Add files using upload-large-folder tool

.gitattributes

@@ -586,3 +586,10 @@ evalkit_internvl/lib/python3.10/site-packages/numpy.libs/libgfortran-040039e1.so
 evalkit_internvl/lib/python3.10/site-packages/huggingface_hub/inference/__pycache__/_client.cpython-310.pyc filter=lfs diff=lfs merge=lfs -text
 evalkit_internvl/lib/python3.10/site-packages/huggingface_hub/inference/_generated/__pycache__/_async_client.cpython-310.pyc filter=lfs diff=lfs merge=lfs -text
 evalkit_internvl/lib/python3.10/site-packages/xlsxwriter/__pycache__/worksheet.cpython-310.pyc filter=lfs diff=lfs merge=lfs -text
+evalkit_tf437/lib/python3.10/site-packages/decord.libs/libavutil-2b26904a.so.56.22.100 filter=lfs diff=lfs merge=lfs -text
+evalkit_tf437/lib/python3.10/site-packages/decord.libs/libavformat-8b46ea57.so.58.20.100 filter=lfs diff=lfs merge=lfs -text
+evalkit_tf437/lib/python3.10/site-packages/decord.libs/libpostproc-88b722f8.so.55.3.100 filter=lfs diff=lfs merge=lfs -text
+evalkit_tf437/lib/python3.10/site-packages/decord.libs/libswscale-8e37dcfd.so.5.3.100 filter=lfs diff=lfs merge=lfs -text
+evalkit_tf437/lib/python3.10/site-packages/decord.libs/libxcb-77222338.so.1.1.0 filter=lfs diff=lfs merge=lfs -text
+evalkit_tf437/lib/python3.10/site-packages/decord.libs/libavfilter-1e2243e2.so.7.40.101 filter=lfs diff=lfs merge=lfs -text
+evalkit_tf437/lib/python3.10/site-packages/decord.libs/libavdevice-bf61e037.so.58.5.100 filter=lfs diff=lfs merge=lfs -text

evalkit_internvl/lib/python3.10/site-packages/open_clip/__init__.py

@@ -0,0 +1,18 @@
+from .version import __version__
+
+from .coca_model import CoCa
+from .constants import OPENAI_DATASET_MEAN, OPENAI_DATASET_STD
+from .factory import create_model, create_model_and_transforms, create_model_from_pretrained, get_tokenizer, create_loss
+from .factory import list_models, add_model_config, get_model_config, load_checkpoint
+from .loss import ClipLoss, DistillClipLoss, CoCaLoss
+from .model import CLIP, CustomTextCLIP, CLIPTextCfg, CLIPVisionCfg, \
+    convert_weights_to_lp, convert_weights_to_fp16, trace_model, get_cast_dtype, get_input_dtype, \
+    get_model_tokenize_cfg, get_model_preprocess_cfg, set_model_preprocess_cfg
+from .openai import load_openai_model, list_openai_models
+from .pretrained import list_pretrained, list_pretrained_models_by_tag, list_pretrained_tags_by_model, \
+    get_pretrained_url, download_pretrained_from_url, is_pretrained_cfg, get_pretrained_cfg, download_pretrained
+from .push_to_hf_hub import push_pretrained_to_hf_hub, push_to_hf_hub
+from .tokenizer import SimpleTokenizer, tokenize, decode
+from .transform import image_transform, AugmentationCfg
+from .zero_shot_classifier import build_zero_shot_classifier, build_zero_shot_classifier_legacy
+from .zero_shot_metadata import OPENAI_IMAGENET_TEMPLATES, SIMPLE_IMAGENET_TEMPLATES, IMAGENET_CLASSNAMES

evalkit_internvl/lib/python3.10/site-packages/open_clip/coca_model.py

@@ -0,0 +1,498 @@
+from typing import Optional
+
+import torch
+from torch import nn
+from torch.nn import functional as F
+import numpy as np
+from dataclasses import dataclass
+
+from .transformer import (
+    LayerNormFp32,
+    LayerNorm,
+    QuickGELU,
+    MultimodalTransformer,
+)
+from .model import CLIPTextCfg, CLIPVisionCfg, _build_vision_tower, _build_text_tower
+
+try:
+    from transformers import (
+        BeamSearchScorer,
+        LogitsProcessorList,
+        TopPLogitsWarper,
+        TopKLogitsWarper,
+        RepetitionPenaltyLogitsProcessor,
