Upload folder using huggingface_hub

args.json

@@ -41,10 +41,8 @@
   "dataset": [
     "/data/group_data/li_lab/siqiouya/datasets/gigaspeech/manifests/train_s_zh_v4_ner_baseline_aligned_rate1.0_k20_final.jsonl"
   ],
-  "val_dataset": [
-    "/data/group_data/li_lab/siqiouya/datasets/gigaspeech/manifests/dev_zh_v4_ner_baseline_aligned_freq_k20_final.jsonl"
-  ],
-  "split_dataset_ratio": 0.0,
+  "val_dataset": [],
+  "split_dataset_ratio": 0.01,
   "data_seed": 42,
   "dataset_num_proc": 8,
   "load_from_cache_file": true,
@@ -166,7 +164,7 @@
   "deterministic_mode": false,
   "train_iters": null,
   "log_interval": 10,
-  "tensorboard_dir": "/data/user_data/siqiouya/ckpts/infinisst-omni/gigaspeech-zh-s_v4_ner_baseline_aligned_rate1.0_k20_final/v0-20251229-061816/runs",
+  "tensorboard_dir": "/data/user_data/siqiouya/ckpts/infinisst-omni/gigaspeech-zh-s_v4_ner_baseline_aligned_rate1.0_k20_final/v1-20260106-064237/runs",
   "no_masked_softmax_fusion": false,
   "no_bias_dropout_fusion": false,
   "no_bias_swiglu_fusion": false,
@@ -200,7 +198,7 @@
   "adam_beta2": 0.95,
   "adam_eps": 1e-08,
   "sgd_momentum": 0.9,
-  "save": "/data/user_data/siqiouya/ckpts/infinisst-omni/gigaspeech-zh-s_v4_ner_baseline_aligned_rate1.0_k20_final/v0-20251229-061816",
+  "save": "/data/user_data/siqiouya/ckpts/infinisst-omni/gigaspeech-zh-s_v4_ner_baseline_aligned_rate1.0_k20_final/v1-20260106-064237",
   "save_interval": 200,
   "no_save_optim": true,
   "no_save_rng": true,
@@ -314,14 +312,14 @@
   "extra_megatron_kwargs": {},
   "add_version": true,
   "rank": 0,
-  "global_world_size": 2,
-  "local_world_size": 2,
+  "global_world_size": 4,
+  "local_world_size": 4,
   "model_suffix": "Qwen3-Omni-30B-A3B-Instruct",
   "model_info": "ModelInfo(model_type='qwen3_omni', model_dir='/data/user_data/siqiouya/ckpts/pretrained/llm/Qwen3-Omni-30B-A3B-Instruct', torch_dtype=torch.bfloat16, max_model_len=65536, quant_method=None, quant_bits=None, rope_scaling={'interleaved': True, 'mrope_section': [24, 20, 20], 'rope_type': 'default', 'type': 'default'}, is_moe_model=True, config=None, task_type='causal_lm', num_labels=None)",
-  "model_meta": "ModelMeta(model_type='qwen3_omni', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Instruct', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Thinking', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Captioner', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Captioner', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[])], template='qwen3_omni', get_function=<function get_model_tokenizer_qwen3_omni at 0x145a0fc46ac0>, model_arch=MultiModelKeys(arch_name='qwen3_omni', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['thinker.model'], aligner=['thinker.audio_tower.proj1', 'thinker.audio_tower.proj2', 'thinker.visual.merger', 'thinker.visual.merger_list'], vision_tower=['thinker.audio_tower', 'thinker.visual'], generator=['talker', 'token2wav']), architectures=['Qwen3OmniMoeForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, task_type=None, ignore_patterns=None, requires=['transformers>=4.57.dev0', 'soundfile', 'decord', 'qwen_omni_utils'], tags=['vision', 'video', 'audio'])",
+  "model_meta": "ModelMeta(model_type='qwen3_omni', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Instruct', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Thinking', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Captioner', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Captioner', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[])], template='qwen3_omni', get_function=<function get_model_tokenizer_qwen3_omni at 0x7f622c92ea20>, model_arch=MultiModelKeys(arch_name='qwen3_omni', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['thinker.model'], aligner=['thinker.audio_tower.proj1', 'thinker.audio_tower.proj2', 'thinker.visual.merger', 'thinker.visual.merger_list'], vision_tower=['thinker.audio_tower', 'thinker.visual'], generator=['talker', 'token2wav']), architectures=['Qwen3OmniMoeForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, task_type=None, ignore_patterns=None, requires=['transformers>=4.57.dev0', 'soundfile', 'decord', 'qwen_omni_utils'], tags=['vision', 'video', 'audio'])",
   "model_dir": "/data/user_data/siqiouya/ckpts/pretrained/llm/Qwen3-Omni-30B-A3B-Instruct",
   "hub": "<class 'swift.hub.hub.MSHub'>",
-  "megatron_model_meta": "MMGPTMegatronModelMeta(megatron_model_type='qwen3_omni', model_types=['qwen3_omni'], convert_mcore2hf=<function convert_mcore2hf_qwen3_omni at 0x1459792d13a0>, convert_hf2mcore=<function convert_hf2mcore_qwen3_omni at 0x1459792d1300>, model_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3VLGPTModel'>, convert_hf_config=<function convert_gpt_hf_config at 0x1459792a1a80>, get_transformer_layer_spec=None, model_provider=<function model_provider at 0x1459792a0860>, visual_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3Omni_Vit'>, extra_args_provider=None)",
