Upload rank_300/config.json with huggingface_hub
Browse files- rank_300/config.json +22 -0
rank_300/config.json
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"experiment": "r300_qwen3teacher_hardnegs",
|
| 3 |
+
"rank": 300,
|
| 4 |
+
"d_model": 300,
|
| 5 |
+
"vocab_size": 82741,
|
| 6 |
+
"seq_len": 512,
|
| 7 |
+
"batch_size": 256,
|
| 8 |
+
"total_steps": 150000,
|
| 9 |
+
"warmup": 1000,
|
| 10 |
+
"lr": 0.0005,
|
| 11 |
+
"min_lr": 1e-05,
|
| 12 |
+
"weight_decay": 0.01,
|
| 13 |
+
"grad_clip": 1.0,
|
| 14 |
+
"tau": 0.05,
|
| 15 |
+
"lambda_mse": 1.0,
|
| 16 |
+
"H_hard_negatives_per_anchor": 64,
|
| 17 |
+
"hard_neg_repo": "jsanzolac/qwen3_emb_512_hard_negatives",
|
| 18 |
+
"warm_start_repo": "jsanzolac/bpe_glove_300_lora_r300_qwen3",
|
| 19 |
+
"warm_start_path": "rank_300/checkpoint_final.pt",
|
| 20 |
+
"glove_repo": "jsanzolac/drifting-glove-distilled-r300",
|
| 21 |
+
"created_at": "2026-05-21T14:16:30.508447"
|
| 22 |
+
}
|