+        MinLengthLogitsProcessor,
+        MaxLengthCriteria,
+        StopStringCriteria,
+        EosTokenCriteria,
+        StoppingCriteriaList
+    )
+
+    GENERATION_TYPES = {
+        "top_k": TopKLogitsWarper,
+        "top_p": TopPLogitsWarper,
+        "beam_search": "beam_search"
+    }
+    _has_transformers = True
+except ImportError as e:
+    GENERATION_TYPES = {
+        "top_k": None,
+        "top_p": None,
+        "beam_search": "beam_search"
+    }
+    _has_transformers = False
+
+
+@dataclass
+class MultimodalCfg(CLIPTextCfg):
+    mlp_ratio: int = 4
+    dim_head: int = 64
+    heads: int = 8
+    n_queries: int = 256
+    attn_pooler_heads: int = 8
+
+
+def _build_text_decoder_tower(
+        embed_dim,
+        multimodal_cfg,
+        quick_gelu: bool = False,
+        cast_dtype: Optional[torch.dtype] = None,
+):
+    multimodal_cfg = MultimodalCfg(**multimodal_cfg) if isinstance(multimodal_cfg, dict) else multimodal_cfg
+    act_layer = QuickGELU if quick_gelu else nn.GELU
+    norm_layer = (
+        LayerNormFp32 if cast_dtype in (torch.float16, torch.bfloat16) else LayerNorm
+    )
+
+    decoder = MultimodalTransformer(
+        context_length=multimodal_cfg.context_length,
+        width=multimodal_cfg.width,
+        heads=multimodal_cfg.heads,
+        layers=multimodal_cfg.layers,
+        ls_init_value=multimodal_cfg.ls_init_value,
+        output_dim=embed_dim,
+        act_layer=act_layer,
+        norm_layer=norm_layer,
+    )
+
+    return decoder
+
+
+def _token_to_tensor(token_id, device: str = "cpu") -> torch.Tensor:
+    if not isinstance(token_id, torch.Tensor):
+        if isinstance(token_id, int):
+            token_id = [token_id]
+        token_id = torch.tensor(token_id, device=device)
+    return token_id
+
+
+class CoCa(nn.Module):
+    def __init__(
+            self,
+            embed_dim,
+            multimodal_cfg: MultimodalCfg,
+            text_cfg: CLIPTextCfg,
+            vision_cfg: CLIPVisionCfg,
+            quick_gelu: bool = False,
+            init_logit_scale: float = np.log(1 / 0.07),
+            init_logit_bias: Optional[float] = None,
+            cast_dtype: Optional[torch.dtype] = None,
+            pad_id: int = 0,
+    ):
+        super().__init__()
+        multimodal_cfg = MultimodalCfg(**multimodal_cfg) if isinstance(multimodal_cfg, dict) else multimodal_cfg
+        text_cfg = CLIPTextCfg(**text_cfg) if isinstance(text_cfg, dict) else text_cfg
+        vision_cfg = CLIPVisionCfg(**vision_cfg) if isinstance(vision_cfg, dict) else vision_cfg
+
+        self.text = _build_text_tower(
+            embed_dim=embed_dim,
+            text_cfg=text_cfg,
+            quick_gelu=quick_gelu,
+            cast_dtype=cast_dtype,
+        )
+
+        vocab_size = (
+            text_cfg.vocab_size  # for hf models
+            if hasattr(text_cfg, "hf_model_name") and text_cfg.hf_model_name is not None
+            else text_cfg.vocab_size
+        )
+
+        self.visual = _build_vision_tower(
+            embed_dim=embed_dim,
+            vision_cfg=vision_cfg,
+            quick_gelu=quick_gelu,
+            cast_dtype=cast_dtype,
+        )
+
+        self.text_decoder = _build_text_decoder_tower(
+            vocab_size,
+            multimodal_cfg=multimodal_cfg,
+            quick_gelu=quick_gelu,
+            cast_dtype=cast_dtype,
+        )
+
+        self.logit_scale = nn.Parameter(torch.ones([]) * init_logit_scale)
+        if init_logit_bias is not None:
