Upload folder using huggingface_hub

config.json

@@ -0,0 +1,36 @@
+{
+  "_name_or_path": "../../models/moe/routing/onlyRouterTrue_aveSVLossFalse_lr5e-03",
+  "architectures": [
+    "MixtralForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 7168,
+  "max_position_embeddings": 4096,
+  "mlp_bias": false,
+  "model_type": "mixtral",
+  "num_attention_heads": 16,
+  "num_experts_per_tok": 2,
+  "num_hidden_layers": 24,
+  "num_key_value_heads": 16,
+  "num_local_experts": 8,
+  "output_router_logits": true,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 10000,
+  "router_aux_loss_coef": 0.0,
+  "router_jitter_noise": 0.0,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.46.0",
+  "use_cache": true,
+  "vocab_size": 99584
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "transformers_version": "4.46.0"
+}

global_step13692/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d676431d531aebb07dbfafb4fba53176e6ac28959f36c84ef80659f6260c325
+size 27800298620

global_step13692/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37e9b557b5a1ad0e487aa7f8adb5eed53764b57c5b9c10424ee93d200d3bf645
+size 27800300284

global_step13692/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08f389447f17a8966fee8f29c93f763e72f4e28f585ecd31cc15526d368eb695
+size 27800300476

global_step13692/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6d4ac4b6cd45e84c57c9abcc65461182324715f9a711ed25ca5bde03cc3f529
+size 27800298684

global_step13692/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e952ef57c2bd4c5a694e079c24a4e90660ce31dfe2cabd83fbce4135c1ada1b7
+size 18533781408

latest

@@ -0,0 +1 @@
+global_step13692

model-00001-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61351c91a915689400aeb5f9a41b0a0d601706ad0fba739ad9f7d51e38d9e43b
+size 4988379392

model-00002-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09d4aea22fe1e0ace2aa78b880f8e26baf0fc1c1a8756d5805a30679799aa098
+size 4987299352

model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1b7e202b6a0a2c9aa482321ad9315891a3b79e7b6d8d57a9c8a28bc4e8f98e3
+size 4991535416

model-00004-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff9af63037c8a97b40004960955d22f5de9391211a71e9383832ab8d2683ebb8
+size 3566400304

model.safetensors.index.json

@@ -0,0 +1,754 @@
+{
+  "metadata": {
+    "total_size": 18533519360
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00004-of-00004.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.1.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.1.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.1.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.6.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.7.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.7.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.experts.7.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.block_sparse_moe.gate.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.1.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.1.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.1.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.6.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.7.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.7.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.experts.7.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.block_sparse_moe.gate.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.1.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.1.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.1.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.block_sparse_moe.gate.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.1.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.1.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.1.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.block_sparse_moe.gate.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.1.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.1.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.1.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.12.block_sparse_moe.gate.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.1.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.1.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.1.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.block_sparse_moe.gate.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.1.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.1.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.1.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.block_sparse_moe.gate.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.1.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.1.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.1.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.block_sparse_moe.gate.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.1.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.1.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.1.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.block_sparse_moe.gate.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.1.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.1.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.1.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.block_sparse_moe.gate.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.1.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.1.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.1.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.block_sparse_moe.gate.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.1.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.1.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.1.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.5.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.6.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.6.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.6.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.7.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.7.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.experts.7.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.19.block_sparse_moe.gate.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.1.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.1.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.1.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.6.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.7.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.7.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.experts.7.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.block_sparse_moe.gate.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.0.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.0.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.0.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.1.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.1.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.1.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.2.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.2.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.2.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.3.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.3.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.3.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.4.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.4.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.4.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.5.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.5.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.5.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.6.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.6.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.6.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.7.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.7.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.experts.7.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.block_sparse_moe.gate.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.0.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.0.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.0.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.1.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.1.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.1.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.2.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.2.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.2.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.3.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.3.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.3.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.4.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.4.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.4.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.5.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.5.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.5.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.6.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.6.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.6.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.7.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.7.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.experts.7.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.block_sparse_moe.gate.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.0.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.0.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.0.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.1.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.1.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.1.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.2.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.2.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.2.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.3.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.3.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.3.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.4.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.4.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.4.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.5.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.5.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.5.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.6.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.6.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.6.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.7.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.7.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.experts.7.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.block_sparse_moe.gate.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.0.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.0.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.0.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.1.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.1.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.1.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.2.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.2.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.2.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.3.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.3.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.3.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.4.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.4.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.4.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.5.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.5.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.5.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.6.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.6.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.6.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.7.w1.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.7.w2.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.experts.7.w3.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.block_sparse_moe.gate.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.1.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.1.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.1.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.6.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.7.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.7.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.experts.7.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.block_sparse_moe.gate.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.1.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.1.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.1.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.6.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.7.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.7.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.experts.7.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.block_sparse_moe.gate.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.1.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.1.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.1.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.6.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.7.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.7.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.experts.7.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.block_sparse_moe.gate.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.1.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.1.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.1.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.block_sparse_moe.gate.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.1.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.1.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.1.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.block_sparse_moe.gate.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.1.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.1.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.1.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.block_sparse_moe.gate.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.1.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.1.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.1.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.block_sparse_moe.gate.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.norm.weight": "model-00004-of-00004.safetensors"
+  }
+}

