Upload folder using huggingface_hub

config.json

@@ -0,0 +1,49 @@
+{
+  "_name_or_path": "./",
+  "architectures": [
+    "QWenLMHeadModel"
+  ],
+  "attn_dropout_prob": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_qwen.QWenConfig",
+    "AutoModelForCausalLM": "modeling_qwen.QWenLMHeadModel"
+  },
+  "bf16": true,
+  "emb_dropout_prob": 0.0,
+  "fp16": false,
+  "fp32": false,
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 22016,
+  "kv_channels": 128,
+  "layer_norm_epsilon": 1e-06,
+  "max_position_embeddings": 8192,
+  "model_type": "qwen",
+  "no_bias": true,
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "onnx_safe": null,
+  "rotary_emb_base": 10000,
+  "rotary_pct": 1.0,
+  "scale_attn_weights": true,
+  "seq_length": 2048,
+  "tie_word_embeddings": false,
+  "tokenizer_type": "QWenTokenizer",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.31.0",
+  "use_cache": true,
+  "use_dynamic_ntk": true,
+  "use_flash_attn": false,
+  "use_logn_attn": true,
+  "visual": {
+    "heads": 16,
+    "image_size": 448,
+    "image_start_id": 151857,
+    "layers": 48,
+    "mlp_ratio": 4.9231,
+    "output_dim": 4096,
+    "patch_size": 14,
+    "width": 1664
+  },
+  "vocab_size": 151936
+}

configuration_qwen.py

@@ -0,0 +1,65 @@
+# Copyright (c) Alibaba Cloud.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+from transformers import PretrainedConfig
+
+
+class QWenConfig(PretrainedConfig):
+    model_type = "qwen"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=151936,
+        hidden_size=4096,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        emb_dropout_prob=0.0,
+        attn_dropout_prob=0.0,
+        layer_norm_epsilon=1e-6,
+        initializer_range=0.02,
+        max_position_embeddings=8192,
+        scale_attn_weights=True,
+        use_cache=True,
+        bf16=False,
+        fp16=False,
+        fp32=False,
+        kv_channels=128,
+        rotary_pct=1.0,
+        rotary_emb_base=10000,
+        use_dynamic_ntk=True,
+        use_logn_attn=True,
+        use_flash_attn="auto",
+        intermediate_size=22016,
+        no_bias=True,
+        tie_word_embeddings=False,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.emb_dropout_prob = emb_dropout_prob
+        self.attn_dropout_prob = attn_dropout_prob
+        self.layer_norm_epsilon = layer_norm_epsilon
+        self.initializer_range = initializer_range
+        self.scale_attn_weights = scale_attn_weights
+        self.use_cache = use_cache
+        self.max_position_embeddings = max_position_embeddings
+        self.bf16 = bf16
+        self.fp16 = fp16
+        self.fp32 = fp32
+        self.kv_channels = kv_channels
+        self.rotary_pct = rotary_pct
+        self.rotary_emb_base = rotary_emb_base
+        self.use_dynamic_ntk = use_dynamic_ntk
+        self.use_logn_attn = use_logn_attn
+        self.use_flash_attn = use_flash_attn
+        self.no_bias = no_bias
+        super().__init__(
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs
+        )

generation_config.json

@@ -0,0 +1,11 @@
+{
+  "chat_format": "chatml",
+  "do_sample": true,
+  "eos_token_id": 151643,
+  "max_new_tokens": 512,
+  "max_window_size": 6144,
+  "pad_token_id": 151643,
+  "top_k": 0,
+  "top_p": 0.5,
+  "transformers_version": "4.31.0"
+}