+            self.logit_bias = nn.Parameter(torch.ones([]) * init_logit_bias)
+        else:
+            self.logit_bias = None
+        self.pad_id = pad_id
+
+        self.context_length = multimodal_cfg.context_length
+
+    @torch.jit.ignore
+    def set_grad_checkpointing(self, enable: bool = True):
+        self.visual.set_grad_checkpointing(enable)
+        self.text.set_grad_checkpointing(enable)
+        self.text_decoder.set_grad_checkpointing(enable)
+
+    def _encode_image(self, images, normalize: bool = True):
+        image_latent, tokens_embs = self.visual(images)
+        image_latent = F.normalize(image_latent, dim=-1) if normalize else image_latent
+        return image_latent, tokens_embs
+
+    def _encode_text(self, text, normalize: bool = True):
+        text_latent, token_emb = self.text(text)
+        text_latent = F.normalize(text_latent, dim=-1) if normalize else text_latent
+        return text_latent, token_emb
+
+    def encode_image(self, images, normalize: bool = True):
+        image_latent, _ = self._encode_image(images, normalize=normalize)
+        return image_latent
+
+    def encode_text(self, text, normalize: bool = True):
+        text_latent, _ = self._encode_text(text, normalize=normalize)
+        return text_latent
+
+    def forward(
+            self,
+            image,
+            text: Optional[torch.Tensor] = None,
+            image_latent: Optional[torch.Tensor] = None,
+            image_embs: Optional[torch.Tensor] = None,
+            output_labels: bool = True,
+    ):
+        if image_latent is None or image_embs is None:
+            image_latent, image_embs = self._encode_image(image)
+
+        if text is None:
+            return {"image_features": image_latent, "image_embs": image_embs}
+
+        text_latent, token_embs = self._encode_text(text)
+
+        # FIXME this isn't an ideal solution, would like to improve -RW
+        labels: Optional[torch.Tensor] = text[:, 1:] if output_labels else None
+        if output_labels:
+            # align text_embs and thus logits with labels for teacher-forcing caption loss
+            token_embs = token_embs[:, :-1]
+
+        logits = self.text_decoder(image_embs, token_embs)
+        out_dict = {
+            "image_features": image_latent,
+            "text_features": text_latent,
+            "logits": logits,
+            "logit_scale": self.logit_scale.exp()
+        }
+        if labels is not None:
+            out_dict["labels"] = labels
+        if self.logit_bias is not None:
+            out_dict["logit_bias"] = self.logit_bias
+        return out_dict
+
+    def generate(
+        self,
+        image,
+        text=None,
+        seq_len=30,
+        max_seq_len=77,
+        temperature=1.,
+        generation_type="beam_search",
+        top_p=0.1,  # keep tokens in the 1 - top_p quantile
+        top_k=1,  # keeps the top_k most probable tokens
+        pad_token_id=None,
+        eos_token_id=None,
+        sot_token_id=None,
+        num_beams=6,
+        num_beam_groups=3,
+        min_seq_len=5,
+        stopping_criteria=None,
+        repetition_penalty=1.0,
+        fixed_output_length=False # if True output.shape == (batch_size, seq_len)
+    ):
+        # taking many ideas and components from HuggingFace GenerationMixin
+        # https://huggingface.co/docs/transformers/main/en/main_classes/text_generation
+        assert _has_transformers, "Please install transformers for generate functionality. `pip install transformers`."