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5cf3bac380addaa516f941e6a276bb95c92ecfb3c08371ff9fba4b2faac6df0
+size 15024

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:219db464e75bbf2b6c41a71470b6182f4de4159f65fd71e3323fd50c658eea3a
+size 15024

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03b8316374c36bf601b0d74a53807a0d460d535e42771ed5f0f6db088f66ab63
+size 15024

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d6c7bdfadb70cba54ce8abee8b9fbcd1d5870419c1ccbd2e7362a47c09ba74e
+size 15024

special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "<CLS|LLM-jp>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "<MASK|LLM-jp>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<PAD|LLM-jp>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "<SEP|LLM-jp>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,83 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<MASK|LLM-jp>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<PAD|LLM-jp>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<CLS|LLM-jp>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<SEP|LLM-jp>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<EOD|LLM-jp>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "cls_token": "<CLS|LLM-jp>",
+  "eod_token": "</s>",
+  "eos_token": "</s>",
+  "extra_ids": 0,
+  "mask_token": "<MASK|LLM-jp>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<PAD|LLM-jp>",
+  "sep_token": "<SEP|LLM-jp>",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "unk_token": "<unk>"
+}

trainer_state.json

@@ -0,0 +1,758 @@
+{
+  "best_metric": 0.2237541824579239,
+  "best_model_checkpoint": "../../models/moe/moe/onlyRouterTrue_aveSVLossFalse_lr5e-03_ralc0.00/checkpoint-13692",
+  "epoch": 1.0,
+  "eval_steps": 3423,
+  "global_step": 13692,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.010005842827928718,
+      "grad_norm": 3.4739856719970703,
+      "learning_rate": 4.9987649590704944e-05,
+      "loss": 0.6176,
+      "step": 137
+    },
+    {
+      "epoch": 0.020011685655857436,
+      "grad_norm": 2.701633930206299,
+      "learning_rate": 4.9950610565428546e-05,
+      "loss": 0.5809,
+      "step": 274
+    },
+    {
+      "epoch": 0.03001752848378615,
+      "grad_norm": 2.7198803424835205,
+      "learning_rate": 4.988891951994058e-05,
+      "loss": 0.572,
+      "step": 411
+    },
+    {
+      "epoch": 0.04002337131171487,
+      "grad_norm": 2.594041109085083,
+      "learning_rate": 4.9802637407013966e-05,
+      "loss": 0.5511,
+      "step": 548
+    },
+    {
+      "epoch": 0.05002921413964359,
+      "grad_norm": 2.6721420288085938,
+      "learning_rate": 4.969184947620146e-05,
+      "loss": 0.5502,
+      "step": 685
+    },
+    {
+      "epoch": 0.0600350569675723,
+      "grad_norm": 2.6376290321350098,
+      "learning_rate": 4.9556665189606316e-05,
+      "loss": 0.5266,
+      "step": 822
+    },
+    {
+      "epoch": 0.07004089979550102,
+      "grad_norm": 2.9528913497924805,
+      "learning_rate": 4.93972181137301e-05,
+      "loss": 0.536,
+      "step": 959
+    },
+    {
+      "epoch": 0.08004674262342974,
+      "grad_norm": 2.715660810470581,
+      "learning_rate": 4.9213665787504655e-05,
+      "loss": 0.5056,
+      "step": 1096
+    },
+    {
+      "epoch": 0.09005258545135846,
+      "grad_norm": 3.309116840362549,
+      "learning_rate": 4.900618956663845e-05,
+      "loss": 0.5124,
+      "step": 1233
+    },
+    {
+      "epoch": 0.10005842827928718,
+      "grad_norm": 2.226123332977295,
