Upload training_run.json with huggingface_hub
Browse files- training_run.json +5 -5
training_run.json
CHANGED
|
@@ -1,18 +1,18 @@
|
|
| 1 |
{
|
| 2 |
-
"timestamp": "2026-02-
|
| 3 |
"model": "google/medgemma-1.5-4b-it",
|
| 4 |
"adapter_repo": "s0close/medgemma-dsa-lora",
|
| 5 |
"dataset_repo": "s0close/medgemma-dsa-train",
|
| 6 |
"training_samples": 432,
|
| 7 |
-
"epochs":
|
| 8 |
-
"lr":
|
| 9 |
"lora_r": 8,
|
| 10 |
"lora_alpha": 16,
|
| 11 |
"batch_size": 1,
|
| 12 |
"grad_accum": 8,
|
| 13 |
"max_seq_length": 1024,
|
| 14 |
-
"training_loss": 0.
|
| 15 |
-
"training_time_seconds":
|
| 16 |
"compute_dtype": "bfloat16",
|
| 17 |
"quantization": "4-bit NF4 double-quant",
|
| 18 |
"environment": "hf_jobs"
|
|
|
|
| 1 |
{
|
| 2 |
+
"timestamp": "2026-02-15T11:29:57.546067",
|
| 3 |
"model": "google/medgemma-1.5-4b-it",
|
| 4 |
"adapter_repo": "s0close/medgemma-dsa-lora",
|
| 5 |
"dataset_repo": "s0close/medgemma-dsa-train",
|
| 6 |
"training_samples": 432,
|
| 7 |
+
"epochs": 2,
|
| 8 |
+
"lr": 2e-05,
|
| 9 |
"lora_r": 8,
|
| 10 |
"lora_alpha": 16,
|
| 11 |
"batch_size": 1,
|
| 12 |
"grad_accum": 8,
|
| 13 |
"max_seq_length": 1024,
|
| 14 |
+
"training_loss": 0.8743757631767679,
|
| 15 |
+
"training_time_seconds": 4394.945178031921,
|
| 16 |
"compute_dtype": "bfloat16",
|
| 17 |
"quantization": "4-bit NF4 double-quant",
|
| 18 |
"environment": "hf_jobs"
|