+  "megatron_model_meta": "MMGPTMegatronModelMeta(megatron_model_type='qwen3_omni', model_types=['qwen3_omni'], convert_mcore2hf=<function convert_mcore2hf_qwen3_omni at 0x7f619c309300>, convert_hf2mcore=<function convert_hf2mcore_qwen3_omni at 0x7f619c309260>, model_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3VLGPTModel'>, convert_hf_config=<function convert_gpt_hf_config at 0x7f619c2dd9e0>, get_transformer_layer_spec=None, model_provider=<function model_provider at 0x7f619c2dc7c0>, visual_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3Omni_Vit'>, extra_args_provider=None)",
   "extra_args": {
     "model": "/data/user_data/siqiouya/ckpts/pretrained/llm/Qwen3-Omni-30B-A3B-Instruct/",
     "model_type": "qwen3_omni",
@@ -364,10 +362,8 @@
     "dataset": [
       "/data/group_data/li_lab/siqiouya/datasets/gigaspeech/manifests/train_s_zh_v4_ner_baseline_aligned_rate1.0_k20_final.jsonl"
     ],
-    "val_dataset": [
-      "/data/group_data/li_lab/siqiouya/datasets/gigaspeech/manifests/dev_zh_v4_ner_baseline_aligned_freq_k20_final.jsonl"
-    ],
-    "split_dataset_ratio": 0.0,
+    "val_dataset": [],
+    "split_dataset_ratio": 0.01,
     "data_seed": 42,
     "dataset_num_proc": 8,
     "load_from_cache_file": true,
@@ -478,7 +474,7 @@
     "mrope_interleaved": true,
     "add_version": true,
     "model_info": "ModelInfo(model_type='qwen3_omni', model_dir='/data/user_data/siqiouya/ckpts/pretrained/llm/Qwen3-Omni-30B-A3B-Instruct', torch_dtype=torch.bfloat16, max_model_len=65536, quant_method=None, quant_bits=None, rope_scaling={'interleaved': True, 'mrope_section': [24, 20, 20], 'rope_type': 'default', 'type': 'default'}, is_moe_model=True, config=None, task_type='causal_lm', num_labels=None)",
-    "model_meta": "ModelMeta(model_type='qwen3_omni', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Instruct', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Thinking', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Captioner', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Captioner', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[])], template='qwen3_omni', get_function=<function get_model_tokenizer_qwen3_omni at 0x145a0fc46ac0>, model_arch=MultiModelKeys(arch_name='qwen3_omni', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['thinker.model'], aligner=['thinker.audio_tower.proj1', 'thinker.audio_tower.proj2', 'thinker.visual.merger', 'thinker.visual.merger_list'], vision_tower=['thinker.audio_tower', 'thinker.visual'], generator=['talker', 'token2wav']), architectures=['Qwen3OmniMoeForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, task_type=None, ignore_patterns=None, requires=['transformers>=4.57.dev0', 'soundfile', 'decord', 'qwen_omni_utils'], tags=['vision', 'video', 'audio'])",
-    "megatron_model_meta": "MMGPTMegatronModelMeta(megatron_model_type='qwen3_omni', model_types=['qwen3_omni'], convert_mcore2hf=<function convert_mcore2hf_qwen3_omni at 0x1459792d13a0>, convert_hf2mcore=<function convert_hf2mcore_qwen3_omni at 0x1459792d1300>, model_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3VLGPTModel'>, convert_hf_config=<function convert_gpt_hf_config at 0x1459792a1a80>, get_transformer_layer_spec=None, model_provider=<function model_provider at 0x1459792a0860>, visual_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3Omni_Vit'>, extra_args_provider=None)"
+    "model_meta": "ModelMeta(model_type='qwen3_omni', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Instruct', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Thinking', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Captioner', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Captioner', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[])], template='qwen3_omni', get_function=<function get_model_tokenizer_qwen3_omni at 0x7f622c92ea20>, model_arch=MultiModelKeys(arch_name='qwen3_omni', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['thinker.model'], aligner=['thinker.audio_tower.proj1', 'thinker.audio_tower.proj2', 'thinker.visual.merger', 'thinker.visual.merger_list'], vision_tower=['thinker.audio_tower', 'thinker.visual'], generator=['talker', 'token2wav']), architectures=['Qwen3OmniMoeForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, task_type=None, ignore_patterns=None, requires=['transformers>=4.57.dev0', 'soundfile', 'decord', 'qwen_omni_utils'], tags=['vision', 'video', 'audio'])",
+    "megatron_model_meta": "MMGPTMegatronModelMeta(megatron_model_type='qwen3_omni', model_types=['qwen3_omni'], convert_mcore2hf=<function convert_mcore2hf_qwen3_omni at 0x7f619c309300>, convert_hf2mcore=<function convert_hf2mcore_qwen3_omni at 0x7f619c309260>, model_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3VLGPTModel'>, convert_hf_config=<function convert_gpt_hf_config at 0x7f619c2dd9e0>, get_transformer_layer_spec=None, model_provider=<function model_provider at 0x7f619c2dc7c0>, visual_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3Omni_Vit'>, extra_args_provider=None)"
   }
 }