+      "learning_rate": 4.877499444443122e-05,
+      "loss": 0.5199,
+      "step": 1370
+    },
+    {
+      "epoch": 0.1100642711072159,
+      "grad_norm": 3.922006130218506,
+      "learning_rate": 4.852030884923388e-05,
+      "loss": 0.4957,
+      "step": 1507
+    },
+    {
+      "epoch": 0.1200701139351446,
+      "grad_norm": 3.2229082584381104,
+      "learning_rate": 4.82423844187538e-05,
+      "loss": 0.4864,
+      "step": 1644
+    },
+    {
+      "epoch": 0.13007595676307332,
+      "grad_norm": 1.936515212059021,
+      "learning_rate": 4.7941495751428536e-05,
+      "loss": 0.4653,
+      "step": 1781
+    },
+    {
+      "epoch": 0.14008179959100203,
+      "grad_norm": 2.966193199157715,
+      "learning_rate": 4.7617940135113606e-05,
+      "loss": 0.4722,
+      "step": 1918
+    },
+    {
+      "epoch": 0.15008764241893077,
+      "grad_norm": 3.2486863136291504,
+      "learning_rate": 4.7272037253352276e-05,
+      "loss": 0.4477,
+      "step": 2055
+    },
+    {
+      "epoch": 0.16009348524685948,
+      "grad_norm": 2.4895386695861816,
+      "learning_rate": 4.690412886951786e-05,
+      "loss": 0.4657,
+      "step": 2192
+    },
+    {
+      "epoch": 0.1700993280747882,
+      "grad_norm": 2.049923896789551,
+      "learning_rate": 4.651457848914021e-05,
+      "loss": 0.4538,
+      "step": 2329
+    },
+    {
+      "epoch": 0.1801051709027169,
+      "grad_norm": 2.0172815322875977,
+      "learning_rate": 4.610377100075045e-05,
+      "loss": 0.4317,
+      "step": 2466
+    },
+    {
+      "epoch": 0.19011101373064562,
+      "grad_norm": 2.2022132873535156,
+      "learning_rate": 4.5672112295598404e-05,
+      "loss": 0.4336,
+      "step": 2603
+    },
+    {
+      "epoch": 0.20011685655857436,
+      "grad_norm": 4.102416038513184,
+      "learning_rate": 4.5220028866618837e-05,
+      "loss": 0.4226,
+      "step": 2740
+    },
+    {
+      "epoch": 0.21012269938650308,
+      "grad_norm": 2.130418539047241,
+      "learning_rate": 4.4747967387042424e-05,
+      "loss": 0.4242,
+      "step": 2877
+    },
+    {
+      "epoch": 0.2201285422144318,
+      "grad_norm": 2.399773120880127,
+      "learning_rate": 4.4256394269067967e-05,
+      "loss": 0.4097,
+      "step": 3014
+    },
+    {
+      "epoch": 0.2301343850423605,
+      "grad_norm": 1.86017644405365,
+      "learning_rate": 4.3745795203031904e-05,
+      "loss": 0.4034,
+      "step": 3151
+    },
+    {
+      "epoch": 0.2401402278702892,
+      "grad_norm": 1.667731761932373,
+      "learning_rate": 4.321667467753034e-05,
+      "loss": 0.4108,
+      "step": 3288
+    },
+    {
+      "epoch": 0.25,
+      "eval_loss": 0.48173779249191284,
+      "eval_runtime": 367.5367,
+      "eval_samples_per_second": 31.393,
+      "eval_steps_per_second": 7.85,
+      "step": 3423
+    },
+    {
+      "epoch": 0.2501460706982179,
+      "grad_norm": 5.194522857666016,
+      "learning_rate": 4.2669555480967783e-05,
+      "loss": 0.3973,
+      "step": 3425
+    },
+    {
+      "epoch": 0.26015191352614664,
+      "grad_norm": 2.7280290126800537,
+      "learning_rate": 4.210497818502509e-05,
+      "loss": 0.409,
+      "step": 3562
+    },
+    {
+      "epoch": 0.27015775635407535,
+      "grad_norm": 2.3584136962890625,
+      "learning_rate": 4.152350061055695e-05,
+      "loss": 0.3855,
+      "step": 3699
+    },
+    {
+      "epoch": 0.28016359918200406,
+      "grad_norm": 3.4275765419006348,
+      "learning_rate": 4.092569727644661e-05,
+      "loss": 0.4029,
+      "step": 3836