modeling_qwen.py

@@ -0,0 +1,1264 @@
+# Copyright (c) Alibaba Cloud.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+import importlib
+import math
+from typing import TYPE_CHECKING, Optional, Tuple, Union, Callable, List, Any, Generator
+
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch.cuda.amp import autocast
+
+from torch.nn import CrossEntropyLoss
+from transformers import PreTrainedTokenizer, GenerationConfig, StoppingCriteriaList
+from transformers.generation.logits_process import LogitsProcessorList
+
+if TYPE_CHECKING:
+    from transformers.generation.streamers import BaseStreamer
+from transformers.generation.utils import GenerateOutput
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPast,
+    CausalLMOutputWithPast,
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.utils import logging
+
+try:
+    from einops import rearrange
+except ImportError:
+    rearrange = None
+from torch import nn
+
+SUPPORT_CUDA = torch.cuda.is_available()
+SUPPORT_BF16 = SUPPORT_CUDA and torch.cuda.is_bf16_supported()
+SUPPORT_FP16 = SUPPORT_CUDA and torch.cuda.get_device_capability(0)[0] >= 7
+
+from .configuration_qwen import QWenConfig
+from .qwen_generation_utils import (
+    HistoryType,
+    make_context,
+    decode_tokens,
+    get_stop_words_ids,
+    StopWordsLogitsProcessor,
+)
+from .visual import VisionTransformer
+
+
+logger = logging.get_logger(__name__)
+
+_CHECKPOINT_FOR_DOC = "qwen"
+_CONFIG_FOR_DOC = "QWenConfig"
+
+QWen_PRETRAINED_MODEL_ARCHIVE_LIST = ["qwen-7b"]
+
+_ERROR_BAD_CHAT_FORMAT = """\
+We detect you are probably using the pretrained model (rather than chat model) for chatting, since the chat_format in generation_config is not "chatml".
+If you are directly using the model downloaded from Huggingface, please make sure you are using our "Qwen/Qwen-7B-Chat" Huggingface model (rather than "Qwen/Qwen-7B") when you call model.chat().
+我们检测到您可能在使用预训练模型（而非chat模型）进行多轮chat，因为您当前在generation_config指定的chat_format，并未设置为我们在对话中所支持的"chatml"格式。
+如果您在直接使用我们从Huggingface提供的模型，请确保您在调用model.chat()时，使用的是"Qwen/Qwen-7B-Chat"模型（而非"Qwen/Qwen-7B"预训练模型）。
+"""
+
+_SENTINEL = object()
+_ERROR_STREAM_IN_CHAT = """\
+Pass argument `stream` to model.chat() is buggy, deprecated, and marked for removal. Please use model.chat_stream(...) instead of model.chat(..., stream=True).
+向model.chat()传入参数stream的用法可能存在Bug，该用法已被废弃，将在未来被移除。请使用model.chat_stream(...)代替model.chat(..., stream=True)。
+"""
+
+apply_rotary_emb_func = None
+rms_norm = None
+flash_attn_unpadded_func = None
+
+
+def _import_flash_attn():
+    global apply_rotary_emb_func, rms_norm, flash_attn_unpadded_func
+    try:
+        from flash_attn.layers.rotary import apply_rotary_emb_func as __apply_rotary_emb_func
+        apply_rotary_emb_func = __apply_rotary_emb_func
+    except ImportError:
+        logger.warn(
+            "Warning: import flash_attn rotary fail, please install FlashAttention rotary to get higher efficiency "
+            "https://github.com/Dao-AILab/flash-attention/tree/main/csrc/rotary"
+        )
+
+    try:
+        from flash_attn.ops.rms_norm import rms_norm as __rms_norm
+        rms_norm = __rms_norm
+    except ImportError:
+        logger.warn(
+            "Warning: import flash_attn rms_norm fail, please install FlashAttention layer_norm to get higher efficiency "
+            "https://github.com/Dao-AILab/flash-attention/tree/main/csrc/layer_norm"
+        )
+
+    try:
+        import flash_attn
+        if not hasattr(flash_attn, '__version__'):
+            from flash_attn.flash_attn_interface import flash_attn_unpadded_func as __flash_attn_unpadded_func
+        else:
+            if int(flash_attn.__version__.split(".")[0]) >= 2:
+                from flash_attn.flash_attn_interface import flash_attn_varlen_func as __flash_attn_unpadded_func
+            else:
+                from flash_attn.flash_attn_interface import flash_attn_unpadded_func as __flash_attn_unpadded_func
+        flash_attn_unpadded_func = __flash_attn_unpadded_func
+    except ImportError:
+        logger.warn(
+            "Warning: import flash_attn fail, please install FlashAttention to get higher efficiency "
+            "https://github.com/Dao-AILab/flash-attention"
+        )
+
+# Copied from transformers.models.bart.modeling_bart._make_causal_mask
+def _make_causal_mask(
+    input_ids_shape: torch.Size, dtype: torch.dtype, device: torch.device, past_key_values_length: int = 0
+):
+    """
+    Make causal mask used for bi-directional self-attention.
+    """
+    bsz, tgt_len = input_ids_shape
+    mask = torch.full((tgt_len, tgt_len), torch.finfo(dtype).min, device=device)
+    mask_cond = torch.arange(mask.size(-1), device=device)
+    mask.masked_fill_(mask_cond < (mask_cond + 1).view(mask.size(-1), 1), 0)
+    mask = mask.to(dtype)
+
+    if past_key_values_length > 0:
+        mask = torch.cat([torch.zeros(tgt_len, past_key_values_length, dtype=dtype, device=device), mask], dim=-1)
+    return mask[None, None, :, :].expand(bsz, 1, tgt_len, tgt_len + past_key_values_length)
+
+
+# Copied from transformers.models.bart.modeling_bart._expand_mask
+def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int] = None):
+    """
+    Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`.
+    """
+    bsz, src_len = mask.size()
+    tgt_len = tgt_len if tgt_len is not None else src_len
+
+    expanded_mask = mask[:, None, None, :].expand(bsz, 1, tgt_len, src_len).to(dtype)
+
+    inverted_mask = 1.0 - expanded_mask
+
+    return inverted_mask.masked_fill(inverted_mask.to(torch.bool), torch.finfo(dtype).min)
+
+
+class FlashSelfAttention(torch.nn.Module):
+    def __init__(
+        self,
+        causal=False,
+        softmax_scale=None,
+        attention_dropout=0.0,
+    ):
+        super().__init__()
+        assert flash_attn_unpadded_func is not None, (
+            "Please install FlashAttention first, " "e.g., with pip install flash-attn"
+        )
+        assert (
+            rearrange is not None
+        ), "Please install einops first, e.g., with pip install einops"
+        self.causal = causal
+        self.softmax_scale = softmax_scale
+        self.dropout_p = attention_dropout
+
+    def forward(self, q, k, v):
+        assert all((i.dtype in [torch.float16, torch.bfloat16] for i in (q, k, v)))
+        assert all((i.is_cuda for i in (q, k, v)))
+        batch_size, seqlen_q = q.shape[0], q.shape[1]
+        seqlen_k = k.shape[1]
+        q, k, v = [rearrange(x, "b s ... -> (b s) ...") for x in [q, k, v]]
+        cu_seqlens_q = torch.arange(
+            0,
+            (batch_size + 1) * seqlen_q,
+            step=seqlen_q,
+            dtype=torch.int32,
+            device=q.device,
+        )
+
+        if self.training:
+            assert seqlen_k == seqlen_q
+
+            is_causal = self.causal
+            cu_seqlens_k = cu_seqlens_q
+        else:
+            is_causal = seqlen_q == seqlen_k
+            cu_seqlens_k = torch.arange(
+                0,
+                (batch_size + 1) * seqlen_k,
+                step=seqlen_k,
+                dtype=torch.int32,
+                device=q.device,
+            )
+            self.dropout_p = 0
+        output = flash_attn_unpadded_func(
+            q,
+            k,
+            v,
+            cu_seqlens_q,
+            cu_seqlens_k,
+            seqlen_q,
+            seqlen_k,
+            self.dropout_p,
+            softmax_scale=self.softmax_scale,
+            causal=is_causal,
+        )
+
+        output = rearrange(output, "(b s) ... -> b s ...", b=batch_size)
+        return output
+
+
+class QWenAttention(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+
+        max_positions = config.max_position_embeddings
+        self.register_buffer(
+            "bias",
+            torch.tril(
+                torch.ones((max_positions, max_positions), dtype=torch.bool)
+            ).view(1, 1, max_positions, max_positions),
+            persistent=False,
+        )
+        self.register_buffer("masked_bias", torch.tensor(-1e4), persistent=False)
+        self.seq_length = config.seq_length
+
+        self.hidden_size = config.hidden_size
+        self.split_size = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.hidden_size // self.num_heads
+
+        self.use_flash_attn = config.use_flash_attn
+        self.scale_attn_weights = True
+
+        self.projection_size = config.kv_channels * config.num_attention_heads
+
+        assert self.projection_size % config.num_attention_heads == 0
+        self.hidden_size_per_attention_head = (
+            self.projection_size // config.num_attention_heads
+        )
+
+        self.c_attn = nn.Linear(config.hidden_size, 3 * self.projection_size)
+
+        self.c_proj = nn.Linear(
+            config.hidden_size, self.projection_size, bias=not config.no_bias
+        )
+
+        self.is_fp32 = not (config.bf16 or config.fp16)
+        if (
+            self.use_flash_attn
+            and flash_attn_unpadded_func is not None
+            and not self.is_fp32
+        ):
+            self.core_attention_flash = FlashSelfAttention(
+                causal=True, attention_dropout=config.attn_dropout_prob
+            )
+
+        self.bf16 = config.bf16
+
+        if config.rotary_pct == 1.0:
+            self.rotary_ndims = None
+        else:
+            assert config.rotary_pct < 1
+            self.rotary_ndims = int(
+                self.hidden_size_per_attention_head * config.rotary_pct
+            )
+        dim = (
+            self.rotary_ndims
+            if self.rotary_ndims is not None
+            else self.hidden_size_per_attention_head
+        )
+        self.rotary_emb = RotaryEmbedding(dim, base=config.rotary_emb_base)
+
+        self.use_dynamic_ntk = config.use_dynamic_ntk
+        self.use_logn_attn = config.use_logn_attn
+
+        logn_list = [
+            math.log(i, self.seq_length) if i > self.seq_length else 1
+            for i in range(1, 32768)
+        ]
+        self.logn_tensor = torch.tensor(logn_list)[None, :, None, None]
+        self._ntk_cached = 1.0
+
+        self.attn_dropout = nn.Dropout(config.attn_dropout_prob)
+
+    def _attn(self, query, key, value, attention_mask=None, head_mask=None):
+        attn_weights = torch.matmul(query, key.transpose(-1, -2))
+
+        if self.scale_attn_weights:
+            attn_weights = attn_weights / torch.full(
+                [],
+                value.size(-1) ** 0.5,
+                dtype=attn_weights.dtype,
+                device=attn_weights.device,
+            )
+
+        query_length, key_length = query.size(-2), key.size(-2)
+        # causal_mask = self.bias[
+        #     :, :, key_length - query_length : key_length, :key_length
+        # ]
+        # mask_value = torch.finfo(attn_weights.dtype).min
+        # mask_value = torch.full([], mask_value, dtype=attn_weights.dtype).to(
+        #     attn_weights.device
+        # )
+        # attn_weights = torch.where(
+        #     causal_mask, attn_weights.to(attn_weights.dtype), mask_value
+        # )
+        attn_weights = attn_weights + attention_mask
+