model-00001-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b843a6dd5e312f7472964292d6a9ddefe591d7db066cbe96a3ab413b6753cc1e
+oid sha256:d81fc6df86edd04a50d4a05cd30eed81f6360b7fb31bcbcb0ff1e62b1c16d736
 size 4997899632

model-00002-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:229bbf4d9ff3fabfbd0a3d319f2925de94781ec7d485125198e5c8de65101136
+oid sha256:a0d25dae8b7014deef9d92b0f1de9339f8f6a00ec5917904cf3513cc14ef5bc4
 size 4997754216

model-00003-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:07bf11be6ba92e098973f2536545a1c3802a699f812205b69b4ecc773799e139
+oid sha256:139f2bceeed6974fdc24048d8d14a5a873956ccfb50554d46802b42246d08f6f
 size 4997754216

model-00004-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fa7d63f13d886a73a6d84daaea9efc1be0cf0d8f24d3b819207cc478f5dcf510
+oid sha256:4b64443f2559ad8b1042d27567d5277c87e61b1437aa84687090ec1c4798a0ec
 size 4997755648

model-00005-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e6864bd25c02b0c71e62d1de30b728ae5006e6b9c6100d29bdb2fced31a1674d
+oid sha256:128fb0751124ab837657148b11c3b55f2462c21d906986205fb554b2398da268
 size 4997755792

model-00006-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2ae2cf8b1c7496b69a9c1d3d5424dc5cab31fb190ab8df43360fde38753c0909
+oid sha256:5c8b50b104e7dd4d6780f6cb6e4e3b8c8f36667dc58064f021c616acef0a3bd8
 size 4997755792

model-00007-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cb021c9920b38b3a6a6cdbdd58651783fd176876ccbe2cc96be067d17b4f5425
+oid sha256:236d0bf86f5da060de8910360c3d4b681a8eb53835f192b368374ad4273b2379
 size 4997755792

model-00008-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:13ba1d85055ddab73176832a3c60834197a720849daf9980790b474e4b8aeaaf
+oid sha256:522d713e1f9fd180ae98c285a3d31effced73ae5ec0206801a03dc71e5511a90
 size 4997755792

model-00009-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0549205207272b170faf2653187d96ce9695f8fb9c38aa04a82307828b9e4a83
+oid sha256:2735c64bc1a2efc1fadd424e51fa0095ff4c4af81331ba1b596ad49ec2705d64
 size 4997755792

model-00010-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ea8545be6f76cfe5800284340fb2de591f16066f502d853acb13960e2a72f979
+oid sha256:a1ca1726bad25ef9cbfa8ba3b655c78a85732579912924d7cb130010b90032f0
 size 4997755792

model-00011-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c10946ae27fc95d321d29bda81ed2316f6bcb7d9b2e6925147d2113fc8f07d8a
+oid sha256:bcb298abc815975dcf3bd4b5db748e2e9aa26424b5bb45d2a29101e7ade79721
 size 4997755792

model-00012-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3deaa81e4b33747557be843421cc6b106a6d37480aa381a371636cbc9e38cf99
+oid sha256:0b14ae3d6846555c05e6f101a5641c8d8910754266a2772b878080745def4f1d
 size 4997755792

model-00013-of-00015.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:46da8f8c0a0b7b5db7df6d3a735beb41b80f602b6c85281ec4d22ef2ae700d53
+oid sha256:de313835b565efa533faccdc4f9050e5c3b2d36af9ef4e6bbbf02305fb483826
 size 4999771808