+        assert seq_len > min_seq_len, "seq_len must be larger than min_seq_len"
+        device = image.device
+
+        with torch.no_grad():
+            sot_token_id = _token_to_tensor(49406 if sot_token_id is None else sot_token_id, device=device)
+            eos_token_id = _token_to_tensor(49407 if eos_token_id is None else eos_token_id, device=device)
+            pad_token_id = self.pad_id if pad_token_id is None else pad_token_id
+            logit_processor = LogitsProcessorList(
+                [
+                    MinLengthLogitsProcessor(min_seq_len, eos_token_id),
+                    RepetitionPenaltyLogitsProcessor(repetition_penalty),
+                ]
+            )
+
+            if stopping_criteria is None:
+                stopping_criteria = [MaxLengthCriteria(max_length=seq_len)]
+            stopping_criteria = StoppingCriteriaList(stopping_criteria)
+
+            if generation_type == "beam_search":
+                output = self._generate_beamsearch(
+                    image_inputs=image,
+                    pad_token_id=pad_token_id,
+                    eos_token_id=eos_token_id,
+                    sot_token_id=sot_token_id,
+                    num_beams=num_beams,
+                    num_beam_groups=num_beam_groups,
+                    min_seq_len=min_seq_len,
+                    stopping_criteria=stopping_criteria,
+                    logit_processor=logit_processor,
+                )
+                if fixed_output_length and output.shape[1] < seq_len:
+                    pad_len = seq_len - output.shape[1]
+                    return torch.cat((
+                            output,
+                            torch.ones(output.shape[0], pad_len, device=device, dtype=output.dtype) * pad_token_id
+                        ),
+                        dim=1
+                    )
+                return output
+
+            elif generation_type == "top_p":
+                logit_warper = GENERATION_TYPES[generation_type](top_p)
+            elif generation_type == "top_k":
+                logit_warper = GENERATION_TYPES[generation_type](top_k)
+            else:
+                raise ValueError(
+                    f"generation_type has to be one of "
+                    f"{'| ' + ' | '.join(list(GENERATION_TYPES.keys())) + ' |'}."
+                )
+
+            image_latent, image_embs = self._encode_image(image)
+
+            if text is None:
+                text = torch.ones((image.shape[0], 1), device=device, dtype=torch.long) * sot_token_id
+
+            was_training = self.training
+            num_dims = len(text.shape)
+
+            if num_dims == 1:
+                text = text[None, :]
+
+            self.eval()
+            out = text
+
+            while True:
+                x = out[:, -max_seq_len:]
+                cur_len = x.shape[1]
+                logits = self(
+                    image,
+                    x,
+                    image_latent=image_latent,
+                    image_embs=image_embs,
+                    output_labels=False,
+                )["logits"][:, -1]
+                mask = (out[:, -1] == eos_token_id) | (out[:, -1] == pad_token_id)
+                sample = torch.ones((out.shape[0], 1), device=device, dtype=torch.long) * pad_token_id
+
+                if mask.all():
+                    if not fixed_output_length:
+                        break
+                else:
+                    logits = logits[~mask, :]
+                    filtered_logits = logit_processor(x[~mask, :], logits)
+                    filtered_logits = logit_warper(x[~mask, :], filtered_logits)
+                    probs = F.softmax(filtered_logits / temperature, dim=-1)
+
+                    if (cur_len + 1 == seq_len):
+                        sample[~mask, :] = torch.ones((sum(~mask), 1), device=device, dtype=torch.long) * eos_token_id
+                    else:
+                        sample[~mask, :] = torch.multinomial(probs, 1)
+
+                out = torch.cat((out, sample), dim=-1)
+
+                cur_len += 1
+
+                if all(stopping_criteria(out, None)):
+                    break
+
+            if num_dims == 1:
+                out = out.squeeze(0)