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1)
+
+        attn_weights = attn_weights.type(value.dtype)
+        attn_weights = self.attn_dropout(attn_weights)
+
+        if head_mask is not None:
+            attn_weights = attn_weights * head_mask
+
+        attn_output = torch.matmul(attn_weights, value)
+        attn_output = attn_output.transpose(1, 2)
+
+        return attn_output, attn_weights
+
+    def _upcast_and_reordered_attn(
+        self, query, key, value, attention_mask=None, head_mask=None
+    ):
+        bsz, num_heads, q_seq_len, dk = query.size()
+        _, _, k_seq_len, _ = key.size()
+
+        attn_weights = torch.empty(
+            bsz * num_heads,
+            q_seq_len,
+            k_seq_len,
+            dtype=torch.float32,
+            device=query.device,
+        )
+
+        scale_factor = 1.0
+        if self.scale_attn_weights:
+            scale_factor /= float(value.size(-1)) ** 0.5
+
+        with autocast(enabled=False):
+            q, k = query.reshape(-1, q_seq_len, dk), key.transpose(-1, -2).reshape(
+                -1, dk, k_seq_len
+            )
+            attn_weights = torch.baddbmm(
+                attn_weights, q.float(), k.float(), beta=0, alpha=scale_factor
+            )
+            attn_weights = attn_weights.reshape(bsz, num_heads, q_seq_len, k_seq_len)
+
+        query_length, key_length = query.size(-2), key.size(-2)
+        causal_mask = self.bias[
+            :, :, key_length - query_length : key_length, :key_length
+        ]
+        mask_value = torch.finfo(attn_weights.dtype).min
+        mask_value = torch.tensor(mask_value, dtype=attn_weights.dtype).to(
+            attn_weights.device
+        )
+        attn_weights = torch.where(causal_mask, attn_weights, mask_value)
+
+        if attention_mask is not None:
+            attn_weights = attn_weights + attention_mask
+
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1)
+
+        if attn_weights.dtype != torch.float32:
+            raise RuntimeError(
+                "Error with upcasting, attn_weights does not have dtype torch.float32"
+            )
+        attn_weights = attn_weights.type(value.dtype)
+        attn_weights = self.attn_dropout(attn_weights)
+
+        if head_mask is not None:
+            attn_weights = attn_weights * head_mask
+
+        attn_output = torch.matmul(attn_weights, value)
+
+        return attn_output, attn_weights
+
+    def _split_heads(self, tensor, num_heads, attn_head_size):
+        new_shape = tensor.size()[:-1] + (num_heads, attn_head_size)
+        tensor = tensor.view(new_shape)
+        return tensor
+
+    def _merge_heads(self, tensor, num_heads, attn_head_size):
+        tensor = tensor.contiguous()
+        new_shape = tensor.size()[:-2] + (num_heads * attn_head_size,)
+        return tensor.view(new_shape)
+
+    def forward(
+        self,
+        hidden_states: Optional[Tuple[torch.FloatTensor]],
+        layer_past: Optional[Tuple[torch.Tensor]] = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+    ):
+
+        mixed_x_layer = self.c_attn(hidden_states)
+        query, key, value = mixed_x_layer.split(self.split_size, dim=2)
+
+        query = self._split_heads(query, self.num_heads, self.head_dim)
+        key = self._split_heads(key, self.num_heads, self.head_dim)
+        value = self._split_heads(value, self.num_heads, self.head_dim)
+
+        kv_seq_len = hidden_states.size()[1]
+        if layer_past:
+            # layer past[0] shape: bs * seq_len * head_num * dim
+            kv_seq_len += layer_past[0].shape[1]
+        if (
+            self.use_dynamic_ntk
+            and kv_seq_len == hidden_states.size()[1]
+            and not self.training
+        ):
+            context_value = math.log(kv_seq_len / self.seq_length, 2) + 1
+            ntk_alpha = 2 ** math.ceil(context_value) - 1
+            ntk_alpha = max(ntk_alpha, 1)
+            self._ntk_cached = ntk_alpha
+        else:
+            ntk_alpha = self._ntk_cached
+        rotary_pos_emb = self.rotary_emb(kv_seq_len, ntk_alpha=ntk_alpha).to(
+            hidden_states.device
+        )
+
+        if rotary_pos_emb is not None:
+            if isinstance(rotary_pos_emb, tuple):
+                rotary_pos_emb = rotary_pos_emb
+            else:
+                rotary_pos_emb = (rotary_pos_emb,) * 2
+
+        if rotary_pos_emb is not None:
+            q_pos_emb, k_pos_emb = rotary_pos_emb
+            # Slice the pos emb for current inference
+            cur_len = query.shape[1]
+            q_pos_emb = q_pos_emb[:, -cur_len:, :, :]
+            k_pos_emb = k_pos_emb[:, -cur_len:, :, :]
+            query = apply_rotary_pos_emb(query, q_pos_emb)
+            key = apply_rotary_pos_emb(key, k_pos_emb)
+
+        if layer_past is not None:
+            past_key, past_value = layer_past[0], layer_past[1]
+            key = torch.cat((past_key, key), dim=1)
+            value = torch.cat((past_value, value), dim=1)
+
+        if use_cache:
+            present = (key, value)
+        else:
+            present = None
+
+        if self.use_logn_attn and not self.training:
+            if self.logn_tensor.device != query.device or self.logn_tensor.dtype != query.dtype:
+                self.logn_tensor = self.logn_tensor.to(query.device).type_as(query)
+            seq_start = key.size(1) - query.size(1)
+            seq_end = key.size(1)
+            logn_tensor = self.logn_tensor[:, seq_start:seq_end, :, :]
+            query = query * logn_tensor.expand_as(query)
+
+        if (
+            self.use_flash_attn
+            and flash_attn_unpadded_func is not None
+            and not self.is_fp32
+            and query.is_cuda
+        ):
+            q, k, v = query, key, value
+            context_layer = self.core_attention_flash(q, k, v)
+
+            context_layer = rearrange(
+                context_layer, "b s h d -> b s (h d)"
+            ).contiguous()
+        else:
+            query = query.permute(0, 2, 1, 3)
+            key = key.permute(0, 2, 1, 3)
+            value = value.permute(0, 2, 1, 3)
+            attn_output, attn_weight = self._attn(
+                query, key, value, attention_mask, head_mask
+            )
+            context_layer = self._merge_heads(
+                attn_output, self.num_heads, self.head_dim
+            )
+
+        attn_output = self.c_proj(context_layer)
+        outputs = (attn_output, present)
+        if output_attentions:
+            if (
+                self.use_flash_attn
+                and flash_attn_unpadded_func is not None
+                and not self.is_fp32
+            ):
+                raise ValueError("Cannot output attentions while using flash-attn")
+            else:
+                outputs += (attn_weight,)
+
+        return outputs
+
+
+class QWenMLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.w1 = nn.Linear(
+            config.hidden_size, config.intermediate_size // 2, bias=not config.no_bias
+        )
+        self.w2 = nn.Linear(
+            config.hidden_size, config.intermediate_size // 2, bias=not config.no_bias
+        )
+        ff_dim_in = config.intermediate_size // 2
+        self.c_proj = nn.Linear(ff_dim_in, config.hidden_size, bias=not config.no_bias)
+
+    def forward(self, hidden_states):
+        a1 = self.w1(hidden_states)
+        a2 = self.w2(hidden_states)
+        intermediate_parallel = a1 * F.silu(a2)
+        output = self.c_proj(intermediate_parallel)
+        return output
+
+
+class QWenBlock(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        hidden_size = config.hidden_size
+        self.bf16 = config.bf16
+
+        self.ln_1 = RMSNorm(
+            hidden_size,
+            eps=config.layer_norm_epsilon,
+        )
+        self.attn = QWenAttention(config)
+        self.ln_2 = RMSNorm(
+            hidden_size,
+            eps=config.layer_norm_epsilon,
+        )
+
+        self.mlp = QWenMLP(config)
+
+    def forward(
+        self,
+        hidden_states: Optional[Tuple[torch.FloatTensor]],
+        layer_past: Optional[Tuple[torch.Tensor]] = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = False,
+        output_attentions: Optional[bool] = False,
+    ):
+        layernorm_output = self.ln_1(hidden_states)
+
+        attn_outputs = self.attn(
+            layernorm_output,
+            layer_past=layer_past,
+            attention_mask=attention_mask,
+            head_mask=head_mask,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+        )
+        attn_output = attn_outputs[0]
+
+        outputs = attn_outputs[1:]
+
+        residual = hidden_states
+        layernorm_input = attn_output + residual
+
+        layernorm_output = self.ln_2(layernorm_input)
+
+        residual = layernorm_input
+        mlp_output = self.mlp(layernorm_output)
+        hidden_states = residual + mlp_output
+
+        if use_cache:
+            outputs = (hidden_states,) + outputs
+        else:
+            outputs = (hidden_states,) + outputs[1:]
+
+        return outputs
+
+
+class QWenPreTrainedModel(PreTrainedModel):
+    config_class = QWenConfig
+    base_model_prefix = "transformer"
+    is_parallelizable = False
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["QWenBlock"]
+
+    def __init__(self, *inputs, **kwargs):
+        super().__init__(*inputs, **kwargs)
+
+    def _init_weights(self, module):
+        """Initialize the weights."""
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, RMSNorm):
+            module.weight.data.fill_(1.0)
+
+        for name, p in module.named_parameters():
+            if name == "c_proj.weight":
+                p.data.normal_(
+                    mean=0.0,
+                    std=(
+                        self.config.initializer_range
+                        / math.sqrt(2 * self.config.num_hidden_layers)
+                    ),
+                )
+
+    def _set_gradient_checkpointing(self, module, value=False):
+        if isinstance(module, QWenModel):
+            module.gradient_checkpointing = value
+
+
+class QWenModel(QWenPreTrainedModel):
+    _keys_to_ignore_on_load_missing = ["attn.masked_bias"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.vocab_size = config.vocab_size
+        self.num_hidden_layers = config.num_hidden_layers
+        self.embed_dim = config.hidden_size
+
+        self.gradient_checkpointing = False
+
+        self.wte = nn.Embedding(self.vocab_size, self.embed_dim)
+
+        self.drop = nn.Dropout(config.emb_dropout_prob)
+        self.h = nn.ModuleList(
+            [
+                QWenBlock(
+                    config,
+                )
+                for i in range(config.num_hidden_layers)
+            ]
+        )
+        self.ln_f = RMSNorm(
+            self.embed_dim,
+            eps=config.layer_norm_epsilon,
+        )
+
+        self.visual = VisionTransformer(**config.visual)
+
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.wte
+
+    def set_input_embeddings(self, new_embeddings):
+        self.wte = new_embeddings
+    
+    # Copied from transformers.models.bart.modeling_bart.BartDecoder._prepare_decoder_attention_mask