+    },
+    {
+      "epoch": 0.29016944200993283,
+      "grad_norm": 2.14209246635437,
+      "learning_rate": 4.031215883196239e-05,
+      "loss": 0.3836,
+      "step": 3973
+    },
+    {
+      "epoch": 0.30017528483786154,
+      "grad_norm": 2.825669288635254,
+      "learning_rate": 3.968349147317693e-05,
+      "loss": 0.3842,
+      "step": 4110
+    },
+    {
+      "epoch": 0.31018112766579026,
+      "grad_norm": 2.3074870109558105,
+      "learning_rate": 3.904031634402552e-05,
+      "loss": 0.3601,
+      "step": 4247
+    },
+    {
+      "epoch": 0.32018697049371897,
+      "grad_norm": 2.8916587829589844,
+      "learning_rate": 3.838326892259564e-05,
+      "loss": 0.38,
+      "step": 4384
+    },
+    {
+      "epoch": 0.3301928133216477,
+      "grad_norm": 1.9778506755828857,
+      "learning_rate": 3.7712998393253786e-05,
+      "loss": 0.3611,
+      "step": 4521
+    },
+    {
+      "epoch": 0.3401986561495764,
+      "grad_norm": 2.735363721847534,
+      "learning_rate": 3.703016700522999e-05,
+      "loss": 0.3653,
+      "step": 4658
+    },
+    {
+      "epoch": 0.3502044989775051,
+      "grad_norm": 3.4953627586364746,
+      "learning_rate": 3.6335449418293985e-05,
+      "loss": 0.3754,
+      "step": 4795
+    },
+    {
+      "epoch": 0.3602103418054338,
+      "grad_norm": 4.839205265045166,
+      "learning_rate": 3.562953203616925e-05,
+      "loss": 0.3468,
+      "step": 4932
+    },
+    {
+      "epoch": 0.37021618463336253,
+      "grad_norm": 2.1916751861572266,
+      "learning_rate": 3.491311232834357e-05,
+      "loss": 0.3524,
+      "step": 5069
+    },
+    {
+      "epoch": 0.38022202746129125,
+      "grad_norm": 3.1856038570404053,
+      "learning_rate": 3.418689814094646e-05,
+      "loss": 0.3501,
+      "step": 5206
+    },
+    {
+      "epoch": 0.39022787028921996,
+      "grad_norm": 1.9597059488296509,
+      "learning_rate": 3.345160699737394e-05,
+      "loss": 0.3474,
+      "step": 5343
+    },
+    {
+      "epoch": 0.4002337131171487,
+      "grad_norm": 2.523240566253662,
+      "learning_rate": 3.2707965389351925e-05,
+      "loss": 0.3355,
+      "step": 5480
+    },
+    {
+      "epoch": 0.41023955594507744,
+      "grad_norm": 16.007722854614258,
+      "learning_rate": 3.195670805913866e-05,
+      "loss": 0.3285,
+      "step": 5617
+    },
+    {
+      "epoch": 0.42024539877300615,
+      "grad_norm": 4.493948936462402,
+      "learning_rate": 3.119857727357527e-05,
+      "loss": 0.3232,
+      "step": 5754
+    },
+    {
+      "epoch": 0.43025124160093486,
+      "grad_norm": 3.0784451961517334,
+      "learning_rate": 3.0434322090701827e-05,
+      "loss": 0.3264,
+      "step": 5891
+    },
+    {
+      "epoch": 0.4402570844288636,
+      "grad_norm": 2.9652538299560547,
+      "learning_rate": 2.9664697619663472e-05,
+      "loss": 0.3147,
+      "step": 6028
+    },
+    {
+      "epoch": 0.4502629272567923,
+      "grad_norm": 2.1190903186798096,
+      "learning_rate": 2.8890464274637876e-05,
+      "loss": 0.3096,
+      "step": 6165
+    },
+    {
+      "epoch": 0.460268770084721,
+      "grad_norm": 7.0982985496521,
+      "learning_rate": 2.8112387023521115e-05,
+      "loss": 0.2968,
+      "step": 6302
+    },
+    {
+      "epoch": 0.4702746129126497,
+      "grad_norm": 1.3744592666625977,
+      "learning_rate": 2.733123463211434e-05,
+      "loss": 0.3036,
+      "step": 6439
+    },
+    {
+      "epoch": 0.4802804557405784,
+      "grad_norm": 1.526167631149292,
