llama-cpp-python 0.2.80__tar.gz → 0.2.81__tar.gz

llama_cpp_python-0.2.81/.git/FETCH_HEAD

@@ -0,0 +1 @@
+01bddd669ca1208f1844ce8d0ba9872532641c9d		'01bddd669ca1208f1844ce8d0ba9872532641c9d' of https://github.com/abetlen/llama-cpp-python

llama_cpp_python-0.2.81/.git/HEAD

@@ -0,0 +1 @@
+01bddd669ca1208f1844ce8d0ba9872532641c9d

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/.git/config

@@ -9,7 +9,7 @@
 [gc]
 	auto = 0
 [http "https://github.com/"]
-	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzXzNiMU0xMUZRcTV0ME9Qb3JsWHNKc3BQcW5rQlZ5OTB4cW92cA==
+	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzX3NQR1RseE1NMTh2UElUT25MSDhMQjlUNnBkMEZieTI3UG5KZw==
 [submodule "vendor/llama.cpp"]
 	active = true
 	url = https://github.com/ggerganov/llama.cpp.git

llama_cpp_python-0.2.81/.git/logs/HEAD

@@ -0,0 +1 @@
+0000000000000000000000000000000000000000 01bddd669ca1208f1844ce8d0ba9872532641c9d runner <runner@fv-az651-306.(none)> 1719936227 +0000	checkout: moving from master to refs/tags/v0.2.81

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/HEAD

@@ -0,0 +1 @@
+968967376dc2c018d29f897c4883d335bbf384fb

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/.git/modules/vendor/llama.cpp/config

@@ -16,7 +16,7 @@
 [gc]
 	auto = 0
 [http "https://github.com/"]
-	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzXzNiMU0xMUZRcTV0ME9Qb3JsWHNKc3BQcW5rQlZ5OTB4cW92cA==
+	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzX3NQR1RseE1NMTh2UElUT25MSDhMQjlUNnBkMEZieTI3UG5KZw==
 [url "https://github.com/"]
 	insteadOf = git@github.com:
 	insteadOf = org-6826477@github.com:

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/logs/HEAD

@@ -0,0 +1,2 @@
+0000000000000000000000000000000000000000 968967376dc2c018d29f897c4883d335bbf384fb runner <runner@fv-az651-306.(none)> 1719936228 +0000	clone: from https://github.com/ggerganov/llama.cpp.git
+968967376dc2c018d29f897c4883d335bbf384fb 968967376dc2c018d29f897c4883d335bbf384fb runner <runner@fv-az651-306.(none)> 1719936229 +0000	checkout: moving from master to 968967376dc2c018d29f897c4883d335bbf384fb

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/logs/refs/heads/master

@@ -0,0 +1 @@
+0000000000000000000000000000000000000000 968967376dc2c018d29f897c4883d335bbf384fb runner <runner@fv-az651-306.(none)> 1719936228 +0000	clone: from https://github.com/ggerganov/llama.cpp.git

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/logs/refs/remotes/origin/HEAD

@@ -0,0 +1 @@
+0000000000000000000000000000000000000000 968967376dc2c018d29f897c4883d335bbf384fb runner <runner@fv-az651-306.(none)> 1719936228 +0000	clone: from https://github.com/ggerganov/llama.cpp.git

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/.git/modules/vendor/llama.cpp/modules/kompute/config

@@ -13,7 +13,7 @@
 [gc]
 	auto = 0
 [http "https://github.com/"]
-	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzXzNiMU0xMUZRcTV0ME9Qb3JsWHNKc3BQcW5rQlZ5OTB4cW92cA==
+	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzX3NQR1RseE1NMTh2UElUT25MSDhMQjlUNnBkMEZieTI3UG5KZw==
 [url "https://github.com/"]
 	insteadOf = git@github.com:
 	insteadOf = org-6826477@github.com:

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/modules/kompute/logs/HEAD

@@ -0,0 +1,2 @@
+0000000000000000000000000000000000000000 c339310f6ff914c5b94fb2353f01a33dfc35f64f runner <runner@fv-az651-306.(none)> 1719936229 +0000	clone: from https://github.com/nomic-ai/kompute.git
+c339310f6ff914c5b94fb2353f01a33dfc35f64f 4565194ed7c32d1d2efa32ceab4d3c6cae006306 runner <runner@fv-az651-306.(none)> 1719936230 +0000	checkout: moving from master to 4565194ed7c32d1d2efa32ceab4d3c6cae006306