+    def _prepare_decoder_attention_mask(self, attention_mask, input_shape, inputs_embeds, past_key_values_length):
+        # create causal mask
+        # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
+        combined_attention_mask = None
+        if input_shape[-1] > 1:
+            combined_attention_mask = _make_causal_mask(
+                input_shape,
+                inputs_embeds.dtype,
+                device=inputs_embeds.device,
+                past_key_values_length=past_key_values_length,
+            )
+
+        if attention_mask is not None:
+            # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
+            expanded_attn_mask = _expand_mask(attention_mask, inputs_embeds.dtype, tgt_len=input_shape[-1]).to(
+                inputs_embeds.device
+            )
+            combined_attention_mask = (
+                expanded_attn_mask if combined_attention_mask is None else expanded_attn_mask + combined_attention_mask
+            )
+
+        return combined_attention_mask
+
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        token_type_ids: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ):
+        if past_key_values is None and torch.any(input_ids == self.config.visual['image_start_id']):
+            bos_pos = torch.where(input_ids == self.config.visual['image_start_id'])
+            eos_pos = torch.where(input_ids == self.config.visual['image_start_id'] + 1)
+            assert (bos_pos[0] == eos_pos[0]).all()
+            img_pos = torch.stack((bos_pos[0], bos_pos[1], eos_pos[1]), dim=1)
+            images = []
+            for i, a, b in img_pos:
+                image = input_ids[i][a + 1 : b - 1].tolist()
+                image = image[ : image.index(self.config.visual['image_start_id'] + 2)]
+                images.append(bytes(image).decode('utf-8'))
+
+            images = self.visual.encode(images)
+            assert images.shape[0] == len(images)
+        else:
+            images = None
+
+        output_attentions = (
+            output_attentions
+            if output_attentions is not None
+            else self.config.output_attentions
+        )
+        output_hidden_states = (
+            output_hidden_states
+            if output_hidden_states is not None
+            else self.config.output_hidden_states
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+        return_dict = (
+            return_dict if return_dict is not None else self.config.use_return_dict
+        )
+
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError(
+                "You cannot specify both input_ids and inputs_embeds at the same time"
+            )
+        elif input_ids is not None:
+            input_shape = input_ids.size()
+            input_ids = input_ids.view(-1, input_shape[-1])
+            batch_size = input_ids.shape[0]
+        elif inputs_embeds is not None:
+            input_shape = inputs_embeds.size()[:-1]
+            batch_size = inputs_embeds.shape[0]
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        device = input_ids.device if input_ids is not None else inputs_embeds.device
+
+        if token_type_ids is not None:
+            token_type_ids = token_type_ids.view(-1, input_shape[-1])
+        if position_ids is not None:
+            position_ids = position_ids.view(-1, input_shape[-1])
+
+        if past_key_values is None:
+            past_length = 0
+            past_key_values = tuple([None] * len(self.h))
+        else:
+            past_length = past_key_values[0][0].size(-2)
+
+        if position_ids is None:
+            position_ids = torch.arange(
+                past_length,
+                input_shape[-1] + past_length,
+                dtype=torch.long,
+                device=device,
+            )
+            position_ids = position_ids.unsqueeze(0).view(-1, input_shape[-1])
+
+        encoder_attention_mask = None
+        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
+
+        if inputs_embeds is None:
+            inputs_embeds = self.wte(input_ids)
+
+        if batch_size <= 0:
+            raise ValueError("batch_size has to be defined and > 0")
+        attention_mask = self._prepare_decoder_attention_mask(
+            attention_mask, input_shape, inputs_embeds, past_length
+        )
+
+        hidden_states = inputs_embeds
+
+        hidden_states = self.drop(hidden_states)
+        if images is not None:
+            for idx, (i, a, b) in enumerate(img_pos):
+                hidden_states[i][a + 1 : b] = images[idx]
+        output_shape = input_shape + (hidden_states.size(-1),)
+
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                )
+                use_cache = False
+
+        presents = () if use_cache else None
+        all_self_attentions = () if output_attentions else None
+        all_hidden_states = () if output_hidden_states else None
+        for i, (block, layer_past) in enumerate(zip(self.h, past_key_values)):
+
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+
+                def create_custom_forward(module):
+                    def custom_forward(*inputs):
+                        # None for past_key_value
+                        return module(*inputs, use_cache, output_attentions)
+
+                    return custom_forward
+
+                outputs = torch.utils.checkpoint.checkpoint(
+                    create_custom_forward(block),
+                    hidden_states,
+                    None,
+                    attention_mask,
+                    head_mask[i],
+                    encoder_hidden_states,
+                    encoder_attention_mask,
+                )
+            else:
+                outputs = block(
+                    hidden_states,
+                    layer_past=layer_past,
+                    attention_mask=attention_mask,
+                    head_mask=head_mask[i],
+                    encoder_hidden_states=encoder_hidden_states,
+                    encoder_attention_mask=encoder_attention_mask,
+                    use_cache=use_cache,
+                    output_attentions=output_attentions,
+                )
+
+            hidden_states = outputs[0]
+            if use_cache is True:
+                presents = presents + (outputs[2 if output_attentions else 1],)
+
+            if output_attentions:
+                all_self_attentions = all_self_attentions + (outputs[1],)
+
+        hidden_states = self.ln_f(hidden_states)
+        hidden_states = hidden_states.view(output_shape)
+        # Add last hidden state
+        if output_hidden_states:
+            all_hidden_states = all_hidden_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(
+                v for v in [hidden_states, presents, all_hidden_states] if v is not None
+            )
+
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=presents,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attentions,
+        )
+
+
+class QWenLMHeadModel(QWenPreTrainedModel):
+    _keys_to_ignore_on_load_missing = [r"h\.\d+\.attn\.rotary_emb\.inv_freq"]
+    _keys_to_ignore_on_load_unexpected = [r"h\.\d+\.attn\.masked_bias"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        assert (
+            config.bf16 + config.fp16 + config.fp32 <= 1
+        ), "Only one of \"bf16\", \"fp16\", \"fp32\" can be true"
+
+        autoset_precision = config.bf16 + config.fp16 + config.fp32 == 0
+
+        if autoset_precision:
+            if SUPPORT_BF16:
+                logger.warn(
+                    "The model is automatically converting to bf16 for faster inference. "
+                    "If you want to disable the automatic precision, please manually add bf16/fp16/fp32=True to \"AutoModelForCausalLM.from_pretrained\"."
+                )
+                config.bf16 = True
+            elif SUPPORT_FP16:
+                logger.warn(
+                    "The model is automatically converting to fp16 for faster inference. "
+                    "If you want to disable the automatic precision, please manually add bf16/fp16/fp32=True to \"AutoModelForCausalLM.from_pretrained\"."
+                )
+                config.fp16 = True
+            else:
+                config.fp32 = True
+
+        if config.bf16 and SUPPORT_CUDA and not SUPPORT_BF16:
+            logger.warn("Your device does NOT seem to support bf16, you can switch to fp16 or fp32 by by passing fp16/fp32=True in \"AutoModelForCausalLM.from_pretrained\".")
+        if config.fp16 and SUPPORT_CUDA and not SUPPORT_FP16:
+            logger.warn("Your device does NOT support faster inference with fp16, please switch to fp32 which is likely to be faster")
+        if config.fp32:
+            if SUPPORT_BF16:
+                logger.warn("Your device support faster inference by passing bf16=True in \"AutoModelForCausalLM.from_pretrained\".")
+            elif SUPPORT_FP16:
+                logger.warn("Your device support faster inference by passing fp16=True in \"AutoModelForCausalLM.from_pretrained\".")
+        
+        if config.use_flash_attn == "auto":
+            if config.bf16 or config.fp16:
+                logger.warn("Try importing flash-attention for faster inference...")
+                config.use_flash_attn = True
+            else:
+                config.use_flash_attn = False
+        if config.use_flash_attn and config.fp32:
+            logger.warn("Flash attention will be disabled because it does NOT support fp32.")
+
+        if config.use_flash_attn:
+            _import_flash_attn()
+
+        self.transformer = QWenModel(config)
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        if config.bf16:
+            self.transformer.bfloat16()
+            self.lm_head.bfloat16()
+        if config.fp16:
+            self.transformer.half()
+            self.lm_head.half()
+        self.post_init()
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    def prepare_inputs_for_generation(
+        self, input_ids, past_key_values=None, inputs_embeds=None, **kwargs
+    ):
+        token_type_ids = kwargs.get("token_type_ids", None)
+        if past_key_values:
+            input_ids = input_ids[:, -1].unsqueeze(-1)
+            if token_type_ids is not None:
+                token_type_ids = token_type_ids[:, -1].unsqueeze(-1)
+
+        attention_mask = kwargs.get("attention_mask", None)
+        position_ids = kwargs.get("position_ids", None)
+
+        if attention_mask is not None and position_ids is None:
+            position_ids = attention_mask.long().cumsum(-1) - 1
+            position_ids.masked_fill_(attention_mask == 0, 1)
+            if past_key_values:
+                position_ids = position_ids[:, -1].unsqueeze(-1)
+        else:
+            position_ids = None
+
+        if inputs_embeds is not None and past_key_values is None:
+            model_inputs = {"inputs_embeds": inputs_embeds}
+        else:
+            model_inputs = {"input_ids": input_ids}
+
+        model_inputs.update(
+            {
+                "past_key_values": past_key_values,
+                "use_cache": kwargs.get("use_cache"),
