Mojiz-DPO

README.md

@@ -15,6 +15,16 @@ should probably proofread and complete it, then remove this comment. -->
 # test
 
 This model is a fine-tuned version of [ahmedabdelwahed/Mojiz-sft](https://huggingface.co/ahmedabdelwahed/Mojiz-sft) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.1612
+- Rewards/chosen: 1.5401
+- Rewards/rejected: -0.4134
+- Rewards/accuracies: 1.0
+- Rewards/margins: 1.9534
+- Logps/rejected: -71.8662
+- Logps/chosen: -323.2329
+- Logits/rejected: -11.6886
+- Logits/chosen: -12.9104
 
 ## Model description
 
@@ -42,6 +52,22 @@ The following hyperparameters were used during training:
 - lr_scheduler_warmup_steps: 150
 - training_steps: 1000
 
+### Training results
+
+| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
+|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
+| 1.3422        | 0.41  | 100  | 0.6447          | 0.0816         | -0.0178          | 1.0                | 0.0994          | -71.0751       | -326.1499    | -11.7285        | -12.9749      |
+| 1.9054        | 0.82  | 200  | 0.5151          | 0.3250         | -0.0755          | 1.0                | 0.4005          | -71.1904       | -325.6630    | -11.7221        | -12.9647      |
+| 1.7913        | 1.22  | 300  | 0.4000          | 0.5897         | -0.1444          | 1.0                | 0.7341          | -71.3283       | -325.1337    | -11.7152        | -12.9536      |
+| 1.6915        | 1.63  | 400  | 0.3189          | 0.8238         | -0.2053          | 1.0                | 1.0290          | -71.4500       | -324.6655    | -11.7087        | -12.9431      |
+| 1.4041        | 2.04  | 500  | 0.2596          | 1.0374         | -0.2638          | 1.0                | 1.3012          | -71.5670       | -324.2383    | -11.7029        | -12.9335      |
+| 1.1989        | 2.45  | 600  | 0.2187          | 1.2160         | -0.3160          | 1.0                | 1.5320          | -71.6715       | -323.8811    | -11.6980        | -12.9256      |
+| 1.2192        | 2.86  | 700  | 0.1915          | 1.3556         | -0.3570          | 1.0                | 1.7126          | -71.7533       | -323.6018    | -11.6939        | -12.9188      |
+| 1.1718        | 3.27  | 800  | 0.1738          | 1.4588         | -0.3884          | 1.0                | 1.8472          | -71.8162       | -323.3954    | -11.6909        | -12.9141      |
+| 0.9623        | 3.67  | 900  | 0.1643          | 1.5190         | -0.4068          | 1.0                | 1.9258          | -71.8531       | -323.2751    | -11.6892        | -12.9113      |
+| 1.0207        | 4.08  | 1000 | 0.1612          | 1.5401         | -0.4134          | 1.0                | 1.9534          | -71.8662       | -323.2329    | -11.6886        | -12.9104      |
+
+
 ### Framework versions
 
 - PEFT 0.7.1

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:93cc34c48cdf032f5dbd222c824637c5b4937bdae2638a8e4b373e1d6956c927
+oid sha256:f074e5482ab7b55736ea422a7cae6b113c35fed4be27bd861a63e0cd28530d77
 size 7098016

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3c9b5296bb1bdbcc33ccb40be1c81db3a36335a1c2b080a5368b48fb7a9eaa66
-size 16330461
+oid sha256:c6159d7ec2f2ff06356659536274e669822a98188a9cf7f009f92a264ab5abf4
+size 16330559

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:77af763fa3f9ba2e354c4f6742d9864e3e73e7734d8c970627d6587442d1eea0
+oid sha256:4b39ddfbc580f2428e71774c75f1bf5d819d38ab012152bd6f322042e89e5044
 size 4219