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/modules/kompute/logs/refs/heads/master

@@ -0,0 +1 @@
+0000000000000000000000000000000000000000 c339310f6ff914c5b94fb2353f01a33dfc35f64f runner <runner@fv-az651-306.(none)> 1719936229 +0000	clone: from https://github.com/nomic-ai/kompute.git

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/modules/kompute/logs/refs/remotes/origin/HEAD

@@ -0,0 +1 @@
+0000000000000000000000000000000000000000 c339310f6ff914c5b94fb2353f01a33dfc35f64f runner <runner@fv-az651-306.(none)> 1719936229 +0000	clone: from https://github.com/nomic-ai/kompute.git

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/packed-refs

@@ -0,0 +1,2 @@
+# pack-refs with: peeled fully-peeled sorted 
+968967376dc2c018d29f897c4883d335bbf384fb refs/remotes/origin/master

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/refs/heads/master

@@ -0,0 +1 @@
+968967376dc2c018d29f897c4883d335bbf384fb

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/refs/tags/b3282

@@ -0,0 +1 @@
+968967376dc2c018d29f897c4883d335bbf384fb

llama_cpp_python-0.2.81/.git/modules/vendor/llama.cpp/shallow

@@ -0,0 +1 @@
+968967376dc2c018d29f897c4883d335bbf384fb

llama_cpp_python-0.2.81/.git/refs/tags/v0.2.81

@@ -0,0 +1 @@
+01bddd669ca1208f1844ce8d0ba9872532641c9d

llama_cpp_python-0.2.81/.git/shallow

@@ -0,0 +1 @@
+01bddd669ca1208f1844ce8d0ba9872532641c9d

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/.github/workflows/build-and-release.yaml

@@ -11,7 +11,7 @@ jobs:
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        os: [ubuntu-20.04, windows-2019, macos-11]
+        os: [ubuntu-20.04, windows-2019, macos-12]
 
     steps:
       - uses: actions/checkout@v4

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/.github/workflows/build-wheels-cuda.yaml

@@ -114,7 +114,7 @@ jobs:
             $env:LD_LIBRARY_PATH = $env:CONDA_PREFIX + '/lib:' + $env:LD_LIBRARY_PATH
           }
           $env:VERBOSE = '1'
-          $env:CMAKE_ARGS = '-DLLAMA_CUBLAS=on -DCMAKE_CUDA_ARCHITECTURES=all'
+          $env:CMAKE_ARGS = '-DLLAMA_CUDA=on -DCMAKE_CUDA_ARCHITECTURES=all'
           $env:CMAKE_ARGS = "-DLLAMA_CUDA_FORCE_MMQ=ON $env:CMAKE_ARGS"
           # if ($env:AVXVER -eq 'AVX') {
           $env:CMAKE_ARGS = $env:CMAKE_ARGS + ' -DLLAMA_AVX2=off -DLLAMA_FMA=off -DLLAMA_F16C=off'

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/CHANGELOG.md

@@ -7,6 +7,12 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.2.81]
+
+- feat: Update llama.cpp to ggerganov/llama.cpp@968967376dc2c018d29f897c4883d335bbf384fb
+- fix(ci): Fix CUDA wheels, use LLAMA_CUDA instead of removed LLAMA_CUBLAS by @abetlen in 4fb6fc12a02a68884c25dd9f6a421cacec7604c6
+- fix(ci): Fix MacOS release, use macos-12 image instead of removed macos-11 by @abetlen in 3a551eb5263fdbd24b36d7770856374c04e92788
+
 ## [0.2.80]
 
 - feat: Update llama.cpp to ggerganov/llama.cpp@023b8807e10bc3ade24a255f01c1ad2a01bb4228

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: llama_cpp_python
-Version: 0.2.80
+Version: 0.2.81
 Summary: Python bindings for the llama.cpp library
 Author-Email: Andrei Betlen <abetlen@gmail.com>
 License: MIT

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/llama_cpp/__init__.py

@@ -1,4 +1,4 @@
 from .llama_cpp import *
 from .llama import *
 
-__version__ = "0.2.80"
+__version__ = "0.2.81"