+                "position_ids": position_ids,
+                "attention_mask": attention_mask,
+                "token_type_ids": token_type_ids,
+            }
+        )
+        return model_inputs
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        token_type_ids: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, CausalLMOutputWithPast]:
+
+        return_dict = (
+            return_dict if return_dict is not None else self.config.use_return_dict
+        )
+
+        transformer_outputs = self.transformer(
+            input_ids,
+            past_key_values=past_key_values,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        hidden_states = transformer_outputs[0]
+
+        lm_logits = self.lm_head(hidden_states)
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(lm_logits.device)
+            shift_logits = lm_logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(
+                shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1)
+            )
+
+        if not return_dict:
+            output = (lm_logits,) + transformer_outputs[1:]
+            return ((loss,) + output) if loss is not None else output
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=lm_logits,
+            past_key_values=transformer_outputs.past_key_values,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )
+
+    @staticmethod
+    def _reorder_cache(
+        past_key_values: Tuple[Tuple[torch.Tensor]], beam_idx: torch.Tensor
+    ) -> Tuple[Tuple[torch.Tensor]]:
+
+        return tuple(
+            tuple(
+                past_state.index_select(0, beam_idx.to(past_state.device))
+                for past_state in layer_past
+            )
+            for layer_past in past_key_values
+        )
+
+    def chat(
+        self,
+        tokenizer: PreTrainedTokenizer,
+        query: str,
+        history: Optional[HistoryType],
+        system: str = "You are a helpful assistant.",
+        append_history: bool = True,
+        stream: Optional[bool] = _SENTINEL,
+        stop_words_ids: Optional[List[List[int]]] = None,
+        **kwargs,
+    ) -> Tuple[str, HistoryType]:
+        assert stream is _SENTINEL, _ERROR_STREAM_IN_CHAT
+        assert self.generation_config.chat_format == 'chatml', _ERROR_BAD_CHAT_FORMAT
+        if history is None:
+            history = []
+        if stop_words_ids is None:
+            stop_words_ids = []
+
+        max_window_size = kwargs.get('max_window_size', None)
+        if max_window_size is None:
+            max_window_size = self.generation_config.max_window_size
+        raw_text, context_tokens = make_context(
+            tokenizer,
+            query,
+            history=history,
+            system=system,
+            max_window_size=max_window_size,
+            chat_format=self.generation_config.chat_format,
+        )
+
+        stop_words_ids.extend(get_stop_words_ids(
+            self.generation_config.chat_format, tokenizer
+        ))
+        input_ids = torch.tensor([context_tokens]).to(self.device)
+        outputs = self.generate(
+                    input_ids,
+                    stop_words_ids = stop_words_ids,
+                    return_dict_in_generate = False,
+                    **kwargs,
+                )
+
+        response = decode_tokens(
+            outputs[0],
+            tokenizer,
+            raw_text_len=len(raw_text),
+            context_length=len(context_tokens),
+            chat_format=self.generation_config.chat_format,
+            verbose=False,
+            errors='replace'
+        )
+
+        if append_history:
+            history.append((query, response))
+
+        return response, history
+
+    def chat_stream(
+            self,
+            tokenizer: PreTrainedTokenizer,
+            query: str,
+            history: Optional[HistoryType],
+            system: str = "You are a helpful assistant.",
+            stop_words_ids: Optional[List[List[int]]] = None,
+            logits_processor: Optional[LogitsProcessorList] = None,
+            **kwargs,
+    ) -> Generator[str, Any, None]:
+        assert self.generation_config.chat_format == 'chatml', _ERROR_BAD_CHAT_FORMAT
+        if history is None:
+            history = []
+        if stop_words_ids is None:
+            stop_words_ids = []
+
+        max_window_size = kwargs.get('max_window_size', None)
+        if max_window_size is None:
+            max_window_size = self.generation_config.max_window_size
+        raw_text, context_tokens = make_context(
+            tokenizer,
+            query,
+            history=history,
+            system=system,
+            max_window_size=max_window_size,
+            chat_format=self.generation_config.chat_format,
+        )
+
+        stop_words_ids.extend(get_stop_words_ids(
+            self.generation_config.chat_format, tokenizer
+        ))
+        if stop_words_ids is not None:
+            stop_words_logits_processor = StopWordsLogitsProcessor(
+                stop_words_ids=stop_words_ids,
+                eos_token_id=self.generation_config.eos_token_id,
+            )
+            if logits_processor is None:
+                logits_processor = LogitsProcessorList([stop_words_logits_processor])
+            else:
+                logits_processor.append(stop_words_logits_processor)
+        input_ids = torch.tensor([context_tokens]).to(self.device)
+
+        from transformers_stream_generator.main import NewGenerationMixin, StreamGenerationConfig
+        self.__class__.generate_stream = NewGenerationMixin.generate
+        self.__class__.sample_stream = NewGenerationMixin.sample_stream
+        stream_config = StreamGenerationConfig(**self.generation_config.to_dict(), do_stream=True)
+        def stream_generator():
+            outputs = []
+            for token in self.generate_stream(
+                    input_ids,
+                    return_dict_in_generate=False,
+                    generation_config=stream_config,
+                    logits_processor=logits_processor,
+                    seed=-1,
+                    **kwargs):
+                outputs.append(token.item())
+                yield tokenizer.decode(outputs, skip_special_tokens=True, errors='ignore')
+
+        return stream_generator()
+
+    def generate(
+        self,
+        inputs: Optional[torch.Tensor] = None,
+        generation_config: Optional[GenerationConfig] = None,
+        logits_processor: Optional[LogitsProcessorList] = None,
+        stopping_criteria: Optional[StoppingCriteriaList] = None,
+        prefix_allowed_tokens_fn: Optional[
+            Callable[[int, torch.Tensor], List[int]]
+        ] = None,
+        synced_gpus: Optional[bool] = None,
+        assistant_model: Optional["PreTrainedModel"] = None,
+        streamer: Optional["BaseStreamer"] = None,
+        **kwargs,
+    ) -> Union[GenerateOutput, torch.LongTensor]:
+        # Process stop_words_ids.
+        stop_words_ids = kwargs.pop("stop_words_ids", None)
+        if stop_words_ids is None and generation_config is not None:
+            stop_words_ids = getattr(generation_config, "stop_words_ids", None)
+        if stop_words_ids is None:
+            stop_words_ids = getattr(self.generation_config, "stop_words_ids", None)
+
+        if stop_words_ids is not None:
+            stop_words_logits_processor = StopWordsLogitsProcessor(
+                stop_words_ids=stop_words_ids,
+                eos_token_id=self.generation_config.eos_token_id,
+            )
+            if logits_processor is None:
+                logits_processor = LogitsProcessorList([stop_words_logits_processor])
+            else:
+                logits_processor.append(stop_words_logits_processor)
+
+        return super().generate(
+            inputs,
+            generation_config=generation_config,
+            logits_processor=logits_processor,
+            stopping_criteria=stopping_criteria,
+            prefix_allowed_tokens_fn=prefix_allowed_tokens_fn,
+            synced_gpus=synced_gpus,
+            assistant_model=assistant_model,
+            streamer=streamer,
+            **kwargs,
+        )
+
+
+class RotaryEmbedding(torch.nn.Module):
+    def __init__(self, dim, base=10000):
+        super().__init__()
+        self.dim = dim
+        self.base = base
+        self.inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float() / dim))
+        if importlib.util.find_spec("einops") is None:
+            raise RuntimeError("einops is required for Rotary Embedding")
+
+        self._rotary_pos_emb_cache = None
+        self._seq_len_cached = 0
+        self._ntk_alpha_cached = 1.0
+
+    def update_rotary_pos_emb_cache(self, max_seq_len, offset=0, ntk_alpha=1.0):
+        seqlen = max_seq_len + offset
+        if seqlen > self._seq_len_cached or ntk_alpha != self._ntk_alpha_cached:
+            base = self.base * ntk_alpha ** (self.dim / (self.dim - 2))
+            self.inv_freq = 1.0 / (
+                base
+                ** (
+                    torch.arange(0, self.dim, 2, device=self.inv_freq.device).float()
+                    / self.dim
+                )
+            )
+            self._seq_len_cached = max(2 * seqlen, 16)
+            self._ntk_alpha_cached = ntk_alpha
+            seq = torch.arange(self._seq_len_cached, device=self.inv_freq.device)
+            freqs = torch.outer(seq.type_as(self.inv_freq), self.inv_freq)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            from einops import rearrange
+
+            self._rotary_pos_emb_cache = rearrange(emb, "n d -> 1 n 1 d")
+
+    def forward(self, max_seq_len, offset=0, ntk_alpha=1.0):
+        self.update_rotary_pos_emb_cache(max_seq_len, offset, ntk_alpha)
+        return self._rotary_pos_emb_cache[:, offset : offset + max_seq_len]
+
+
+def _rotate_half(x):
+    from einops import rearrange
+
+    x = rearrange(x, "... (j d) -> ... j d", j=2)
+    x1, x2 = x.unbind(dim=-2)
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(t, freqs):
+    if apply_rotary_emb_func is not None and t.is_cuda:
+        t_ = t.float()
+        freqs = freqs.squeeze(0).squeeze(1)
+        cos = freqs[:, : freqs.shape[-1] // 2].cos()
+        sin = freqs[:, : freqs.shape[-1] // 2].sin()
+        output = apply_rotary_emb_func(t_, cos, sin).type_as(t)
+        return output
+    else:
+        rot_dim = freqs.shape[-1]
+        t_, t_pass_ = t[..., :rot_dim], t[..., rot_dim:]
+        t_ = t_.float()
+        t_pass_ = t_pass_.float()
+        t_ = (t_ * freqs.cos()) + (_rotate_half(t_) * freqs.sin())
+        return torch.cat((t_, t_pass_), dim=-1).type_as(t)
+
+
+class RMSNorm(torch.nn.Module):
+    def __init__(self, dim: int, eps: float = 1e-6):
+        super().__init__()
+        self.eps = eps
+        self.weight = nn.Parameter(torch.ones(dim))
+
+    def _norm(self, x):
+        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)
+
+    def forward(self, x):
+        if rms_norm is not None and x.is_cuda:
+            return rms_norm(x, self.weight, self.eps)
+        else:
+            output = self._norm(x.float()).type_as(x)
+            return output * self.weight