+      "learning_rate": 2.6547778904558018e-05,
+      "loss": 0.3069,
+      "step": 6576
+    },
+    {
+      "epoch": 0.49028629856850714,
+      "grad_norm": 1.452222466468811,
+      "learning_rate": 2.5762793920764124e-05,
+      "loss": 0.2992,
+      "step": 6713
+    },
+    {
+      "epoch": 0.5,
+      "eval_loss": 0.3556436002254486,
+      "eval_runtime": 367.6162,
+      "eval_samples_per_second": 31.386,
+      "eval_steps_per_second": 7.848,
+      "step": 6846
+    },
+    {
+      "epoch": 0.5002921413964359,
+      "grad_norm": 2.3508574962615967,
+      "learning_rate": 2.4977055271599893e-05,
+      "loss": 0.2923,
+      "step": 6850
+    },
+    {
+      "epoch": 0.5102979842243646,
+      "grad_norm": 2.161405324935913,
+      "learning_rate": 2.4191339292578617e-05,
+      "loss": 0.2866,
+      "step": 6987
+    },
+    {
+      "epoch": 0.5203038270522933,
+      "grad_norm": 1.180467963218689,
+      "learning_rate": 2.340642229681474e-05,
+      "loss": 0.2874,
+      "step": 7124
+    },
+    {
+      "epoch": 0.530309669880222,
+      "grad_norm": 2.1960806846618652,
+      "learning_rate": 2.262307980800109e-05,
+      "loss": 0.2649,
+      "step": 7261
+    },
+    {
+      "epoch": 0.5403155127081507,
+      "grad_norm": 1.662766933441162,
+      "learning_rate": 2.1842085794166068e-05,
+      "loss": 0.268,
+      "step": 7398
+    },
+    {
+      "epoch": 0.5503213555360794,
+      "grad_norm": 1.6623494625091553,
+      "learning_rate": 2.1064211902967904e-05,
+      "loss": 0.2538,
+      "step": 7535
+    },
+    {
+      "epoch": 0.5603271983640081,
+      "grad_norm": 1.6007744073867798,
+      "learning_rate": 2.02902266992815e-05,
+      "loss": 0.2654,
+      "step": 7672
+    },
+    {
+      "epoch": 0.570333041191937,
+      "grad_norm": 2.0483551025390625,
+      "learning_rate": 1.9520894905831154e-05,
+      "loss": 0.2567,
+      "step": 7809
+    },
+    {
+      "epoch": 0.5803388840198657,
+      "grad_norm": 1.1370702981948853,
+      "learning_rate": 1.8756976647619502e-05,
+      "loss": 0.2558,
+      "step": 7946
+    },
+    {
+      "epoch": 0.5903447268477944,
+      "grad_norm": 1.3925893306732178,
+      "learning_rate": 1.7999226700899093e-05,
+      "loss": 0.239,
+      "step": 8083
+    },
+    {
+      "epoch": 0.6003505696757231,
+      "grad_norm": 1.1552609205245972,
+      "learning_rate": 1.7248393747428747e-05,
+      "loss": 0.2389,
+      "step": 8220
+    },
+    {
+      "epoch": 0.6103564125036518,
+      "grad_norm": 1.9827595949172974,
+      "learning_rate": 1.6505219634751472e-05,
+      "loss": 0.2439,
+      "step": 8357
+    },
+    {
+      "epoch": 0.6203622553315805,
+      "grad_norm": 1.7899774312973022,
+      "learning_rate": 1.5770438643224794e-05,
+      "loss": 0.2471,
+      "step": 8494
+    },
+    {
+      "epoch": 0.6303680981595092,
+      "grad_norm": 1.239159345626831,
+      "learning_rate": 1.5044776760527727e-05,
+      "loss": 0.2329,
+      "step": 8631
+    },
+    {
+      "epoch": 0.6403739409874379,
+      "grad_norm": 1.22818922996521,
+      "learning_rate": 1.4328950964361143e-05,
+      "loss": 0.2316,
+      "step": 8768
+    },
+    {
+      "epoch": 0.6503797838153667,
+      "grad_norm": 1.210986614227295,
+      "learning_rate": 1.3623668514050391e-05,
+      "loss": 0.2221,
+      "step": 8905
+    },
+    {
+      "epoch": 0.6603856266432954,
+      "grad_norm": 3.446089506149292,
+      "learning_rate": 1.2929626251749854e-05,
+      "loss": 0.2269,
