Upload LlamaForCausalLM

config.json

@@ -0,0 +1,44 @@
+{
+  "_name_or_path": "Stockmark-2-100B-Instruct-beta-AWQ",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 0,
+  "eos_token_id": 1,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 9216,
+  "initializer_range": 0.02,
+  "intermediate_size": 32768,
+  "max_position_embeddings": 4096,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 72,
+  "num_hidden_layers": 86,
+  "num_key_value_heads": 8,
+  "pad_token_id": 2,
+  "pretraining_tp": 1,
+  "quantization_config": {
+    "backend": "autoawq",
+    "bits": 4,
+    "do_fuse": false,
+    "exllama_config": null,
+    "fuse_max_seq_len": null,
+    "group_size": 128,
+    "modules_to_fuse": null,
+    "modules_to_not_convert": null,
+    "quant_method": "awq",
+    "version": "gemm",
+    "zero_point": true
+  },
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 10000,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float16",
+  "transformers_version": "4.44.2",
+  "use_cache": false,
+  "vocab_size": 100352
+}

generation_config.json

@@ -0,0 +1,11 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 0,
+  "do_sample": true,
+  "eos_token_id": [
+    1,
+    7
+  ],
+  "pad_token_id": 2,
+  "transformers_version": "4.44.2"
+}

model-00001-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27720afeec0a9cf37bf1233264d75218940ae23bc2c8aab4be4e8afb100f53e0
+size 4948527648

model-00002-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:231f310bf69432e0821b630ed532d06b22ec17e235f0fd89470f5c512dc970e9
+size 4962101992

model-00003-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1f79c4179ea93469273e60722b56136cc282af2d4af6d74fa2d383f421aaa4a
+size 4864005376

model-00004-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3081a4b7ed094d31e643d341b64c52aff30e2056695c2fb69de37d9c167737e
+size 4962102096

model-00005-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54d68898417a53a042492b4047569f2772e3070cb05d7ffc0994799f2319e156
+size 4962102096

model-00006-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15a5a2b23373d137cab20035ecfac6ba8037451179d5ac4550d6979962348474
+size 4864005376

model-00007-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d21fecf3fc193b6a0960c2f6191984df3cd0b438375c935e5aa5f59b868bfd02
+size 4962102096

model-00008-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:595d4d4ee4fa8670f78fa3bc1c444ad732fbfe08171121d3a34699b17a453ea8
+size 4962102096

model-00009-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf5eababa6e04e1fecc4ce912f2bb3d35f0c4ba1364e2767c70347b48a717af2
+size 4864005376

model-00010-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f13282b7533a0eaad63a4239bf7fa205f5e937c75b685a0c3ae40c9f73fd5cbc
+size 4962102096

model-00011-of-00011.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53dc5e1ab24abeb50bdb1b3ce8f04aaf5a0dbe38cdc3ee345204c0102ecf3352
+size 3301085400