pytorch_model-00001-of-00010.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f85037d904705766b58b76c5265a568591f288925a081488a0ec3f3f752c7fce
+size 1964070447

pytorch_model-00002-of-00010.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe54b166888161c8598615eec862e6ad2c86b3d3eaac501daa6f86d3b0174850
+size 1933792141

pytorch_model-00003-of-00010.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ce49783fa46b607f112df9d27e7be2f313041cd3f8414226f6d8bdbeae86f4c
+size 1933792141

pytorch_model-00004-of-00010.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e64e1b5d87d41e2da7dd6b42ad5dd92feac963d01ab5481e86b4da198a670264
+size 1990406779

pytorch_model-00005-of-00010.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d1b139edfde201c5ff4b592b968bc2baf00cba4b42ecc5e858a1fc898d3c434
+size 1923281531

pytorch_model-00006-of-00010.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:206f37b9ee4747f53a5c66c1a388c7bcb8236315bec83e9ca105b350acae0781
+size 1933783675

pytorch_model-00007-of-00010.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cf880e42cb11a45985f8b1e2caa0ba9afa98ab13ce38be5150641bb4cbb85fd
+size 1933792205

pytorch_model-00008-of-00010.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f8c5659e3d3934ccc92519039168ca84c945dec227f8477fbcafd2426264156
+size 1975364669

pytorch_model-00009-of-00010.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02b652b95c6dffc2754d2cc3d8166b35c542726a571b979fc2b59417c218f196
+size 1994925011

pytorch_model-00010-of-00010.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:508bc4b5caab78e08d9961000cd00411f0e56f0f016828bcf9e78c132d960c24
+size 1730968463

pytorch_model.bin.index.json

@@ -0,0 +1,860 @@
+{
+  "metadata": {
+    "total_size": 19313870336
+  },
+  "weight_map": {
+    "lm_head.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.h.0.attn.c_attn.bias": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.0.attn.c_attn.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.0.attn.c_proj.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.0.ln_1.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.0.ln_2.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.0.mlp.c_proj.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.0.mlp.w1.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.0.mlp.w2.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.1.attn.c_attn.bias": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.1.attn.c_attn.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.1.attn.c_proj.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.1.ln_1.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.1.ln_2.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.1.mlp.c_proj.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.1.mlp.w1.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.1.mlp.w2.weight": "pytorch_model-00001-of-00010.bin",
+    "transformer.h.10.attn.c_attn.bias": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.10.attn.c_attn.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.10.attn.c_proj.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.10.ln_1.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.10.ln_2.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.10.mlp.c_proj.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.10.mlp.w1.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.10.mlp.w2.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.11.attn.c_attn.bias": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.11.attn.c_attn.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.11.attn.c_proj.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.11.ln_1.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.11.ln_2.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.11.mlp.c_proj.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.11.mlp.w1.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.11.mlp.w2.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.12.attn.c_attn.bias": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.12.attn.c_attn.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.12.attn.c_proj.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.12.ln_1.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.12.ln_2.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.12.mlp.c_proj.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.12.mlp.w1.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.12.mlp.w2.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.13.attn.c_attn.bias": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.13.attn.c_attn.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.13.attn.c_proj.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.13.ln_1.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.13.ln_2.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.13.mlp.c_proj.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.13.mlp.w1.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.13.mlp.w2.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.14.attn.c_attn.bias": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.14.attn.c_attn.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.14.attn.c_proj.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.14.ln_1.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.14.ln_2.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.14.mlp.c_proj.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.14.mlp.w1.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.14.mlp.w2.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.15.attn.c_attn.bias": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.15.attn.c_attn.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.15.attn.c_proj.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.15.ln_1.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.15.ln_2.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.15.mlp.c_proj.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.15.mlp.w1.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.15.mlp.w2.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.16.attn.c_attn.bias": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.16.attn.c_attn.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.16.attn.c_proj.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.16.ln_1.weight": "pytorch_model-00004-of-00010.bin",
+    "transformer.h.16.ln_2.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.16.mlp.c_proj.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.16.mlp.w1.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.16.mlp.w2.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.17.attn.c_attn.bias": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.17.attn.c_attn.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.17.attn.c_proj.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.17.ln_1.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.17.ln_2.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.17.mlp.c_proj.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.17.mlp.w1.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.17.mlp.w2.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.18.attn.c_attn.bias": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.18.attn.c_attn.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.18.attn.c_proj.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.18.ln_1.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.18.ln_2.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.18.mlp.c_proj.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.18.mlp.w1.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.18.mlp.w2.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.19.attn.c_attn.bias": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.19.attn.c_attn.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.19.attn.c_proj.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.19.ln_1.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.19.ln_2.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.19.mlp.c_proj.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.19.mlp.w1.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.19.mlp.w2.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.2.attn.c_attn.bias": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.2.attn.c_attn.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.2.attn.c_proj.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.2.ln_1.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.2.ln_2.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.2.mlp.c_proj.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.2.mlp.w1.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.2.mlp.w2.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.20.attn.c_attn.bias": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.20.attn.c_attn.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.20.attn.c_proj.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.20.ln_1.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.20.ln_2.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.20.mlp.c_proj.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.20.mlp.w1.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.20.mlp.w2.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.21.attn.c_attn.bias": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.21.attn.c_attn.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.21.attn.c_proj.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.21.ln_1.weight": "pytorch_model-00005-of-00010.bin",
+    "transformer.h.21.ln_2.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.21.mlp.c_proj.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.21.mlp.w1.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.21.mlp.w2.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.22.attn.c_attn.bias": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.22.attn.c_attn.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.22.attn.c_proj.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.22.ln_1.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.22.ln_2.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.22.mlp.c_proj.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.22.mlp.w1.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.22.mlp.w2.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.23.attn.c_attn.bias": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.23.attn.c_attn.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.23.attn.c_proj.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.23.ln_1.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.23.ln_2.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.23.mlp.c_proj.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.23.mlp.w1.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.23.mlp.w2.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.24.attn.c_attn.bias": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.24.attn.c_attn.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.24.attn.c_proj.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.24.ln_1.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.24.ln_2.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.24.mlp.c_proj.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.24.mlp.w1.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.24.mlp.w2.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.25.attn.c_attn.bias": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.25.attn.c_attn.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.25.attn.c_proj.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.25.ln_1.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.25.ln_2.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.25.mlp.c_proj.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.25.mlp.w1.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.25.mlp.w2.weight": "pytorch_model-00006-of-00010.bin",
+    "transformer.h.26.attn.c_attn.bias": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.26.attn.c_attn.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.26.attn.c_proj.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.26.ln_1.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.26.ln_2.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.26.mlp.c_proj.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.26.mlp.w1.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.26.mlp.w2.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.27.attn.c_attn.bias": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.27.attn.c_attn.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.27.attn.c_proj.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.27.ln_1.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.27.ln_2.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.27.mlp.c_proj.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.27.mlp.w1.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.27.mlp.w2.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.28.attn.c_attn.bias": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.28.attn.c_attn.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.28.attn.c_proj.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.28.ln_1.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.28.ln_2.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.28.mlp.c_proj.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.28.mlp.w1.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.28.mlp.w2.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.29.attn.c_attn.bias": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.29.attn.c_attn.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.29.attn.c_proj.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.29.ln_1.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.29.ln_2.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.29.mlp.c_proj.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.29.mlp.w1.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.29.mlp.w2.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.3.attn.c_attn.bias": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.3.attn.c_attn.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.3.attn.c_proj.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.3.ln_1.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.3.ln_2.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.3.mlp.c_proj.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.3.mlp.w1.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.3.mlp.w2.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.30.attn.c_attn.bias": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.30.attn.c_attn.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.30.attn.c_proj.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.30.ln_1.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.30.ln_2.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.30.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.h.30.mlp.w1.weight": "pytorch_model-00007-of-00010.bin",
+    "transformer.h.30.mlp.w2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.h.31.attn.c_attn.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.h.31.attn.c_attn.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.h.31.attn.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.h.31.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.h.31.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.h.31.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.h.31.mlp.w1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.h.31.mlp.w2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.h.4.attn.c_attn.bias": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.4.attn.c_attn.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.4.attn.c_proj.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.4.ln_1.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.4.ln_2.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.4.mlp.c_proj.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.4.mlp.w1.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.4.mlp.w2.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.5.attn.c_attn.bias": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.5.attn.c_attn.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.5.attn.c_proj.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.5.ln_1.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.5.ln_2.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.5.mlp.c_proj.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.5.mlp.w1.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.5.mlp.w2.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.6.attn.c_attn.bias": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.6.attn.c_attn.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.6.attn.c_proj.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.6.ln_1.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.6.ln_2.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.6.mlp.c_proj.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.6.mlp.w1.weight": "pytorch_model-00002-of-00010.bin",
+    "transformer.h.6.mlp.w2.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.7.attn.c_attn.bias": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.7.attn.c_attn.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.7.attn.c_proj.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.7.ln_1.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.7.ln_2.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.7.mlp.c_proj.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.7.mlp.w1.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.7.mlp.w2.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.8.attn.c_attn.bias": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.8.attn.c_attn.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.8.attn.c_proj.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.8.ln_1.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.8.ln_2.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.8.mlp.c_proj.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.8.mlp.w1.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.8.mlp.w2.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.9.attn.c_attn.bias": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.9.attn.c_attn.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.9.attn.c_proj.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.9.ln_1.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.9.ln_2.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.9.mlp.c_proj.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.9.mlp.w1.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.h.9.mlp.w2.weight": "pytorch_model-00003-of-00010.bin",
+    "transformer.ln_f.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.attn_pool.attn.in_proj_bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.attn_pool.attn.in_proj_weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.attn_pool.attn.out_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.attn_pool.attn.out_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.attn_pool.kv_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.attn_pool.ln_kv.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.attn_pool.ln_kv.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.attn_pool.ln_q.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.attn_pool.ln_q.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.attn_pool.pos_embed": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.attn_pool.query": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.conv1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.ln_post.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.ln_post.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.ln_pre.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.ln_pre.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.positional_embedding": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.proj": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.0.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.1.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.10.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.11.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.12.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.13.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.14.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.15.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.16.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.17.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.18.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.19.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.2.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.20.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.21.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.22.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.23.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.24.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.25.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.26.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.27.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.28.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.29.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.3.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.30.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.31.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.32.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.33.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.34.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.35.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.36.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.37.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.38.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.39.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.4.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.40.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.41.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.mlp.c_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.42.mlp.c_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.attn.in_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.attn.in_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.attn.out_proj.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.attn.out_proj.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.ln_1.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.ln_1.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.ln_2.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.ln_2.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.mlp.c_fc.bias": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.mlp.c_fc.weight": "pytorch_model-00009-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.mlp.c_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.43.mlp.c_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.attn.in_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.attn.in_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.attn.out_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.attn.out_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.ln_1.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.ln_1.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.ln_2.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.ln_2.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.mlp.c_fc.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.mlp.c_fc.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.mlp.c_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.44.mlp.c_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.attn.in_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.attn.in_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.attn.out_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.attn.out_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.ln_1.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.ln_1.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.ln_2.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.ln_2.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.mlp.c_fc.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.mlp.c_fc.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.mlp.c_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.45.mlp.c_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.attn.in_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.attn.in_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.attn.out_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.attn.out_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.ln_1.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.ln_1.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.ln_2.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.ln_2.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.mlp.c_fc.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.mlp.c_fc.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.mlp.c_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.46.mlp.c_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.attn.in_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.attn.in_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.attn.out_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.attn.out_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.ln_1.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.ln_1.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.ln_2.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.ln_2.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.mlp.c_fc.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.mlp.c_fc.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.mlp.c_proj.bias": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.47.mlp.c_proj.weight": "pytorch_model-00010-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.5.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.6.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.7.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.8.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.attn.in_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.attn.in_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.attn.out_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.attn.out_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.ln_1.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.ln_1.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.ln_2.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.ln_2.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.mlp.c_fc.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.mlp.c_fc.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.mlp.c_proj.bias": "pytorch_model-00008-of-00010.bin",
+    "transformer.visual.transformer.resblocks.9.mlp.c_proj.weight": "pytorch_model-00008-of-00010.bin",
+    "transformer.wte.weight": "pytorch_model-00001-of-00010.bin"
+  }
+}