+      "step": 9042
+    },
+    {
+      "epoch": 0.6703914694712241,
+      "grad_norm": 2.4413158893585205,
+      "learning_rate": 1.2247509913940128e-05,
+      "loss": 0.2209,
+      "step": 9179
+    },
+    {
+      "epoch": 0.6803973122991528,
+      "grad_norm": 1.9606636762619019,
+      "learning_rate": 1.1577993453897933e-05,
+      "loss": 0.221,
+      "step": 9316
+    },
+    {
+      "epoch": 0.6904031551270815,
+      "grad_norm": 0.8826603293418884,
+      "learning_rate": 1.0921738375808166e-05,
+      "loss": 0.2041,
+      "step": 9453
+    },
+    {
+      "epoch": 0.7004089979550102,
+      "grad_norm": 1.2856823205947876,
+      "learning_rate": 1.0279393081176122e-05,
+      "loss": 0.2022,
+      "step": 9590
+    },
+    {
+      "epoch": 0.7104148407829389,
+      "grad_norm": 0.5094468593597412,
+      "learning_rate": 9.651592228185622e-06,
+      "loss": 0.2023,
+      "step": 9727
+    },
+    {
+      "epoch": 0.7204206836108676,
+      "grad_norm": 1.4797954559326172,
+      "learning_rate": 9.038956104635871e-06,
+      "loss": 0.1961,
+      "step": 9864
+    },
+    {
+      "epoch": 0.7304265264387964,
+      "grad_norm": 1.850120186805725,
+      "learning_rate": 8.442090015076842e-06,
+      "loss": 0.2063,
+      "step": 10001
+    },
+    {
+      "epoch": 0.7404323692667251,
+      "grad_norm": 1.438025712966919,
+      "learning_rate": 7.861583682748586e-06,
+      "loss": 0.1903,
+      "step": 10138
+    },
+    {
+      "epoch": 0.75,
+      "eval_loss": 0.254038542509079,
+      "eval_runtime": 367.4989,
+      "eval_samples_per_second": 31.396,
+      "eval_steps_per_second": 7.85,
+      "step": 10269
+    },
+    {
+      "epoch": 0.7504382120946538,
+      "grad_norm": 1.2179102897644043,
+      "learning_rate": 7.298010666915303e-06,
+      "loss": 0.1885,
+      "step": 10275
+    },
+    {
+      "epoch": 0.7604440549225825,
+      "grad_norm": 1.099284291267395,
+      "learning_rate": 6.751927796170044e-06,
+      "loss": 0.2005,
+      "step": 10412
+    },
+    {
+      "epoch": 0.7704498977505112,
+      "grad_norm": 1.2008761167526245,
+      "learning_rate": 6.2238746182698375e-06,
+      "loss": 0.1702,
+      "step": 10549
+    },
+    {
+      "epoch": 0.7804557405784399,
+      "grad_norm": 1.5368109941482544,
+      "learning_rate": 5.7143728670448095e-06,
+      "loss": 0.1947,
+      "step": 10686
+    },
+    {
+      "epoch": 0.7904615834063686,
+      "grad_norm": 1.0241951942443848,
+      "learning_rate": 5.223925946908093e-06,
+      "loss": 0.1809,
+      "step": 10823
+    },
+    {
+      "epoch": 0.8004674262342975,
+      "grad_norm": 1.369369387626648,
+      "learning_rate": 4.7530184354757675e-06,
+      "loss": 0.1889,
+      "step": 10960
+    },
+    {
+      "epoch": 0.8104732690622262,
+      "grad_norm": 0.7070077061653137,
+      "learning_rate": 4.30211560478837e-06,
+      "loss": 0.1796,
+      "step": 11097
+    },
+    {
+      "epoch": 0.8204791118901549,
+      "grad_norm": 1.7389978170394897,
+      "learning_rate": 3.871662961606784e-06,
+      "loss": 0.1843,
+      "step": 11234
+    },
+    {
+      "epoch": 0.8304849547180836,
+      "grad_norm": 0.9540761709213257,
+      "learning_rate": 3.4620858072370504e-06,
+      "loss": 0.1764,
+      "step": 11371
+    },
+    {
+      "epoch": 0.8404907975460123,
+      "grad_norm": 1.6178034543991089,
+      "learning_rate": 3.073788817318707e-06,
+      "loss": 0.1785,
+      "step": 11508
+    },
+    {
+      "epoch": 0.850496640373941,