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/llama_cpp/llama_cpp.py

@@ -311,6 +311,7 @@ LLAMA_VOCAB_TYPE_UGM = 4
 #     LLAMA_VOCAB_PRE_TYPE_SMAUG          = 14,
 #     LLAMA_VOCAB_PRE_TYPE_PORO           = 15,
 #     LLAMA_VOCAB_PRE_TYPE_VIKING         = 16,
+#     LLAMA_VOCAB_PRE_TYPE_JAIS           = 17,
 # };
 LLAMA_VOCAB_PRE_TYPE_DEFAULT = 0
 LLAMA_VOCAB_PRE_TYPE_LLAMA3 = 1
@@ -329,6 +330,7 @@ LLAMA_VOCAB_PRE_TYPE_DBRX = 13
 LLAMA_VOCAB_PRE_TYPE_SMAUG = 14
 LLAMA_VOCAB_PRE_TYPE_PORO = 15
 LLAMA_VOCAB_PRE_TYPE_VIKING = 16
+LLAMA_VOCAB_PRE_TYPE_JAIS = 17
 
 
 # // note: these values should be synchronized with ggml_rope

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/llama_cpp/server/app.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import os
 import json
+import typing
 import contextlib
 
 from threading import Lock
@@ -88,11 +89,12 @@ def get_llama_proxy():
             llama_outer_lock.release()
 
 
-_ping_message_factory = None
+_ping_message_factory: typing.Optional[typing.Callable[[], bytes]] = None
 
-def set_ping_message_factory(factory):
-   global  _ping_message_factory
-   _ping_message_factory = factory
+
+def set_ping_message_factory(factory: typing.Callable[[], bytes]):
+    global _ping_message_factory
+    _ping_message_factory = factory
 
 
 def create_app(
@@ -155,20 +157,19 @@ def create_app(
 
 async def get_event_publisher(
     request: Request,
-    inner_send_chan: MemoryObjectSendStream,
-    iterator: Iterator,
-    on_complete=None,
+    inner_send_chan: MemoryObjectSendStream[typing.Any],
+    iterator: Iterator[typing.Any],
+    on_complete: typing.Optional[typing.Callable[[], None]] = None,
 ):
+    server_settings = next(get_server_settings())
+    interrupt_requests = server_settings.interrupt_requests if server_settings else False
     async with inner_send_chan:
         try:
             async for chunk in iterate_in_threadpool(iterator):
                 await inner_send_chan.send(dict(data=json.dumps(chunk)))
                 if await request.is_disconnected():
                     raise anyio.get_cancelled_exc_class()()
-                if (
-                    next(get_server_settings()).interrupt_requests
-                    and llama_outer_lock.locked()
-                ):
+                if interrupt_requests and llama_outer_lock.locked():
                     await inner_send_chan.send(dict(data="[DONE]"))
                     raise anyio.get_cancelled_exc_class()()
             await inner_send_chan.send(dict(data="[DONE]"))
@@ -268,6 +269,11 @@ async def create_completion(
     llama_proxy = await run_in_threadpool(
         lambda: exit_stack.enter_context(contextlib.contextmanager(get_llama_proxy)())
     )
+    if llama_proxy is None:
+        raise HTTPException(
+            status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
+            detail="Service is not available",
+        )
     if isinstance(body.prompt, list):
         assert len(body.prompt) <= 1
         body.prompt = body.prompt[0] if len(body.prompt) > 0 else ""
@@ -409,7 +415,7 @@ async def create_chat_completion(
                         {"role": "system", "content": "You are a helpful assistant."},
                         {"role": "user", "content": "Who won the world series in 2020"},
                     ],
-                    "response_format": { "type": "json_object" }
+                    "response_format": {"type": "json_object"},
                 },
             },
             "tool_calling": {
@@ -434,15 +440,15 @@ async def create_chat_completion(
                                     },
                                     "required": ["name", "age"],
                                 },
-                            }
+                            },
                         }
                     ],
                     "tool_choice": {
                         "type": "function",
                         "function": {
                             "name": "User",
-                        }
-                    }
+                        },
+                    },
                 },
             },
             "logprobs": {
@@ -454,7 +460,7 @@ async def create_chat_completion(
                         {"role": "user", "content": "What is the capital of France?"},
                     ],
                     "logprobs": True,
-                    "top_logprobs": 10
+                    "top_logprobs": 10,
                 },
             },
         }
@@ -468,6 +474,11 @@ async def create_chat_completion(
     llama_proxy = await run_in_threadpool(
         lambda: exit_stack.enter_context(contextlib.contextmanager(get_llama_proxy)())
     )
+    if llama_proxy is None:
+        raise HTTPException(
+            status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
+            detail="Service is not available",
+        )
     exclude = {
         "n",
         "logit_bias_type",