qwen_generation_utils.py

@@ -0,0 +1,420 @@
+# Copyright (c) Alibaba Cloud.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+"""Generation support."""
+
+from typing import Tuple, List, Union, Iterable
+
+import numpy as np
+import torch
+import torch.nn.functional as F
+from transformers import PreTrainedTokenizer
+from transformers import logging
+from transformers.generation import LogitsProcessor
+
+logger = logging.get_logger(__name__)
+
+# Types.
+HistoryType = List[Tuple[str, str]]
+TokensType = List[int]
+BatchTokensType = List[List[int]]
+
+
+def pad_batch(batch: BatchTokensType, pad_id: int, seq_length: int) -> BatchTokensType:
+    for tokens in batch:
+        context_length = len(tokens)
+        if context_length < seq_length:
+            tokens.extend([pad_id] * (seq_length - context_length))
+    return batch
+
+
+def get_ltor_masks_and_position_ids(
+    data,
+    eod_token,
+    reset_position_ids,
+    reset_attention_mask,
+    eod_mask_loss,
+):
+    """Build masks and position id for left to right model."""
+
+    # Extract batch size and sequence length.
+    micro_batch_size, seq_length = data.size()
+
+    # Attention mask (lower triangular).
+    if reset_attention_mask:
+        att_mask_batch = micro_batch_size
+    else:
+        att_mask_batch = 1
+    attention_mask = torch.tril(
+        torch.ones((att_mask_batch, seq_length, seq_length), device=data.device)
+    ).view(att_mask_batch, 1, seq_length, seq_length)
+
+    # Loss mask.
+    loss_mask = torch.ones(data.size(), dtype=torch.float, device=data.device)
+    if eod_mask_loss:
+        loss_mask[data == eod_token] = 0.0
+
+    # Position ids.
+    position_ids = torch.arange(seq_length, dtype=torch.long, device=data.device)
+    position_ids = position_ids.unsqueeze(0).expand_as(data)
+    # We need to clone as the ids will be modifed based on batch index.
+    if reset_position_ids:
+        position_ids = position_ids.clone()
+
+    if reset_position_ids or reset_attention_mask:
+        # Loop through the batches:
+        for b in range(micro_batch_size):
+
+            # Find indecies where EOD token is.
+            eod_index = position_ids[b, data[b] == eod_token]
+            # Detach indecies from positions if going to modify positions.
+            if reset_position_ids:
+                eod_index = eod_index.clone()
+
+            # Loop through EOD indecies:
+            prev_index = 0
+            for j in range(eod_index.size()[0]):
+                i = eod_index[j]
+                # Mask attention loss.
+                if reset_attention_mask:
+                    attention_mask[b, 0, (i + 1) :, : (i + 1)] = 0
+                # Reset positions.
+                if reset_position_ids:
+                    position_ids[b, (i + 1) :] -= i + 1 - prev_index
+                    prev_index = i + 1
+
+    # Convert attention mask to binary:
+    attention_mask = attention_mask < 0.5
+
+    return attention_mask, loss_mask, position_ids
+
+
+def get_batch(context_tokens: torch.LongTensor, eod_id: int):
+    """Generate batch from context tokens."""
+    # Move to GPU.
+    tokens = context_tokens.contiguous().to(context_tokens.device)
+    # Get the attention mask and postition ids.
+    attention_mask, _, position_ids = get_ltor_masks_and_position_ids(
+        tokens,
+        eod_id,
+        reset_position_ids=False,
+        reset_attention_mask=False,
+        eod_mask_loss=False,
+    )
+    return tokens, attention_mask, position_ids
+
+
+def get_stop_words_ids(chat_format, tokenizer):
+    if chat_format == "raw":
+        stop_words_ids = [tokenizer.encode("Human:"), [tokenizer.eod_id]]
+    elif chat_format == "chatml":
+        stop_words_ids = [[tokenizer.im_end_id], [tokenizer.im_start_id]]
+    else:
+        raise NotImplementedError(f"Unknown chat format {chat_format!r}")
+    return stop_words_ids
+
+
+def make_context(
+    tokenizer: PreTrainedTokenizer,
+    query: str,
+    history: List[Tuple[str, str]] = None,
+    system: str = "",
+    max_window_size: int = 6144,
+    chat_format: str = "chatml",
+):
+    if history is None:
+        history = []
+
+    if chat_format == "chatml":
+        im_start, im_end = "<|im_start|>", "<|im_end|>"
+        im_start_tokens = [tokenizer.im_start_id]
+        im_end_tokens = [tokenizer.im_end_id]
+        nl_tokens = tokenizer.encode("\n")
+
+        def _tokenize_str(role, content):
+            return f"{role}\n{content}", tokenizer.encode(
+                role, allowed_special=set(tokenizer.IMAGE_ST)
+            ) + nl_tokens + tokenizer.encode(content, allowed_special=set(tokenizer.IMAGE_ST))
+
+        system_text, system_tokens_part = _tokenize_str("system", system)
+        system_tokens = im_start_tokens + system_tokens_part + im_end_tokens
+
+        raw_text = ""
+        context_tokens = []
+
+        for turn_query, turn_response in reversed(history):
+            query_text, query_tokens_part = _tokenize_str("user", turn_query)
+            query_tokens = im_start_tokens + query_tokens_part + im_end_tokens
+            if turn_response is not None:
+                response_text, response_tokens_part = _tokenize_str(
+                    "assistant", turn_response
+                )
+                response_tokens = im_start_tokens + response_tokens_part + im_end_tokens
+
+                next_context_tokens = nl_tokens + query_tokens + nl_tokens + response_tokens
+                prev_chat = (
+                    f"\n{im_start}{query_text}{im_end}\n{im_start}{response_text}{im_end}"
+                )
+            else:
+                next_context_tokens = nl_tokens + query_tokens + nl_tokens
+                prev_chat = f"\n{im_start}{query_text}{im_end}\n"
+
+            current_context_size = (
+                len(system_tokens) + len(next_context_tokens) + len(context_tokens)
+            )
+            if current_context_size < max_window_size:
+                context_tokens = next_context_tokens + context_tokens
+                raw_text = prev_chat + raw_text
+            else:
+                break
+
+        context_tokens = system_tokens + context_tokens
+        raw_text = f"{im_start}{system_text}{im_end}" + raw_text
+        context_tokens += (
+            nl_tokens
+            + im_start_tokens
+            + _tokenize_str("user", query)[1]
+            + im_end_tokens
+            + nl_tokens
+            + im_start_tokens
+            + tokenizer.encode("assistant")
+            + nl_tokens
+        )
+        raw_text += f"\n{im_start}user\n{query}{im_end}\n{im_start}assistant\n"
+
+    elif chat_format == "raw":
+        raw_text = query
+        context_tokens = tokenizer.encode(raw_text)
+    else:
+        raise NotImplementedError(f"Unknown chat format {chat_format!r}")
+
+    return raw_text, context_tokens
+
+
+def _decode_default(
+    tokens: List[int],
+    *,
+    stop_words: List[str],
+    eod_words: List[str],
+    tokenizer: PreTrainedTokenizer,
+    raw_text_len: int,
+    verbose: bool = False,
+    return_end_reason: bool = False,
+    errors: str='replace',
+):
+    trim_decode_tokens = tokenizer.decode(tokens, errors=errors)[raw_text_len:]
+    if verbose:
+        print("\nRaw Generate: ", trim_decode_tokens)
+
+    end_reason = f"Gen length {len(tokens)}"
+    for stop_word in stop_words:
+        trim_decode_tokens = trim_decode_tokens.replace(stop_word, "").strip()
+    for eod_word in eod_words:
+        if eod_word in trim_decode_tokens:
+            end_reason = f"Gen {eod_word!r}"
+        trim_decode_tokens = trim_decode_tokens.split(eod_word)[0]
+    trim_decode_tokens = trim_decode_tokens.strip()
+    if verbose:
+        print("\nEnd Reason:", end_reason)
+        print("\nGenerate: ", trim_decode_tokens)
+
+    if return_end_reason:
+        return trim_decode_tokens, end_reason
+    else:
+        return trim_decode_tokens
+
+
+def _decode_chatml(
+    tokens: List[int],
+    *,
+    stop_words: List[str],
+    eod_token_ids: List[int],
+    tokenizer: PreTrainedTokenizer,
+    raw_text_len: int,
+    context_length: int,
+    verbose: bool = False,
+    return_end_reason: bool = False,
+    errors: str='replace'
+):
+    end_reason = f"Gen length {len(tokens)}"
+    eod_token_idx = context_length
+    for eod_token_idx in range(context_length, len(tokens)):
+        if tokens[eod_token_idx] in eod_token_ids:
+            end_reason = f"Gen {tokenizer.decode([tokens[eod_token_idx]])!r}"
+            break
+
+    trim_decode_tokens = tokenizer.decode(tokens[:eod_token_idx], errors=errors)[raw_text_len:]
+    if verbose:
+        print("\nRaw Generate w/o EOD:", tokenizer.decode(tokens, errors=errors)[raw_text_len:])
+        print("\nRaw Generate:", trim_decode_tokens)
+        print("\nEnd Reason:", end_reason)
+    for stop_word in stop_words:
+        trim_decode_tokens = trim_decode_tokens.replace(stop_word, "").strip()
+    trim_decode_tokens = trim_decode_tokens.strip()
+    if verbose:
+        print("\nGenerate:", trim_decode_tokens)
+
+    if return_end_reason:
+        return trim_decode_tokens, end_reason
+    else:
+        return trim_decode_tokens
+
+
+def decode_tokens(
+    tokens: Union[torch.LongTensor, TokensType],
+    tokenizer: PreTrainedTokenizer,
+    raw_text_len: int,
+    context_length: int,
+    chat_format: str,
+    verbose: bool = False,
+    return_end_reason: bool = False,
+    errors: str="replace",
+) -> str:
+    if torch.is_tensor(tokens):
+        tokens = tokens.cpu().numpy().tolist()
+
+    if chat_format == "chatml":
+        return _decode_chatml(
+            tokens,
+            stop_words=[],
+            eod_token_ids=[tokenizer.im_start_id, tokenizer.im_end_id],
+            tokenizer=tokenizer,
+            raw_text_len=raw_text_len,
+            context_length=context_length,
+            verbose=verbose,
+            return_end_reason=return_end_reason,
+            errors=errors,
+        )
+    elif chat_format == "raw":
+        return _decode_default(
+            tokens,
+            stop_words=["<|endoftext|>"],
+            eod_words=["<|endoftext|>"],
+            tokenizer=tokenizer,
+            raw_text_len=raw_text_len,
+            verbose=verbose,
+            return_end_reason=return_end_reason,
+            errors=errors,
+        )
+    else:
+        raise NotImplementedError(f"Unknown chat format {chat_format!r}")
+
+
+class StopWordsLogitsProcessor(LogitsProcessor):
+    """
+    :class:`transformers.LogitsProcessor` that enforces that when specified sequences appear, stop geration.
+
+    Args:
+        stop_words_ids (:obj:`List[List[int]]`):
+            List of list of token ids of stop ids. In order to get the tokens of the words
+            that should not appear in the generated text, use :obj:`tokenizer(bad_word,
+            add_prefix_space=True).input_ids`.
+        eos_token_id (:obj:`int`):
+            The id of the `end-of-sequence` token.
+    """
+
+    def __init__(self, stop_words_ids: Iterable[Iterable[int]], eos_token_id: int):
+
+        if not isinstance(stop_words_ids, List) or len(stop_words_ids) == 0:
+            raise ValueError(
+                f"`stop_words_ids` has to be a non-emtpy list, but is {stop_words_ids}."
+            )
+        if any(not isinstance(bad_word_ids, list) for bad_word_ids in stop_words_ids):
+            raise ValueError(
+                f"`stop_words_ids` has to be a list of lists, but is {stop_words_ids}."
+            )
+        if any(
+            any(
+                (not isinstance(token_id, (int, np.integer)) or token_id < 0)
+                for token_id in stop_word_ids
+            )
+            for stop_word_ids in stop_words_ids
+        ):
+            raise ValueError(
+                f"Each list in `stop_words_ids` has to be a list of positive integers, but is {stop_words_ids}."
+            )
+
+        self.stop_words_ids = list(
+            filter(
+                lambda bad_token_seq: bad_token_seq != [eos_token_id], stop_words_ids
+            )
+        )
+        self.eos_token_id = eos_token_id
+        for stop_token_seq in self.stop_words_ids:
+            assert (
+                len(stop_token_seq) > 0
+            ), "Stop words token sequences {} cannot have an empty list".format(
+                stop_words_ids
+            )
+
+    def __call__(
+        self, input_ids: torch.LongTensor, scores: torch.FloatTensor
+    ) -> torch.FloatTensor:
+        stopped_samples = self._calc_stopped_samples(input_ids)
+        for i, should_stop in enumerate(stopped_samples):
+            if should_stop:
+                scores[i, self.eos_token_id] = float(2**15)
+        return scores
+
+    def _tokens_match(self, prev_tokens: torch.LongTensor, tokens: List[int]) -> bool:
+        if len(tokens) == 0:
+            # if bad word tokens is just one token always ban it
+            return True
+        elif len(tokens) > len(prev_tokens):
+            # if bad word tokens are longer then prev input_ids they can't be equal
+            return False
+        elif prev_tokens[-len(tokens) :].tolist() == tokens:
+            # if tokens match
+            return True
+        else:
+            return False
+
+    def _calc_stopped_samples(self, prev_input_ids: Iterable[int]) -> Iterable[int]:
+        stopped_samples = []
+        for prev_input_ids_slice in prev_input_ids:
+            match = False
+            for stop_token_seq in self.stop_words_ids:
+                if self._tokens_match(prev_input_ids_slice, stop_token_seq):
+                    # if tokens do not match continue
+                    match = True
+                    break
+            stopped_samples.append(match)
+
+        return stopped_samples
+
+
+def top_k_logits(logits, top_k=0, top_p=0.0, filter_value=-float("Inf")):
+    """This function has been mostly taken from huggingface conversational
+    ai code at
+        https://medium.com/huggingface/how-to-build-a-state-of-the-art-
+             conversational-ai-with-transfer-learning-2d818ac26313"""
+
+    if top_k > 0:
+        # Remove all tokens with a probability less than the
+        # last token of the top-k
+        indices_to_remove = logits < torch.topk(logits, top_k)[0][..., -1, None]
+        logits[indices_to_remove] = filter_value
+
+    if top_p > 0.0:
+        # Cconvert to 1D
+        sorted_logits, sorted_indices = torch.sort(logits, descending=True, dim=-1)
+        cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
+
+        # Remove tokens with cumulative probability above the threshold
+        sorted_indices_to_remove = cumulative_probs > top_p
+        # Shift the indices to the right to keep also the first token
+        # above the threshold
+        sorted_indices_to_remove[..., 1:] = sorted_indices_to_remove[..., :-1].clone()
+        sorted_indices_to_remove[..., 0] = 0
+        for i in range(sorted_indices.size(0)):
+            indices_to_remove = sorted_indices[i][sorted_indices_to_remove[i]]
+            logits[i][indices_to_remove] = filter_value
+
+    return logits
+
+
+def switch(val1, val2, boolean):
+    boolean = boolean.type_as(val1)
+    return (1 - boolean) * val1 + boolean * val2

