bart-finetuned-loaf-512-lyrictoplan

README.md

@@ -15,7 +15,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [facebook/bart-large](https://huggingface.co/facebook/bart-large) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.4458
+- Loss: 0.3928
 
 ## Model description
 
@@ -47,12 +47,12 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | Validation Loss |
 |:-------------:|:-----:|:----:|:---------------:|
-| 1.0218        | 0.45  | 500  | 0.5545          |
-| 0.5582        | 0.89  | 1000 | 0.5083          |
-| 0.4996        | 1.34  | 1500 | 0.4844          |
-| 0.4744        | 1.78  | 2000 | 0.4609          |
-| 0.449         | 2.23  | 2500 | 0.4560          |
-| 0.4287        | 2.67  | 3000 | 0.4458          |
+| 0.7003        | 0.45  | 500  | 0.5146          |
+| 0.5357        | 0.89  | 1000 | 0.4827          |
+| 0.4799        | 1.34  | 1500 | 0.4387          |
+| 0.4354        | 1.78  | 2000 | 0.4132          |
+| 0.4047        | 2.23  | 2500 | 0.4049          |
+| 0.3739        | 2.67  | 3000 | 0.3928          |
 
 
 ### Framework versions

generation_config.json

@@ -8,14 +8,14 @@
   "forced_bos_token_id": 0,
   "forced_eos_token_id": 2,
   "max_new_tokens": 512,
-  "min_new_tokens": 128,
+  "min_new_tokens": 432,
   "n_examples": null,
   "no_repeat_ngram_size": 3,
   "num_beams": 4,
   "pad_token_id": 1,
   "padding": "max_length",
   "renormalize_logits": true,
-  "repetition_penalty": 0.6,
+  "repetition_penalty": 0.4,
   "skip_special_tokens": true,
   "temperature": 0.85,
   "top_k": 0,

runs/Apr09_07-06-43_870b8dce1033/events.out.tfevents.1712646406.870b8dce1033.46497.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6c638f8c384a212ae5db7056fa2811fd9e3beb2e024556d8944eff0075bddc8c
-size 9033
+oid sha256:ed393bd04a082b5e8d71507b63b5d0b08f491b18c02e37a4d7dfdba25dff88ee
+size 9387