Training in progress, step 2000

adapter_config.json

@@ -3,7 +3,7 @@
   "alpha_pattern": {},
   "arrow_config": null,
   "auto_mapping": null,
-  "base_model_name_or_path": "mazesmazes/tiny-audio",
+  "base_model_name_or_path": "Qwen/Qwen3-0.6B",
   "bias": "none",
   "corda_config": null,
   "ensure_weight_tying": false,
@@ -16,9 +16,9 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 256,
+  "lora_alpha": 32,
   "lora_bias": false,
-  "lora_dropout": 0.05,
+  "lora_dropout": 0.0,
   "megatron_config": null,
   "megatron_core": "megatron.core",
   "modules_to_save": null,
@@ -29,8 +29,8 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "q_proj",
-    "v_proj"
+    "v_proj",
+    "q_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b42236aaf96efc41e22f9dfed16985e994aebfe5dbd7d005da5c25598ac7a4cf
+oid sha256:f8492345b1762cee01629ee589f91551b15ed7050fb203c8f919fc89131a4d75
 size 36715216

config.json

@@ -277,13 +277,13 @@
   ],
   "encoder_dim": 1280,
   "freeze_projector": false,
-  "freq_mask_length": 10,
+  "freq_mask_length": 43,
   "inference_warmup_tokens": 10,
   "label_smoothing": 0.0,
   "length_penalty": 1.0,
   "llm_dim": 1024,
-  "lora_alpha": 256,
-  "lora_dropout": 0.05,
+  "lora_alpha": 32,
+  "lora_dropout": 0.0,
   "lora_rank": 64,
   "lora_target_modules": [
     "q_proj",
@@ -300,7 +300,6 @@
   "num_freq_masks": 2,
   "num_time_masks": 2,
   "pipeline_tag": "automatic-speech-recognition",
-  "pretrained_model_path": "mazesmazes/tiny-audio",
   "projector_dropout": 0.0,
   "projector_hidden_dim": null,
   "projector_init_std": 0.02,
@@ -378,7 +377,7 @@
     "vocab_size": 151670
   },
   "text_model_id": "Qwen/Qwen3-0.6B",
-  "time_mask_length": 15,
+  "time_mask_length": 100,
   "transformers_version": "5.0.0.dev0",
   "use_cache": false,
   "use_lora": true,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:451ac117081e6354a030959d0f922af2259e18bd87c6eba299188fd4bda93752
+oid sha256:0347348d2f9cd64839ab1875ed302a1e1c18f65a070a3643f044869e50d2dce4
 size 25172384

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a42bd1aaa0bffdaabd32fd8a4af9544e0ed2608a2116ad462a6945ee4b9eefff
+oid sha256:1540bbe90c48101bd1f6fb63273e8b75f53d50d743e6bdb8f473bd0c470b4aba
 size 5201