visual.py

@@ -0,0 +1,375 @@
+from collections import OrderedDict
+import math
+import requests
+from io import BytesIO
+from functools import partial
+from PIL import Image
+from typing import Callable, Optional, Sequence, Tuple, List
+
+import torch
+from torch import nn
+from torch.nn import functional as F
+from torch.utils.checkpoint import checkpoint
+from torch.nn.init import trunc_normal_
+from torchvision import transforms
+from torchvision.transforms import InterpolationMode
+
+
+def get_abs_pos(abs_pos, tgt_size):
+    # abs_pos: L, C
+    # tgt_size: M
+    # return: M, C
+    src_size = int(math.sqrt(abs_pos.size(0)))
+    tgt_size = int(math.sqrt(tgt_size))
+    dtype = abs_pos.dtype
+
+    if src_size != tgt_size:
+        return F.interpolate(
+            abs_pos.float().reshape(1, src_size, src_size, -1).permute(0, 3, 1, 2),
+            size=(tgt_size, tgt_size),
+            mode="bicubic",
+            align_corners=False,
+        ).permute(0, 2, 3, 1).flatten(0, 2).to(dtype=dtype)
+    else:
+        return abs_pos
+
+
+class Resampler(nn.Module):
+    def __init__(
+            self,
+            grid_size,
+            embed_dim,
+            num_heads,
+            kv_dim=None,
+            norm_layer=nn.LayerNorm
+    ):
+        super().__init__()
+        self.num_queries = grid_size ** 2
+        self.embed_dim = embed_dim
+        self.num_heads = num_heads
+
+        self.pos_embed = nn.Parameter(torch.randn(embed_dim, grid_size)).requires_grad_(False)
+
+        self.query = nn.Parameter(torch.zeros(self.num_queries, embed_dim))
+        trunc_normal_(self.query, std=.02)
+
+        if kv_dim is not None and kv_dim != embed_dim:
+            self.kv_proj = nn.Linear(kv_dim, embed_dim, bias=False)
+        else:
+            self.kv_proj = nn.Identity()
+
+        self.attn = nn.MultiheadAttention(embed_dim, num_heads)
+        self.ln_q = norm_layer(embed_dim)
+        self.ln_kv = norm_layer(embed_dim)
+        
+        self.apply(self._init_weights)
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            trunc_normal_(m.weight, std=.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+
+    def forward(self, x, attn_mask=None):
+
+        pos_embed = get_abs_pos(self.pos_embed, x.size(1))
+
+        x = self.kv_proj(x)
+        x = self.ln_kv(x).permute(1, 0, 2)
+
+        N = x.shape[1]
+        q = self.ln_q(self.query)
+        out = self.attn(
+            self._repeat(q, N) + self.pos_embed.unsqueeze(1),
+            x + pos_embed.unsqueeze(1),
+            x,
+            attn_mask=attn_mask)[0]
+        return out.permute(1, 0, 2)
+
+    def _repeat(self, query, N: int):
+        return query.unsqueeze(1).repeat(1, N, 1)
+
+
+class VisualAttention(nn.Module):
+    """self-attention layer class.
+
+    Self-attention layer takes input with size [s, b, h]
+    and returns output of the same size.
+    """
+
+    def __init__(self, embed_dim, num_heads,
+                 bias=True, kdim=None, vdim=None):
+        super(VisualAttention, self).__init__()
+        self.embed_dim = embed_dim
+        self.kdim = kdim if kdim is not None else embed_dim
+        self.vdim = vdim if vdim is not None else embed_dim
+        self._qkv_same_embed_dim = self.kdim == embed_dim and self.vdim == embed_dim
+
+        self.num_heads = num_heads
+
+        # Per attention head and per partition values.
+        assert embed_dim % num_heads == 0
+        self.hidden_size_per_attention_head = embed_dim // num_heads
+        self.num_attention_heads_per_partition = num_heads
+        self.hidden_size_per_partition = embed_dim
+
+        # Strided linear layer.
+        assert self._qkv_same_embed_dim, 'Only Support SelfAttention Currently'
+        self.in_proj = nn.Linear(embed_dim, 3 * embed_dim)
+        self.out_proj = nn.Linear(embed_dim, embed_dim)
+        self.norm_factor = math.sqrt(self.hidden_size_per_attention_head)
+
+    def forward(self, query, key, value, attn_mask = None):
+        # query/key/value: [sq, b, h]
+        sq, b, _ = query.size()
+
+        assert query is key, 'Only Support Self-Attention Currently'
+        sk = sq
+        mixed_x_layer = self.in_proj(query)
+
+        # [sq, b, (np * 3 * hn)] --> [sq, b, np, 3 * hn]
+        new_tensor_shape = mixed_x_layer.size()[:-1] + \
+            (self.num_attention_heads_per_partition,
+             3 * self.hidden_size_per_attention_head)
+        mixed_x_layer = mixed_x_layer.view(*new_tensor_shape)
+
+        # [sq, b, np, 3 * hn] --> 3 [sq, b, np, hn]
+        query_layer, key_layer, value_layer = mixed_x_layer.split(
+            self.hidden_size_per_attention_head, dim=-1)
+
+        # [sq, b, np, hn] -> [sq, b * np, hn]
+        query_layer = query_layer.view(sq,
+            b * self.num_attention_heads_per_partition,
+            self.hidden_size_per_attention_head).transpose(0, 1)
+        # [sk, b, np, hn] -> [sk, b * np, hn]
+        key_layer = key_layer.view(sk,
+            b * self.num_attention_heads_per_partition,
+            self.hidden_size_per_attention_head).transpose(0, 1)
+
+        q_scaled = query_layer / self.norm_factor
+        if attn_mask is not None:
+            attention_probs = torch.baddbmm(attn_mask, q_scaled, key_layer.transpose(-2, -1))
+        else:
+            attention_probs = torch.bmm(q_scaled, key_layer.transpose(-2, -1))
+        attention_probs = attention_probs.softmax(dim=-1)
+
+        value_layer = value_layer.view(sk,
+            b * self.num_attention_heads_per_partition,
+            self.hidden_size_per_attention_head).transpose(0, 1)
+
+        # matmul: [b * np, sq, hn]
+        context_layer = torch.bmm(attention_probs, value_layer)
+
+        # change view [b, np, sq, hn]
+        context_layer = context_layer.view(b,
+            self.num_attention_heads_per_partition,
+            sq, self.hidden_size_per_attention_head)
+
+        # [b, np, sq, hn] --> [sq, b, np, hn]
+        context_layer = context_layer.permute(2, 0, 1, 3).contiguous()
+
+        # [sq, b, np, hn] --> [sq, b, hp]
+        new_context_layer_shape = context_layer.size()[:-2] + \
+            (self.hidden_size_per_partition,)
+        context_layer = context_layer.view(*new_context_layer_shape)
+
+        output = self.out_proj(context_layer)
+
+        return output
+
+
+class VisualAttentionBlock(nn.Module):
+    def __init__(
+            self,
+            d_model: int,
+            n_head: int,
+            mlp_ratio: float = 4.0,
+            act_layer: Callable = nn.GELU,
+            norm_layer: Callable = nn.LayerNorm,
+            is_cross_attention: bool = False,
+    ):
+        super().__init__()
+
+        self.ln_1 = norm_layer(d_model)
+        if is_cross_attention:
+            self.ln_1_kv = norm_layer(d_model)
+
+        self.ln_2 = norm_layer(d_model)
+        mlp_width = int(d_model * mlp_ratio)
+        self.attn = VisualAttention(d_model, n_head)
+        self.mlp = nn.Sequential(OrderedDict([
+            ("c_fc", nn.Linear(d_model, mlp_width)),
+            ("gelu", act_layer()),
+            ("c_proj", nn.Linear(mlp_width, d_model))
+        ]))
+
+    def attention(
+            self,
+            q_x: torch.Tensor,
+            k_x: Optional[torch.Tensor] = None,
+            v_x: Optional[torch.Tensor] = None,
+            attn_mask: Optional[torch.Tensor] = None,
+    ):
+        k_x = k_x if k_x is not None else q_x
+        v_x = v_x if v_x is not None else q_x
+
+        attn_mask = attn_mask.to(q_x.dtype) if attn_mask is not None else None
+        return self.attn(q_x, k_x, v_x, attn_mask=attn_mask)
+
+    def forward(
+            self,
+            q_x: torch.Tensor,
+            k_x: Optional[torch.Tensor] = None,
+            v_x: Optional[torch.Tensor] = None,
+            attn_mask: Optional[torch.Tensor] = None,
+    ):
+        k_x = self.ln_1_kv(k_x) if hasattr(self, "ln_1_kv") and k_x is not None else None
+        v_x = self.ln_1_kv(v_x) if hasattr(self, "ln_1_kv") and v_x is not None else None
+
+        x = q_x + self.attention(q_x=self.ln_1(q_x), k_x=k_x, v_x=v_x, attn_mask=attn_mask)
+        x = x + self.mlp(self.ln_2(x))
+        return x
+
+
+class Transformer(nn.Module):
+    def __init__(
+            self,
+            width: int,
+            layers: int,
+            heads: int,
+            mlp_ratio: float = 4.0,
+            act_layer: Callable = nn.GELU,
+            norm_layer: Callable = nn.LayerNorm,
+    ):
+        super().__init__()
+        self.width = width
+        self.layers = layers
+        self.grad_checkpointing = False
+
+        self.resblocks = nn.ModuleList([
+            VisualAttentionBlock(
+                width, heads, mlp_ratio, act_layer=act_layer, norm_layer=norm_layer)
+            for _ in range(layers)
+        ])
+
+    def get_cast_dtype(self) -> torch.dtype:
+        return self.resblocks[0].mlp.c_fc.weight.dtype
+
+    def get_cast_device(self) -> torch.device:
+        return self.resblocks[0].mlp.c_fc.weight.device
+
+    def forward(self, x: torch.Tensor, attn_mask: Optional[torch.Tensor] = None):
+        for r in self.resblocks:
+            if self.grad_checkpointing and not torch.jit.is_scripting():
+                # TODO: handle kwargs https://github.com/pytorch/pytorch/issues/79887#issuecomment-1161758372
+                x = checkpoint(r, x, None, None, attn_mask)
+            else:
+                x = r(x, attn_mask=attn_mask)
+        return x
+
+
+class VisionTransformer(nn.Module):
+
+    def __init__(
+            self,
+            image_size: int,
+            patch_size: int,
+            width: int,
+            layers: int,
+            heads: int,
+            mlp_ratio: float,
+            n_queries: int = 256,
+            output_dim: int = 512,
+            **kwargs
+    ):
+        super().__init__()
+        image_height, image_width = self.image_size = (image_size, image_size)
+        patch_height, patch_width = self.patch_size = (patch_size, patch_size)
+        self.grid_size = (image_height // patch_height, image_width // patch_width)
+        self.output_dim = output_dim
+
+        mean = (0.48145466, 0.4578275, 0.40821073)
+        std = (0.26862954, 0.26130258, 0.27577711)
+        self.image_transform = transforms.Compose([
+            transforms.Resize(
+                (image_size, image_size),
+                interpolation=InterpolationMode.BICUBIC
+            ),
+            transforms.ToTensor(),
+            transforms.Normalize(mean=mean, std=std),
+        ])
+
+        self.conv1 = nn.Conv2d(in_channels=3, out_channels=width, kernel_size=patch_size, stride=patch_size, bias=False)
+
+        # class embeddings and positional embeddings
+        scale = width ** -0.5
+        self.positional_embedding = nn.Parameter(scale * torch.randn(self.grid_size[0] * self.grid_size[1], width))
+
+        norm_layer = partial(nn.LayerNorm, eps=1e-6)
+        act_layer = nn.GELU
+
+        self.ln_pre = norm_layer(width)
+        self.transformer = Transformer(
+            width,
+            layers,
+            heads,
+            mlp_ratio,
+            act_layer=act_layer,
+            norm_layer=norm_layer,
+        )
+
+        self.attn_pool = Resampler(
+            grid_size=int(math.sqrt(n_queries)),
+            embed_dim=output_dim,
+            num_heads=output_dim // 128,
+            kv_dim=width,
+            norm_layer=norm_layer,
+        )
+        self.ln_post = norm_layer(output_dim)
+        self.proj = nn.Parameter((output_dim** -0.5) * torch.randn(output_dim, output_dim))
+
+    @torch.jit.ignore
+    def set_grad_checkpointing(self, enable=True):
+        self.transformer.grad_checkpointing = enable
+
+    def forward(self, x: torch.Tensor):
+        x = x.to(
+            dtype=self.transformer.get_cast_dtype(),
+            device=self.transformer.get_cast_device(),
+        )
+        # to patches
+        x = self.conv1(x)  # shape = [*, width, grid, grid]
+        x = x.reshape(x.shape[0], x.shape[1], -1)  # shape = [*, width, grid ** 2]
+        x = x.permute(0, 2, 1)  # shape = [*, grid ** 2, width]
+
+        x = x + get_abs_pos(self.positional_embedding, x.size(1))
+
+        x = self.ln_pre(x)
+
+        x = x.permute(1, 0, 2)  # NLD -> LND
+        x = self.transformer(x)
+        x = x.permute(1, 0, 2)  # LND -> NLD
+
+        if self.attn_pool:
+            x = self.attn_pool(x)
+        x = self.ln_post(x)
+        x = x @ self.proj
+
+        return x
+
+    def encode(self, image_paths: List[str]):
+        images = []
+        for image_path in image_paths:
+            if image_path.startswith("http://") or image_path.startswith("https://"):
+                image = Image.open(requests.get(image_path, stream=True).raw)
+            elif image_path.startswith("oss://"):
+                raise NotImplementedError
+            else:
+                image = Image.open(image_path)
+            image = image.convert("RGB")
+            images.append(self.image_transform(image))
+        images = torch.stack(images, dim=0)
+        return self(images)