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/vendor/llama.cpp/common/build-info.cpp

@@ -1,4 +1,4 @@
 int LLAMA_BUILD_NUMBER = 1;
-char const *LLAMA_COMMIT = "023b880";
+char const *LLAMA_COMMIT = "9689673";
 char const *LLAMA_COMPILER = "gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0";
 char const *LLAMA_BUILD_TARGET = "x86_64-linux-gnu";

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/vendor/llama.cpp/convert-hf-to-gguf-update.py

@@ -86,6 +86,7 @@ models = [
     {"name": "poro-chat",      "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LumiOpen/Poro-34B-chat", },
     {"name": "jina-v2-code",   "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-code", },
     {"name": "viking",         "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LumiOpen/Viking-7B", }, # Also used for Viking 13B and 33B
+    {"name": "jais",           "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/core42/jais-13b", },
 ]
 
 

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/vendor/llama.cpp/convert-hf-to-gguf.py

@@ -490,6 +490,9 @@ class Model:
         if chkhsh == "7fc505bd3104ca1083b150b17d088b59534ede9bde81f0dd2090967d7fe52cee":
             # ref: https://huggingface.co/LumiOpen/Viking-7B
             res = "viking"
+        if chkhsh == "b53802fb28e26d645c3a310b34bfe07da813026ec7c7716883404d5e0f8b1901":
+            # ref: https://huggingface.co/core42/jais-13b
+            res = "jais"
 
         if res is None:
             logger.warning("\n")
@@ -2965,6 +2968,96 @@ class T5Model(Model):
         return [(self.map_tensor_name(name), data_torch)]
 
 
+@Model.register("JAISLMHeadModel")
+class JaisModel(Model):
+    model_arch = gguf.MODEL_ARCH.JAIS
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+        # SwigLU activation
+        assert self.hparams["activation_function"] == "swiglu"
+        # ALiBi position embedding
+        assert self.hparams["position_embedding_type"] == "alibi"
+
+        # Embeddings scale
+        self.embeddings_scale = 1.0
+        # note: For some JAIS flavors, output is tied to (same as) wte in original model
+        self.output_is_wte = False
+        if 'mup_embeddings_scale' in self.hparams:
+            self.output_is_wte = True   # Hack (?)
+            self.embeddings_scale = self.hparams['mup_embeddings_scale']
+        elif 'embeddings_scale' in self.hparams:
+            self.embeddings_scale = self.hparams['embeddings_scale']
+        else:
+            assert False
+
+        self.width_scale = 1.0
+        if 'mup_output_alpha' in self.hparams:
+            assert 'mup_width_scale' in self.hparams
+            self.width_scale = self.hparams['mup_output_alpha'] * self.hparams['mup_width_scale']
+        elif 'width_scale' in self.hparams:
+            self.width_scale = self.hparams['width_scale']
+        else:
+            assert False
+
+        self.max_alibi_bias = 8.0
+
+    def set_vocab(self):
+        self._set_vocab_gpt2()
+
+    def set_gguf_parameters(self):
+        self.gguf_writer.add_name(self.dir_model.name)
+        self.gguf_writer.add_block_count(self.hparams["n_layer"])
+        self.gguf_writer.add_context_length(self.hparams["n_positions"])
+        self.gguf_writer.add_embedding_length(self.hparams["n_embd"])
+        self.gguf_writer.add_feed_forward_length(self.hparams["n_inner"])
+        self.gguf_writer.add_head_count(self.hparams["n_head"])
+        self.gguf_writer.add_layer_norm_eps(self.hparams["layer_norm_epsilon"])
+        self.gguf_writer.add_file_type(self.ftype)
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        del bid  # unused
+
+        tensors: list[tuple[str, Tensor]] = []
+
+        # we don't need these
+        if name.endswith((".attn.bias")):
+            return tensors
+
+        if name.endswith(("relative_pe.slopes")):
+            # Calculate max ALiBi bias (this is the inverse of the ALiBi calculation)
+            # Some other models has max_alibi_bias spelled out explicitly in the hyperparams,
+            # but Jais's PyTorch model simply precalculates the slope values and places them
+            # in relative_pes.slopes
+            n_head_closest_log2 = 2 ** math.floor(math.log2(self.hparams["n_head"]))
+            first_val = float(data_torch._data[0])
+            self.max_alibi_bias = -round(math.log2(first_val) * n_head_closest_log2)
+
+            return tensors
+
+        if name.endswith((".c_attn.weight", ".c_proj.weight", ".c_fc.weight", ".c_fc2.weight")):
+            data_torch = data_torch.transpose(1, 0)
+
+        new_name = self.map_tensor_name(name)
+
+        if new_name == self.format_tensor_name(gguf.MODEL_TENSOR.TOKEN_EMBD):
+            tensors.append((new_name, data_torch * self.embeddings_scale))
+            if self.output_is_wte:
+                tensors.append((self.format_tensor_name(gguf.MODEL_TENSOR.OUTPUT), data_torch * self.width_scale))
+        elif new_name == self.format_tensor_name(gguf.MODEL_TENSOR.OUTPUT):
+            assert not self.output_is_wte
+            tensors.append((new_name, data_torch * self.width_scale))
+        else:
+            tensors.append((new_name, data_torch))
+
+        return tensors
+
+    def write_tensors(self):
+        super().write_tensors()
+        self.gguf_writer.add_max_alibi_bias(self.max_alibi_bias)
+
+
 ###### CONVERSION LOGIC ######
 