+      "grad_norm": 1.0763949155807495,
+      "learning_rate": 2.7071556419920514e-06,
+      "loss": 0.1715,
+      "step": 11645
+    },
+    {
+      "epoch": 0.8605024832018697,
+      "grad_norm": 1.0416216850280762,
+      "learning_rate": 2.3625485268391893e-06,
+      "loss": 0.1609,
+      "step": 11782
+    },
+    {
+      "epoch": 0.8705083260297984,
+      "grad_norm": 1.1292046308517456,
+      "learning_rate": 2.040307954973572e-06,
+      "loss": 0.1739,
+      "step": 11919
+    },
+    {
+      "epoch": 0.8805141688577272,
+      "grad_norm": 1.6674988269805908,
+      "learning_rate": 1.7407523106315244e-06,
+      "loss": 0.155,
+      "step": 12056
+    },
+    {
+      "epoch": 0.8905200116856559,
+      "grad_norm": 0.8487703800201416,
+      "learning_rate": 1.4641775645981849e-06,
+      "loss": 0.1687,
+      "step": 12193
+    },
+    {
+      "epoch": 0.9005258545135846,
+      "grad_norm": 0.9765934944152832,
+      "learning_rate": 1.210856981778688e-06,
+      "loss": 0.1561,
+      "step": 12330
+    },
+    {
+      "epoch": 0.9105316973415133,
+      "grad_norm": 0.9121004343032837,
+      "learning_rate": 9.810408512034908e-07,
+      "loss": 0.1621,
+      "step": 12467
+    },
+    {
+      "epoch": 0.920537540169442,
+      "grad_norm": 1.1236246824264526,
+      "learning_rate": 7.749562387346088e-07,
+      "loss": 0.1749,
+      "step": 12604
+    },
+    {
+      "epoch": 0.9305433829973707,
+      "grad_norm": 1.0239607095718384,
+      "learning_rate": 5.928067627171158e-07,
+      "loss": 0.165,
+      "step": 12741
+    },
+    {
+      "epoch": 0.9405492258252994,
+      "grad_norm": 1.127655029296875,
+      "learning_rate": 4.347723927975417e-07,
+      "loss": 0.1662,
+      "step": 12878
+    },
+    {
+      "epoch": 0.9505550686532281,
+      "grad_norm": 0.9149182438850403,
+      "learning_rate": 3.01009272107991e-07,
+      "loss": 0.164,
+      "step": 13015
+    },
+    {
+      "epoch": 0.9605609114811569,
+      "grad_norm": 1.0919889211654663,
+      "learning_rate": 1.9164956299158322e-07,
+      "loss": 0.1723,
+      "step": 13152
+    },
+    {
+      "epoch": 0.9705667543090856,
+      "grad_norm": 2.5338237285614014,
+      "learning_rate": 1.0680131642176183e-07,
+      "loss": 0.1584,
+      "step": 13289
+    },
+    {
+      "epoch": 0.9805725971370143,
+      "grad_norm": 0.7331089377403259,
+      "learning_rate": 4.6548365244375446e-08,
+      "loss": 0.17,
+      "step": 13426
+    },
+    {
+      "epoch": 0.990578439964943,
+      "grad_norm": 1.282274603843689,
+      "learning_rate": 1.0950241348084422e-08,
+      "loss": 0.1564,
+      "step": 13563
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 0.2237541824579239,
+      "eval_runtime": 366.8443,
+      "eval_samples_per_second": 31.452,
+      "eval_steps_per_second": 7.864,
+      "step": 13692
+    }
+  ],
+  "logging_steps": 137,
+  "max_steps": 13692,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 3423,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.405329205216464e+19,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f587753ccef3612b7d0f7899e81af61598b3e06ec3aaf4ff41332eed819d9e8
+size 7416

zero_to_fp32.py

@@ -0,0 +1,674 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import json
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        state_dict_split = split_torch_state_dict_into_shards(state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard, output_path)
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)