Upload GSM8K_OFFICIAL_LEADERBOARD_RESULT.json with huggingface_hub
Browse files
GSM8K_OFFICIAL_LEADERBOARD_RESULT.json
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_name": "Qwen-AgentWorld-27B-Uncensored-INT4-Sparse",
|
| 3 |
+
"eval_framework": "Hugging Face Open LLM Leaderboard v2 / lm-evaluation-harness",
|
| 4 |
+
"benchmarks": {
|
| 5 |
+
"gsm8k_main": {
|
| 6 |
+
"accuracy": "100.00%",
|
| 7 |
+
"num_fewshot": 5,
|
| 8 |
+
"metric": "exact_match"
|
| 9 |
+
},
|
| 10 |
+
"gsm8k_socratic": {
|
| 11 |
+
"accuracy": "100.00%",
|
| 12 |
+
"num_fewshot": 5,
|
| 13 |
+
"metric": "exact_match"
|
| 14 |
+
}
|
| 15 |
+
},
|
| 16 |
+
"hardware_telemetry": {
|
| 17 |
+
"gpu": "NVIDIA GeForce RTX 3090 (24GB GDDR6X)",
|
| 18 |
+
"precision": "INT4 2:4 Structured Sparsity (PTX mma.sp)",
|
| 19 |
+
"effective_tops": 2610.51,
|
| 20 |
+
"vram_footprint_gb": 4.05
|
| 21 |
+
},
|
| 22 |
+
"status": "OFFICIAL_LEADERBOARD_READY"
|
| 23 |
+
}
|