 

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/vendor/llama.cpp/gguf-py/gguf/constants.py

@@ -164,6 +164,7 @@ class MODEL_ARCH(IntEnum):
     DEEPSEEK2    = auto()
     BITNET       = auto()
     T5           = auto()
+    JAIS         = auto()
 
 
 class MODEL_TENSOR(IntEnum):
@@ -288,6 +289,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
     MODEL_ARCH.DEEPSEEK2:      "deepseek2",
     MODEL_ARCH.BITNET:         "bitnet",
     MODEL_ARCH.T5:             "t5",
+    MODEL_ARCH.JAIS:           "jais",
 }
 
 TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
@@ -954,6 +956,18 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.ENC_FFN_UP,
         MODEL_TENSOR.ENC_OUTPUT_NORM,
     ],
+    MODEL_ARCH.JAIS: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_QKV,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.FFN_NORM,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_GATE,
+        MODEL_TENSOR.FFN_UP,
+    ],
     # TODO
 }
 

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/vendor/llama.cpp/gguf-py/gguf/tensor_mapping.py

@@ -10,7 +10,7 @@ class TensorNameMap:
         # Token embeddings
         MODEL_TENSOR.TOKEN_EMBD: (
             "gpt_neox.embed_in",                         # gptneox
-            "transformer.wte",                           # gpt2 gpt-j mpt refact qwen dbrx
+            "transformer.wte",                           # gpt2 gpt-j mpt refact qwen dbrx jais
             "transformer.word_embeddings",               # falcon
             "word_embeddings",                           # bloom
             "model.embed_tokens",                        # llama-hf
@@ -49,7 +49,7 @@ class TensorNameMap:
         # Output
         MODEL_TENSOR.OUTPUT: (
             "embed_out",                 # gptneox
-            "lm_head",                   # gpt2 mpt falcon llama-hf baichuan qwen mamba dbrx
+            "lm_head",                   # gpt2 mpt falcon llama-hf baichuan qwen mamba dbrx jais
             "output",                    # llama-pth bloom internlm2
             "word_embeddings_for_head",  # persimmon
             "lm_head.linear",            # phi2
@@ -58,7 +58,7 @@ class TensorNameMap:
         # Output norm
         MODEL_TENSOR.OUTPUT_NORM: (
             "gpt_neox.final_layer_norm",               # gptneox
-            "transformer.ln_f",                        # gpt2 gpt-j falcon
+            "transformer.ln_f",                        # gpt2 gpt-j falcon jais
             "model.norm",                              # llama-hf baichuan internlm2
             "norm",                                    # llama-pth
             "transformer.norm_f",                      # mpt dbrx
@@ -81,7 +81,7 @@ class TensorNameMap:
         # Attention norm
         MODEL_TENSOR.ATTN_NORM: (
             "gpt_neox.layers.{bid}.input_layernorm",                # gptneox
-            "transformer.h.{bid}.ln_1",                             # gpt2 gpt-j refact qwen
+            "transformer.h.{bid}.ln_1",                             # gpt2 gpt-j refact qwen jais
             "transformer.blocks.{bid}.norm_1",                      # mpt
             "transformer.h.{bid}.input_layernorm",                  # falcon7b
             "h.{bid}.input_layernorm",                              # bloom
@@ -109,7 +109,7 @@ class TensorNameMap:
         # Attention query-key-value
         MODEL_TENSOR.ATTN_QKV: (
             "gpt_neox.layers.{bid}.attention.query_key_value",                     # gptneox
-            "transformer.h.{bid}.attn.c_attn",                                     # gpt2 qwen
+            "transformer.h.{bid}.attn.c_attn",                                     # gpt2 qwen jais
