Training in progress, step 1000

config.json

@@ -156,10 +156,18 @@
     ]
   ],
   "encoder_dim": 1280,
+  "freeze_projector": false,
   "inference_warmup_tokens": 10,
   "label_smoothing": 0.0,
   "length_penalty": 1.0,
   "llm_dim": 1024,
+  "lora_alpha": 32,
+  "lora_dropout": 0.0,
+  "lora_rank": 8,
+  "lora_target_modules": [
+    "q_proj",
+    "v_proj"
+  ],
   "mask_feature_length": 27,
   "mask_feature_min_masks": 2,
   "mask_feature_prob": 0.0,
@@ -188,7 +196,7 @@
   "qformer_window_size": 15,
   "repetition_penalty": 1.0,
   "router_aux_loss_coef": 0.01,
-  "system_prompt": "/no_think /system_override",
+  "system_prompt": "You are a helpful speech transcription assistant.",
   "text_config": {
     "_name_or_path": "Qwen/Qwen3-0.6B",
     "architectures": [
@@ -254,6 +262,7 @@
   "text_model_id": "Qwen/Qwen3-0.6B",
   "transformers_version": "5.0.0.dev0",
   "use_cache": false,
+  "use_lora": false,
   "use_specaugment": false,
   "user_prompt": "Please transcribe this English audio into text: <audio>",
   "vocab_size": 151670

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bc90ebd7f1e678a85a0a5e78f37c8a70357d296b72ad3a9a457d9359cc2ae7d4
+oid sha256:c47553e6d0e49b19f336273387688dbc869149abb13baf4bae6a72ebb19cae43
 size 25172384