Upload DeepseekV4ForCausalLM (#7)

- Upload DeepseekV4ForCausalLM (7f3a06fd45f01dd17bc405d02f3246d9ad0fc666)

chat_template.jinja

@@ -60,6 +60,16 @@
     {%- endif -%}
 {%- endfor -%}
 
+{%- if tools -%}
+    {%- set anchor_ns = namespace(found=false) -%}
+    {%- for m in mns.list -%}
+        {%- if m.role == "system" or m.role == "developer" -%}{%- set anchor_ns.found = true -%}{%- endif -%}
+    {%- endfor -%}
+    {%- if not anchor_ns.found -%}
+        {%- set mns.list = [{"role": "system", "content": ""}] + mns.list -%}
+    {%- endif -%}
+{%- endif -%}
+
 {%- set lu = namespace(idx=-1) -%}
 {%- for m in mns.list -%}
     {%- if m.role == "user" or m.role == "developer" -%}

config.json

@@ -5,26 +5,13 @@
   "attention_bias": false,
   "attention_dropout": 0.0,
   "bos_token_id": 0,
-  "compress_ratios": [
-    0,
-    0,
-    4,
-    128
-  ],
-  "compress_rope_parameters": {
-    "beta_fast": 32,
-    "beta_slow": 1,
-    "factor": 16,
-    "original_max_position_embeddings": 65536,
-    "partial_rotary_factor": 0.125,
-    "rope_theta": 160000.0,
-    "rope_type": "yarn",
-    "type": "yarn"
+  "compress_rates": {
+    "compressed_sparse_attention": 4,
+    "heavily_compressed_attention": 128
   },
   "compress_rope_theta": 160000.0,
   "dtype": "bfloat16",
   "eos_token_id": 1,
-  "first_k_dense_replace": 2,
   "hc_eps": 1e-06,
   "hc_mult": 4,
   "hc_sinkhorn_iters": 20,
@@ -35,18 +22,27 @@
   "index_n_heads": 64,
   "index_topk": 512,
   "initializer_range": 0.02,
-  "intermediate_size": 32,
-  "kv_lora_rank": null,
+  "layer_types": [
+    "sliding_attention",
+    "sliding_attention",
+    "compressed_sparse_attention",
+    "heavily_compressed_attention"
+  ],
   "max_position_embeddings": 1048576,
+  "mlp_bias": false,
+  "mlp_layer_types": [
+    "hash_moe",
+    "hash_moe",
+    "hash_moe",
+    "moe"
+  ],
   "model_type": "deepseek_v4",
-  "moe_intermediate_size": 2048,
-  "n_group": null,
-  "n_routed_experts": 4,
+  "moe_intermediate_size": 32,
+  "n_routed_experts": 256,
   "n_shared_experts": 1,
   "norm_topk_prob": true,
   "num_attention_heads": 4,
-  "num_experts_per_tok": 2,
-  "num_hash_layers": 3,
+  "num_experts_per_tok": 6,
   "num_hidden_layers": 4,
   "num_key_value_heads": 1,
   "num_nextn_predict_layers": 1,
@@ -55,21 +51,43 @@
   "output_router_logits": false,
   "pad_token_id": null,
   "partial_rotary_factor": 0.125,
-  "pretraining_tp": 1,
   "q_lora_rank": 1024,
-  "qk_nope_head_dim": 448,
   "qk_rope_head_dim": 64,
+  "quantization_config": {
+    "activation_scheme": "dynamic",
+    "fmt": "e4m3",
+    "quant_method": "fp8",
+    "scale_fmt": "ue8m0",
+    "weight_block_size": [
+      128,
+      128
+    ]
+  },
   "rms_norm_eps": 1e-06,
-  "rope_interleave": true,
   "rope_parameters": {
-    "beta_fast": 32,
-    "beta_slow": 1,
-    "factor": 16,
-    "original_max_position_embeddings": 65536,
+    "compress": {
+      "beta_fast": 32,
+      "beta_slow": 1,
+      "factor": 16,
+      "original_max_position_embeddings": 65536,
+      "partial_rotary_factor": 0.125,
+      "rope_theta": 160000,
+      "rope_type": "yarn",
+      "type": "yarn"
+    },
+    "main": {
+      "beta_fast": 32,
+      "beta_slow": 1,
+      "factor": 16,
+      "original_max_position_embeddings": 65536,
+      "partial_rotary_factor": 0.125,
+      "rope_theta": 10000,
+      "rope_type": "yarn",
+      "type": "yarn"
+    },
     "partial_rotary_factor": 0.125,
     "rope_theta": 10000.0,
-    "rope_type": "yarn",
-    "type": "yarn"
+    "rope_type": "default"
   },
   "rope_theta": 10000.0,
   "routed_scaling_factor": 1.5,
@@ -79,10 +97,8 @@
   "sliding_window": 128,
   "swiglu_limit": 10.0,
   "tie_word_embeddings": false,
-  "topk_group": null,
   "topk_method": "noaux_tc",
-  "transformers_version": "5.7.0.dev0",
+  "transformers_version": "5.8.0",
   "use_cache": true,
-  "v_head_dim": null,
   "vocab_size": 129280
 }

generation_config.json

@@ -5,5 +5,5 @@
   "eos_token_id": 1,
   "temperature": 1.0,
   "top_p": 1.0,
-  "transformers_version": "5.7.0.dev0"
+  "transformers_version": "5.8.0"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4076740c031a3b6670e03573ec7f9ea4095e54c5dde682ac8e2eb629572cd39c
-size 37758460
+oid sha256:c87b575cc8bfd39f7ded48954bb03fc673799196e342d1a49332a849a309d664
+size 75882058