SmolLM-360M-Instruct_fsdp_qlora_nf4_adapter

README.md

@@ -20,7 +20,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [HuggingFaceTB/SmolLM-360M-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM-360M-Instruct) on the generator dataset.
 It achieves the following results on the evaluation set:
-- Loss: 2.1967
+- Loss: 1.8125
 
 ## Model description
 
@@ -51,13 +51,32 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.03
-- num_epochs: 1
+- num_epochs: 20
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss |
-|:-------------:|:------:|:----:|:---------------:|
-| 2.2811        | 0.9524 | 10   | 2.1967          |
+| Training Loss | Epoch   | Step | Validation Loss |
+|:-------------:|:-------:|:----:|:---------------:|
+| 2.3025        | 0.9524  | 10   | 2.1503          |
+| 2.1107        | 2.0     | 21   | 2.0371          |
+| 2.0147        | 2.9524  | 31   | 1.9716          |
+| 1.9577        | 4.0     | 42   | 1.9244          |
+| 1.9087        | 4.9524  | 52   | 1.8953          |
+| 1.8825        | 6.0     | 63   | 1.8716          |
+| 1.8667        | 6.9524  | 73   | 1.8558          |
+| 1.8488        | 8.0     | 84   | 1.8429          |
+| 1.8285        | 8.9524  | 94   | 1.8343          |
+| 1.8201        | 10.0    | 105  | 1.8270          |
+| 1.8129        | 10.9524 | 115  | 1.8219          |
+| 1.8028        | 12.0    | 126  | 1.8179          |
+| 1.7987        | 12.9524 | 136  | 1.8155          |
+| 1.7938        | 14.0    | 147  | 1.8137          |
+| 1.7899        | 14.9524 | 157  | 1.8129          |
+| 1.7903        | 16.0    | 168  | 1.8126          |
+| 1.7884        | 16.9524 | 178  | 1.8125          |
+| 1.7892        | 18.0    | 189  | 1.8124          |
+| 1.7825        | 18.9524 | 199  | 1.8125          |
+| 1.7906        | 19.0476 | 200  | 1.8125          |
 
 
 ### Framework versions

adapter_config.json

@@ -20,12 +20,12 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "up_proj",
+    "down_proj",
+    "q_proj",
     "k_proj",
+    "up_proj",
     "v_proj",
-    "down_proj",
     "gate_proj",
-    "q_proj",
     "o_proj"
   ],
   "task_type": "CAUSAL_LM",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d83b4ec8af7d70272df4b4e55b9a6299c66bc7dd90361443688b58cf9a7f1c47
+oid sha256:60eccafbdd7e27e07e3068c9edd230824847642c908f3e2a4bec9fb098dd47bf
 size 17426248

runs/Aug29_14-50-40_algo-2/events.out.tfevents.1724943055.algo-2.67.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ef8cfdcd5588ee3d488035f009ea9cb6db63b898577be1ac2ec83e2b4e7415e
+size 15293

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dd95ff3e5837b1ef37da1017c5cb06b5565c78c981567f7be3613d9f8789d47f
+oid sha256:d66baec8168be38945aef6f74e01084dbd1792505d90526651d84ce0814c4c51
 size 5240