Upload folder using huggingface_hub

0000120_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12ad72c3fcb9e363947f657aae9d73e4dfbad7c0af79e80497bbdffbe1985f15
+size 10563230

adapter_config.json

@@ -0,0 +1,41 @@
+{
+    "adapter_path": "/Users/onyx/crasher/slm-training/adapters/deepseek-r1-1.5b/seed42",
+    "batch_size": 4,
+    "clear_cache_threshold": 0,
+    "config": "/Users/onyx/crasher/slm-training/adapters/deepseek-r1-1.5b/seed42/lora_config.yaml",
+    "data": "/Users/onyx/crasher/slm-training/data",
+    "fine_tune_type": "lora",
+    "grad_accumulation_steps": 1,
+    "grad_checkpoint": false,
+    "iters": 120,
+    "learning_rate": 0.0001,
+    "lora_parameters": {
+        "rank": 8,
+        "scale": 20.0,
+        "dropout": 0.0
+    },
+    "lr_schedule": null,
+    "mask_prompt": false,
+    "max_seq_length": 1024,
+    "model": "mlx-community/DeepSeek-R1-Distill-Qwen-1.5B-4bit",
+    "num_layers": 8,
+    "optimizer": "adam",
+    "optimizer_config": {
+        "adam": {},
+        "adamw": {},
+        "muon": {},
+        "sgd": {},
+        "adafactor": {}
+    },
+    "project_name": null,
+    "report_to": null,
+    "resume_adapter_file": null,
+    "save_every": 120,
+    "seed": 42,
+    "steps_per_eval": 40,
+    "steps_per_report": 20,
+    "test": false,
+    "test_batches": 500,
+    "train": true,
+    "val_batches": 8
+}

adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12ad72c3fcb9e363947f657aae9d73e4dfbad7c0af79e80497bbdffbe1985f15
+size 10563230

lora_config.yaml

@@ -0,0 +1,20 @@
+model: mlx-community/DeepSeek-R1-Distill-Qwen-1.5B-4bit
+train: true
+data: /Users/onyx/crasher/slm-training/data
+seed: 42
+fine_tune_type: lora
+num_layers: 8
+batch_size: 4
+iters: 120
+learning_rate: 0.0001
+max_seq_length: 1024
+steps_per_report: 20
+steps_per_eval: 40
+val_batches: 8
+save_every: 120
+grad_checkpoint: false
+adapter_path: /Users/onyx/crasher/slm-training/adapters/deepseek-r1-1.5b/seed42
+lora_parameters:
+  rank: 8
+  scale: 20.0
+  dropout: 0.0