may be some race condition that is causing this failure. Running with 16 should show in logs that each island gets a single worker and may avoid it.

ok，thank you very much。I try it at once。

[mushuang]

Ok, I will take a look at it.

Can you try with 16 islands?

The challenge is when we have parallel_evalution how do we decide from which island to sample the programs from, if the number of parallel evaluators is more than the number of islands we will end up sampling from the same island. We try to pin the evaluter to a particular island to maintain isolation between the islands. There may be some race condition that is causing this failure. Running with 16 should show in logs that each island gets a single worker and may avoid it.

@codelion
hi，It still happened，when num_islands: 12 and parallel_evaluations: 12。openevolve version is v0.2.11.
my system info：
intel(R) Core(TM) i9-14900K 3.20 GHz 128 GB 64 , x64 Windows 10

this is log：
[STDERR] 2025-09-03 01:39:28,043 - INFO - Saved checkpoint at iteration 251 to quantitative_trading\openevolve_output\checkpoints\checkpoint_251 [STDERR] 2025-09-03 01:39:30,206 - ERROR - Error processing result from iteration 249: A process in the process pool was terminated abruptly while the future was running or pending. [STDERR] 2025-09-03 01:39:30,211 - ERROR - Error submitting iteration 276: A child process terminated abruptly, the process pool is not usable anymore [STDERR] 2025-09-03 01:39:30,212 - ERROR - Error processing result from iteration 250: A process in the process pool was terminated abruptly while the future was running or pending. [STDERR] 2025-09-03 01:39:30,221 - ERROR - Error submitting iteration 276: A child process terminated abruptly, the process pool is not usable anymore [STDERR] 2025-09-03 01:39:30,228 - ERROR - Error submitting iteration 276: A child process terminated abruptly, the process pool is not usable anymore [STDERR] 2025-09-03 01:39:30,229 - ERROR - Error processing result from iteration 254: A process in the process pool was terminated abruptly while the future was running or pending. [STDERR] 2025-09-03 01:39:30,235 - ERROR - Error submitting iteration 276: A child process terminated abruptly, the process pool is not usable anymore [STDERR] 2025-09-03 01:39:30,240 - ERROR - Error submitting iteration 276: A child process terminated abruptly, the process pool is not usable anymore

this is my config：

`# Configuration for function minimization example
max_iterations: 500
checkpoint_interval: 1
log_level: "INFO"

Evolution settings

diff_based_evolution: true # Use diff-based evolution (true) or full rewrites (false)

max_code_length: 30000 # Maximum allowed code length in characters

LLM configuration

llm:

Models for evolution

models:
# List of available models with their weights
- name: "deepseek-reasoner"
weight: 0.8
- name: "deepseek-reasoner"
weight: 0.2

Models for LLM feedback

evaluator_models:
# List of available models with their weights
- name: "deepseek-reasoner"
weight: 0.8
- name: "deepseek-reasoner"
weight: 0.2

api_base: "https://api.deepseek.com/v1"
api_key: null

temperature: 0.7
top_p: 0.95
max_tokens: 4096

Request parameters

timeout: 1200 # Timeout for API requests in seconds
retries: 5 # Number of retries for failed requests
retry_delay: 5 # Delay between retries in seconds

Prompt configuration

prompt:
system_message: |
One, Identity
You are a highly experienced programmer specializing in optimization algorithms, a top-tier Ph.D. in algorithms.
You are also a senior quantitative trading strategy expert proficient in all trading techniques throughout human history,
with exceptional analytical and innovative capabilities.
（...more）

num_top_programs: 3
use_template_stochasticity: true

Database configuration

database:
population_size: 70
archive_size: 20
num_islands: 12
elite_selection_ratio: 0.2
exploitation_ratio: 0.7

You can mix built-in features with custom metrics from your evaluator:

feature_dimensions: # Dimensions for MAP-Elites feature map (for diversity, NOT fitness)
- "complexity" # Code length (built-in)
- "diversity" # Code diversity (built-in)

Evaluator configuration

evaluator:

General settings

timeout: 600 # Maximum evaluation time in seconds
max_retries: 5 # Maximum number of retries for evaluation

Evaluation strategies

cascade_evaluation: false # Use cascade evaluation to filter bad solutions early
cascade_thresholds: # Thresholds for advancing to next evaluation stage
- 0.5 # First stage threshold
- 0.75 # Second stage threshold
- 0.9 # Third stage threshold
parallel_evaluations: 12
use_llm_feedback: false

Evolution settings

diff_based_evolution: false
allow_full_rewrites: false
`

[codelion]

Thanks for testing, I will work on a fix.