Model save

README.md

@@ -2,8 +2,6 @@
 license: cc-by-nc-4.0
 library_name: peft
 tags:
-- text-to-audio
-- ylacombe/tiny-punk
 - generated_from_trainer
 base_model: facebook/musicgen-melody
 model-index:
@@ -16,10 +14,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # musicgen-melody-punk-lora
 
-This model is a fine-tuned version of [facebook/musicgen-melody](https://huggingface.co/facebook/musicgen-melody) on the YLACOMBE/TINY-PUNK - DEFAULT dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.2634
-- Clap: nan
+This model is a fine-tuned version of [facebook/musicgen-melody](https://huggingface.co/facebook/musicgen-melody) on an unknown dataset.
 
 ## Model description
 

adapter_config.json

@@ -23,22 +23,22 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "enc_to_dec_proj",
-    "out_proj",
-    "fc2",
-    "lm_heads.3",
-    "lm_heads.2",
+    "q_proj",
     "embed_tokens.0",
     "lm_heads.1",
-    "q_proj",
+    "lm_heads.3",
+    "lm_heads.2",
+    "audio_enc_to_dec_proj",
+    "v_proj",
     "k_proj",
-    "fc1",
-    "embed_tokens.2",
+    "out_proj",
     "embed_tokens.3",
-    "v_proj",
     "lm_heads.0",
-    "audio_enc_to_dec_proj",
-    "embed_tokens.1"
+    "fc2",
+    "embed_tokens.2",
+    "fc1",
+    "embed_tokens.1",
+    "enc_to_dec_proj"
   ],
   "task_type": null,
   "use_dora": false,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:78b66138da22e547a9045f9b37fbf37b27e087670b0dc0350582fa254d5ee26b
+oid sha256:65ed04caf4a8554877af5cd683c5a4014ea967aa9617be556c959836879781c8
 size 87103456

all_results.json

@@ -1,15 +1,15 @@
 {
     "epoch": 3.764705882352941,
-    "eval_clap": NaN,
-    "eval_loss": 0.2634366452693939,
-    "eval_runtime": 177.4407,
+    "eval_clap": 0.3319118022918701,
+    "eval_loss": 4.704505920410156,
+    "eval_runtime": 149.2288,
     "eval_samples": 5,
-    "eval_samples_per_second": 0.028,
-    "eval_steps_per_second": 0.028,
-    "total_flos": 35328595626816.0,
-    "train_loss": 6.771849155426025,
-    "train_runtime": 276.9653,
+    "eval_samples_per_second": 0.034,
+    "eval_steps_per_second": 0.034,
+    "total_flos": 59875624258368.0,
+    "train_loss": 8.819560050964355,
+    "train_runtime": 362.7337,
     "train_samples": 33,
-    "train_samples_per_second": 0.477,
-    "train_steps_per_second": 0.029
+    "train_samples_per_second": 0.364,
+    "train_steps_per_second": 0.022
 }

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "facebook/musicgen-melody-large",
+  "_name_or_path": "facebook/musicgen-melody",
   "architectures": [
     "MusicgenMelodyForConditionalGeneration"
   ],
@@ -120,11 +120,11 @@
     "encoder_no_repeat_ngram_size": 0,
     "eos_token_id": null,
     "exponential_decay_length_penalty": null,
-    "ffn_dim": 8192,
+    "ffn_dim": 6144,
     "finetuning_task": null,
     "forced_bos_token_id": null,
     "forced_eos_token_id": null,
-    "hidden_size": 2048,
+    "hidden_size": 1536,
     "id2label": {
       "0": "LABEL_0",
       "1": "LABEL_1"
@@ -143,7 +143,7 @@
     "min_length": 0,
     "model_type": "musicgen_melody_decoder",
     "no_repeat_ngram_size": 0,
-    "num_attention_heads": 32,
+    "num_attention_heads": 24,
     "num_beam_groups": 1,
     "num_beams": 1,
     "num_codebooks": 4,

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 3.764705882352941,
-    "eval_clap": NaN,
-    "eval_loss": 0.2634366452693939,
-    "eval_runtime": 177.4407,
+    "eval_clap": 0.3319118022918701,
+    "eval_loss": 4.704505920410156,
+    "eval_runtime": 149.2288,
     "eval_samples": 5,
-    "eval_samples_per_second": 0.028,
-    "eval_steps_per_second": 0.028
+    "eval_samples_per_second": 0.034,
+    "eval_steps_per_second": 0.034
 }

model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a6610b21d2b0da10c212aa63ddaa43b9d9578c2606a5fcc6f10e818f4b39be3e
+oid sha256:8a3bef520a63852637a9d716027d50208a0524f8503e3f6ce755d246d13eb1c7
 size 4966269512

model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bdf7b00866034c7908b8bcff3b69ddeba3743235afb667a529411a6370a161d0
+oid sha256:fa44115d209c4f4ae102700ce5f17cc229f56c57a69c36ff035cf32c25d3ff18
 size 1263380400

train_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 3.764705882352941,
-    "total_flos": 35328595626816.0,
-    "train_loss": 6.771849155426025,
-    "train_runtime": 276.9653,
+    "total_flos": 59875624258368.0,
+    "train_loss": 8.819560050964355,
+    "train_runtime": 362.7337,
     "train_samples": 33,
-    "train_samples_per_second": 0.477,
-    "train_steps_per_second": 0.029
+    "train_samples_per_second": 0.364,
+    "train_steps_per_second": 0.022
 }

trainer_state.json

@@ -11,11 +11,11 @@
     {
       "epoch": 3.764705882352941,
       "step": 8,
-      "total_flos": 35328595626816.0,
-      "train_loss": 6.771849155426025,
-      "train_runtime": 276.9653,
-      "train_samples_per_second": 0.477,
-      "train_steps_per_second": 0.029
+      "total_flos": 59875624258368.0,
+      "train_loss": 8.819560050964355,
+      "train_runtime": 362.7337,
+      "train_samples_per_second": 0.364,
+      "train_steps_per_second": 0.022
     }
   ],
   "logging_steps": 25,
@@ -23,7 +23,7 @@
   "num_input_tokens_seen": 0,
   "num_train_epochs": 4,
   "save_steps": 500,
-  "total_flos": 35328595626816.0,
+  "total_flos": 59875624258368.0,
   "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f1ccac7c14909717b76ae18f6c6e89f17008eafe9462a646abafd7de11fc3992
+oid sha256:451ddedc8999e13d13a4ae6be9bc3dd43f434d3e71772902d308707b411dfdc3
 size 5112