Update app.py

app.py

@@ -7,30 +7,9 @@ TOKEN = os.environ.get("BULK_ENERGY_TOKEN")
 API = HfApi(token=TOKEN)
 
 REPO_ID = "AIEnergyScore/BulkCalcSpace"
+RESULTS_DSET = "AIEnergyScore/BulkCalcResults"
 app = FastAPI()
 
-def check_for_traceback(run_dir):
-    # run_dir="./runs/${experiment_name}/${backend_model}/${now}"
-    found_error = False
-    error_message = ""
-    try:
-        # Read error message
-        with open(f"{run_dir}/error.log", 'r') as f:
-            # There may be a better way to do this that finds the
-            # index of Traceback, then prints from there : end-of-file index (the file length-1).
-            for line in f:
-                # Question: Do we even need to check for this? The presence of the
-                # error file, or at least a non-empty one,
-                # means there's been an error, no?
-                if 'Traceback (most recent call last):' in line:
-                    found_error = True
-                if found_error:
-                    error_message += line
-    except FileNotFoundError as e:
-        # When does this happen?
-        print(f"Could not find {run_dir}/error.log")
-    return error_message
-    
 @app.get("/")
 def start_train():
     model_file = open("models.txt", "r+").readlines()
@@ -38,7 +17,7 @@ def start_train():
     hardware_file = open("hardware.txt", "r+").readlines()
     for hardware in hardware_file:
         hardware = hardware.strip()
-        print(f"Hardware is {hardware}")
+        print(f"Requested hardware is {hardware}")
         curr_runtime = API.get_space_runtime(repo_id=REPO_ID)
         print(f"Current hardware is {curr_runtime}")
         if curr_runtime != hardware:
@@ -46,18 +25,22 @@ def start_train():
             API.request_space_hardware(repo_id=REPO_ID, hardware=hardware)
         for model in model_file:
             model = model.strip()
+            print(f"Attempting to benchmark model {model}.")
             for task in task_file:
                 task = task.strip()
+                print(f"Attempting to benchmark model {model} on task {task}.")
                 # Create the name of the directory for output.
                 now = time.time()
-                run_dir = f"/runs/${task}/${model}/${now}"
+                run_dir = f"/runs/{task}/{model}/{now}"
                 os.system(f"./entrypoint.sh {REPO_ID} {model} {task} {hardware} {run_dir}")
-    #space_variables = API.get_space_variables(repo_id=REPO_ID)
-    #if 'STATUS' not in space_variables or space_variables['STATUS'] != 'COMPUTING':
-    #    print("Beginning processing.")
-    #    API.add_space_variable(repo_id=REPO_ID, key='STATUS', value='COMPUTING')
-    #os.system(f"./entrypoint.sh {REPO_ID} {model} {task} {hardware}")
-    #API.add_space_variable(repo_id=REPO_ID, key='STATUS', value='NOT_COMPUTING')
-    #print("Pausing space")
-    #API.pause_space(REPO_ID)
-    return {"Status": "Done"}#space_variables['STATUS']}
+                # Uploads all run output to the results dataset.
+                print(f"Uploading {run_dir} to {RESULTS_DSET}")
+                try:
+                    API.create_repo(repo_id=f"{RESULTS_DSET}", repo_type="dataset",)
+                    print(f"Created results dataset repository")
+                except:
+                    print(f"Using pre-existing dataset respository.")
+                API.upload_folder(folder_path=run_dir, repo_id=f"{RESULTS_DSET}", repo_type="dataset",)
+    print("Pausing space")
+    API.pause_space(REPO_ID)
+    #return {"Status": "Done"}