Training in progress, step 10000

config.json

@@ -281,7 +281,7 @@
   "inference_warmup_tokens": 10,
   "label_smoothing": 0.0,
   "length_penalty": 1.0,
-  "llm_dim": 1024,
+  "llm_dim": 2048,
   "lora_alpha": 32,
   "lora_dropout": 0.0,
   "lora_rank": 8,
@@ -321,7 +321,7 @@
   "router_aux_loss_coef": 0.01,
   "system_prompt": "",
   "text_config": {
-    "_name_or_path": "Qwen/Qwen3-0.6B",
+    "_name_or_path": "Qwen/Qwen3-1.7B",
     "architectures": [
       "Qwen3ForCausalLM"
     ],
@@ -331,9 +331,9 @@
     "eos_token_id": 151645,
     "head_dim": 128,
     "hidden_act": "silu",
-    "hidden_size": 1024,
+    "hidden_size": 2048,
     "initializer_range": 0.02,
-    "intermediate_size": 3072,
+    "intermediate_size": 6144,
     "layer_types": [
       "full_attention",
       "full_attention",
@@ -382,7 +382,7 @@
     "use_sliding_window": false,
     "vocab_size": 151670
   },
-  "text_model_id": "Qwen/Qwen3-0.6B",
+  "text_model_id": "Qwen/Qwen3-1.7B",
   "time_mask_length": 100,
   "transformers_version": "5.0.0.dev0",
   "use_cache": false,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d53c23dd303affd6818a84362bfde2a1d5a5df17dc8df88610bf12647f8fc252
-size 12587424
+oid sha256:768d4130a1c141e8c626ec9815cc5de075420e2970c37d7ce52763a18ca205cb
+size 58732960

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:18572cd996e881f07a0575e01bb0e0151fd4feb2cf38350b9c02d1ee7f580ec9
+oid sha256:6a7b7022481c58a22f6d25ca3fc9a90f2337145b17d68d21726cddba4714e179
 size 5265