change Repo names

Dockerfile

@@ -1,5 +1,5 @@
 # Use specific python image
-FROM registry.hf.space/pingandpasquale-moe-llm-gpu-poor-leaderboard:latest
+FROM registry.hf.space/sparse-generative-ai-open-moe-llm-leaderboard:latest
 
 RUN pip install -i https://test.pypi.org/simple/ --extra-index-url https://pypi.org/simple/ moe-infinity --no-cache-dir
 # To fix pydantic version

cli/create_request_file.py

@@ -9,7 +9,7 @@ from colorama import Fore
 from huggingface_hub import HfApi, snapshot_download
 
 EVAL_REQUESTS_PATH = "eval-queue"
-QUEUE_REPO = "PingAndPasquale/requests"
+QUEUE_REPO = "sparse-generative-ai/requests"
 
 precisions = ("float16", "bfloat16", "8bit (LLM.int8)", "4bit (QLoRA / FP4)", "GPTQ")
 model_types = ("pretrained", "fine-tuned", "RL-tuned", "instruction-tuned")

src/backend/run_eval_suite.py

@@ -12,8 +12,7 @@ from src.backend.tasks.cnndm.task_v2 import CNNDMv2
 from src.backend.tasks.selfcheckgpt.task import SelfCheckGPT
 
 from src.backend.huggingface_generate_until import HFLMwithChatTemplate
-from src.backend.moe_infinity import MoEHFLM
-
+from src.backend.moe_infinity import MoEHFLM  # MoEInfinity
 
 def run_evaluation(
     eval_request: EvalRequest,

src/envs.py

@@ -5,15 +5,15 @@ from huggingface_hub import HfApi
 # clone / pull the lmeh eval data
 H4_TOKEN = os.environ.get("H4_TOKEN", None)
 
-# REPO_ID = "pminervini/PingAndPasquale"
-REPO_ID = "PingAndPasquale/MOE-LLM-GPU-Poor-Leaderboard"
+# REPO_ID = "pminervini/sparse-generative-ai"
+REPO_ID = "sparse-generative-ai/open-moe-llm-leaderboard"
 
-QUEUE_REPO = "PingAndPasquale/requests"
+QUEUE_REPO = "sparse-generative-ai/requests"
 QUEUE_REPO_OPEN_LLM = "open-llm-leaderboard/requests"
-RESULTS_REPO = "PingAndPasquale/results"
+RESULTS_REPO = "sparse-generative-ai/results"
 
-PRIVATE_QUEUE_REPO = "PingAndPasquale/private-requests"
-PRIVATE_RESULTS_REPO = "PingAndPasquale/private-results"
+PRIVATE_QUEUE_REPO = "sparse-generative-ai/private-requests"
+PRIVATE_RESULTS_REPO = "sparse-generative-ai/private-results"
 
 IS_PUBLIC = bool(os.environ.get("IS_PUBLIC", True))
 
@@ -26,7 +26,7 @@ EVAL_REQUESTS_PATH_OPEN_LLM = os.path.join(CACHE_PATH, "eval-queue-open-llm")
 EVAL_REQUESTS_PATH_PRIVATE = "eval-queue-private"
 EVAL_RESULTS_PATH_PRIVATE = "eval-results-private"
 
-PATH_TO_COLLECTION = "PingAndPasquale/llm-leaderboard-best-models-652d6c7965a4619fb5c27a03"
+PATH_TO_COLLECTION = "sparse-generative-ai/llm-leaderboard-best-models-652d6c7965a4619fb5c27a03"
 
 # Rate limit variables
 RATE_LIMIT_PERIOD = 7

src/submission/submit.py

@@ -113,13 +113,14 @@ def add_new_eval(
     }
 
     # Check for duplicate submission
-    if f"{model}_{revision}_{precision}" in REQUESTED_MODELS:
+    if f"{model}_{revision}_{precision}_{inference_framework}" in REQUESTED_MODELS:
         return styled_warning("This model has been already submitted.")
 
     print("Creating eval file")
     OUT_DIR = f"{EVAL_REQUESTS_PATH}/{user_name}"
     os.makedirs(OUT_DIR, exist_ok=True)
-    out_path = f"{OUT_DIR}/{model_path}_eval_request_{private}_{precision}_{weight_type}.json"
+    # out_path = f"{OUT_DIR}/{model_path}_eval_request_{private}_{precision}_{weight_type}.json"
+    out_path = f"{OUT_DIR}/{model_path}_eval_request_{private}_{precision}_{weight_type}_{inference_framework}.json"
 
     with open(out_path, "w") as f:
         f.write(json.dumps(eval_entry))