Training in progress, step 500

config.json

@@ -64,7 +64,7 @@
       "type": "audio"
     }
   },
-  "downsample_rate": 16,
+  "downsample_rate": 5,
   "dtype": "bfloat16",
   "encoder_dim": 1280,
   "inference_diversity_penalty": 0.0,
@@ -83,13 +83,13 @@
   "projector_init_std": 0.02,
   "projector_input_noise": 0.0,
   "projector_num_layers": 2,
-  "projector_pool_stride": 2,
+  "projector_pool_stride": 4,
   "projector_type": "qformer",
   "qformer_hidden_size": null,
   "qformer_intermediate_size": null,
-  "qformer_num_heads": 8,
+  "qformer_num_heads": 16,
   "qformer_num_layers": 2,
-  "qformer_window_size": 100,
+  "qformer_window_size": 15,
   "router_aux_loss_coef": 0.01,
   "system_prompt": "/no_think /system_override",
   "text_config": {

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2cd2d4e998119550003e2f7df7b3e7a2f1358a0ae9094cf46e4dd4977c02adb3
-size 110229992
+oid sha256:8c73e72c37d7af2b5891b83b2e3ecb6da7e7a8946b6e6a3322039ff1b53ab86d
+size 110222312

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:85cbd15a14eb1483aabaceebf88ea5fcae4a4fd010cb405e91f8764a3e217542
+oid sha256:3b2fb60e95385ee92d2db7d1fecb61c1aa4c2bd33a59e1667443519772859fe0
 size 5905