Commit 4259c4a0 authored by Brewer, Wes's avatar Brewer, Wes
Browse files

Fix a few bugs related to the addition of account management

parent 9395b8d3

main.py

+0 −1
Original line number Diff line number Diff line
@@ -104,7 +104,6 @@ if args.replay: 
    else: # custom data loader

        print(*args.replay)

        jobs = td.load_data(args.replay)

        for job in jobs: job['priority'] = sc.policy.aging_boost(job['nodes_required'])

        td.save_snapshot(jobs, filename=DIR_NAME)



    # Set number of timesteps based on the last job running which we assume

raps/account.py

+4 −1
Original line number Diff line number Diff line
@@ -32,6 +32,9 @@ class Account: 
        self.total_jobs += 1

        self.time_allocated += jobstats.run_time

        self.energy_allocated += jobstats.energy

        if self.time_allocated == 0:

            self.avg_power = 0

        else:

            self.avg_power = self.energy_allocated / self.time_allocated

        if average_user.avg_power == 0:  # If this is the first job use own power

            average_user.avg_power = self.avg_power

raps/dataloaders/adastraMI250.py

+2 −1
Original line number Diff line number Diff line
@@ -82,6 +82,7 @@ def load_data_from_df(jobs_df: pd.DataFrame, **kwargs): 
    # Map dataframe to job state. Add results to jobs list

    for jidx in tqdm(range(num_jobs - 1), total=num_jobs, desc="Processing Jobs"):



        account = jobs_df.loc[jidx, 'user_id'] # or 'group_id'

        job_id = jobs_df.loc[jidx, 'job_id']



        if not jid == '*':
@@ -163,7 +164,7 @@ def load_data_from_df(jobs_df: pd.DataFrame, **kwargs): 


        if time_offset >= 0  and wall_time >  0:

            #print("start_time",time_start,"\tend_time",time_end,"\twall_time",wall_time,"\tquanta wall time",gpu_trace.size * TRACE_QUANTA )

            job_info = job_dict(nodes_required, name, cpu_trace, gpu_trace, [],[],wall_time,

            job_info = job_dict(nodes_required, name, account, cpu_trace, gpu_trace, [],[],wall_time,

                                end_state, scheduled_nodes, time_offset, job_id, priority)

            jobs.append(job_info)

        else:

raps/dataloaders/frontier.py

+3 −1
Original line number Diff line number Diff line
@@ -95,6 +95,8 @@ def load_data_from_df(jobs_df: pd.DataFrame, jobprofile_df: pd.DataFrame, **kwar 
    # Map dataframe to job state. Add results to jobs list

    for jidx in tqdm(range(num_jobs - 1), total=num_jobs, desc="Processing Jobs"):



        user = jobs_df.loc[jidx, 'user']

        account = jobs_df.loc[jidx, 'account']

        job_id = jobs_df.loc[jidx, 'job_id']

        allocation_id = jobs_df.loc[jidx, 'allocation_id']

        nodes_required = jobs_df.loc[jidx, 'node_count']
@@ -154,7 +156,7 @@ def load_data_from_df(jobs_df: pd.DataFrame, jobprofile_df: pd.DataFrame, **kwar 
                scheduled_nodes.append(indices)



        if gpu_trace.size > 0 and (jid == job_id or jid == '*') and time_offset > 0:

            job_info = job_dict(nodes_required, name, cpu_trace, gpu_trace, [], [], wall_time, 

            job_info = job_dict(nodes_required, name, account, cpu_trace, gpu_trace, [], [], wall_time, 

                                end_state, scheduled_nodes, time_offset, job_id, priority)

            jobs.append(job_info)

raps/dataloaders/fugaku.py

+2 −0
Original line number Diff line number Diff line
@@ -68,6 +68,7 @@ def load_data_from_df(df, **kwargs): 
    # Loop through the DataFrame rows to extract job information

    for _, row in tqdm(df.iterrows(), total=len(df), desc="Processing Jobs"):

        nodes_required = row['nnumr'] if 'nnumr' in df.columns else 0

        account = row['usr']

        name = row['jnam'] if 'jnam' in df.columns else 'unknown'



        if validate:
@@ -95,6 +96,7 @@ def load_data_from_df(df, **kwargs): 
        job_info = job_dict(

            nodes_required=nodes_required,

            name=name,

            account=account,

            cpu_trace=cpu_trace,

            gpu_trace=gpu_trace,

            ntx_trace=[],