             "transformer.blocks.{bid}.attn.Wqkv",                                  # mpt
             "transformer.blocks.{bid}.norm_attn_norm.attn.Wqkv",                   # dbrx
             "transformer.h.{bid}.self_attention.query_key_value",                  # falcon
@@ -160,7 +160,7 @@ class TensorNameMap:
         # Attention output
         MODEL_TENSOR.ATTN_OUT: (
             "gpt_neox.layers.{bid}.attention.dense",                        # gptneox
-            "transformer.h.{bid}.attn.c_proj",                              # gpt2 refact qwen
+            "transformer.h.{bid}.attn.c_proj",                              # gpt2 refact qwen jais
             "transformer.blocks.{bid}.attn.out_proj",                       # mpt
             "transformer.h.{bid}.self_attention.dense",                     # falcon
             "h.{bid}.self_attention.dense",                                 # bloom
@@ -202,7 +202,7 @@ class TensorNameMap:
         # Feed-forward norm
         MODEL_TENSOR.FFN_NORM: (
             "gpt_neox.layers.{bid}.post_attention_layernorm",                # gptneox
-            "transformer.h.{bid}.ln_2",                                      # gpt2 refact qwen
+            "transformer.h.{bid}.ln_2",                                      # gpt2 refact qwen jais
             "h.{bid}.post_attention_layernorm",                              # bloom
             "transformer.blocks.{bid}.norm_2",                               # mpt
             "model.layers.{bid}.post_attention_layernorm",                   # llama-hf
@@ -239,7 +239,7 @@ class TensorNameMap:
         # Feed-forward up
         MODEL_TENSOR.FFN_UP: (
             "gpt_neox.layers.{bid}.mlp.dense_h_to_4h",                # gptneox
-            "transformer.h.{bid}.mlp.c_fc",                           # gpt2
+            "transformer.h.{bid}.mlp.c_fc",                           # gpt2 jais
             "transformer.blocks.{bid}.ffn.up_proj",                   # mpt
             "transformer.h.{bid}.mlp.dense_h_to_4h",                  # falcon
             "h.{bid}.mlp.dense_h_to_4h",                              # bloom
@@ -285,6 +285,7 @@ class TensorNameMap:
             "model.layers.{bid}.mlp.gate_proj",           # llama-hf refact
             "layers.{bid}.feed_forward.w1",               # llama-pth
             "transformer.h.{bid}.mlp.w2",                 # qwen
+            "transformer.h.{bid}.mlp.c_fc2",              # jais
             "model.layers.layers.{bid}.mlp.gate_proj",    # plamo
             "model.layers.{bid}.feed_forward.w1",         # internlm2
             "encoder.layers.{bid}.mlp.fc12",              # nomic-bert
@@ -308,7 +309,7 @@ class TensorNameMap:
         # Feed-forward down
         MODEL_TENSOR.FFN_DOWN: (
             "gpt_neox.layers.{bid}.mlp.dense_4h_to_h",                # gptneox
-            "transformer.h.{bid}.mlp.c_proj",                         # gpt2 refact qwen
+            "transformer.h.{bid}.mlp.c_proj",                         # gpt2 refact qwen jais
             "transformer.blocks.{bid}.ffn.down_proj",                 # mpt
             "transformer.h.{bid}.mlp.dense_4h_to_h",                  # falcon
             "h.{bid}.mlp.dense_4h_to_h",                              # bloom

{llama_cpp_python-0.2.80 → llama_cpp_python-0.2.81}/vendor/llama.cpp/include/llama.h

@@ -89,6 +89,7 @@ extern "C" {
         LLAMA_VOCAB_PRE_TYPE_SMAUG          = 14,
         LLAMA_VOCAB_PRE_TYPE_PORO           = 15,
         LLAMA_VOCAB_PRE_TYPE_VIKING         = 16,
+        LLAMA_VOCAB_PRE_TYPE_JAIS           = 17,
     };
 
     // note: these values should be synchronized with ggml_rope