+
+            self.train(was_training)
+            return out
+
+    def _generate_beamsearch(
+            self,
+            image_inputs,
+            pad_token_id=None,
+            eos_token_id=None,
+            sot_token_id=None,
+            num_beams=6,
+            num_beam_groups=3,
+            min_seq_len=5,
+            stopping_criteria=None,
+            logit_processor=None,
+            logit_warper=None,
+    ):
+        device = image_inputs.device
+        batch_size = image_inputs.shape[0]
+        image_inputs = torch.repeat_interleave(image_inputs, num_beams, dim=0)
+        image_latent, image_embs = self._encode_image(image_inputs)
+
+        input_ids = torch.ones((batch_size * num_beams, 1), device=device, dtype=torch.long)
+        input_ids = input_ids * sot_token_id
+        beam_scorer = BeamSearchScorer(
+            batch_size=batch_size,
+            num_beams=num_beams,
+            device=device,
+            num_beam_groups=num_beam_groups,
+        )
+        # instantiate logits processors
+        logits_processor = (
+            LogitsProcessorList([MinLengthLogitsProcessor(min_seq_len, eos_token_id=eos_token_id)])
+            if logit_processor is None
+            else logit_processor
+        )
+
+        num_beams = beam_scorer.num_beams
+        num_beam_groups = beam_scorer.num_beam_groups
+        num_sub_beams = num_beams // num_beam_groups
+        batch_size = len(beam_scorer._beam_hyps) // num_beam_groups
+        batch_beam_size, cur_len = input_ids.shape
+        beam_indices = None
+
+        if num_beams * batch_size != batch_beam_size:
+            raise ValueError(
+                f"Batch dimension of `input_ids` should be {num_beams * batch_size}, but is {batch_beam_size}."
+            )
+
+        beam_scores = torch.full((batch_size, num_beams), -1e9, dtype=torch.float, device=device)
+        # initialise score of first beam of each group with 0 and the rest with 1e-9. This ensures that the beams in
+        # the same group don't produce same tokens everytime.
+        beam_scores[:, ::num_sub_beams] = 0
+        beam_scores = beam_scores.view((batch_size * num_beams,))
+
+        while True:
+
+            # predicted tokens in cur_len step
+            current_tokens = torch.zeros(batch_size * num_beams, dtype=input_ids.dtype, device=device)
+
+            # indices which will form the beams in the next time step
+            reordering_indices = torch.zeros(batch_size * num_beams, dtype=torch.long, device=device)
+
+            # do one decoder step on all beams of all sentences in batch
+            model_inputs = prepare_inputs_for_generation(input_ids=input_ids, image_inputs=image_inputs)
+            outputs = self(
+                model_inputs['images'],
+                model_inputs['text'],
+                image_latent=image_latent,
+                image_embs=image_embs,
+                output_labels=False,
+            )
+
+            for beam_group_idx in range(num_beam_groups):
+                group_start_idx = beam_group_idx * num_sub_beams
+                group_end_idx = min(group_start_idx + num_sub_beams, num_beams)
+                group_size = group_end_idx - group_start_idx
+
+                # indices of beams of current group among all sentences in batch
+                batch_group_indices = []
+
+                for batch_idx in range(batch_size):
+                    batch_group_indices.extend(
+                        [batch_idx * num_beams + idx for idx in range(group_start_idx, group_end_idx)]
+                    )
+                group_input_ids = input_ids[batch_group_indices]
+
+                # select outputs of beams of currentg group only
+                next_token_logits = outputs['logits'][batch_group_indices, -1, :]
+                vocab_size = next_token_logits.shape[-1]
+
+                next_token_scores_processed = logits_processor(
+                    group_input_ids, next_token_logits, current_tokens=current_tokens, beam_group_idx=beam_group_idx
+                )
+                next_token_scores = next_token_scores_processed + beam_scores[batch_group_indices].unsqueeze(-1)
+                next_token_scores = next_token_scores.expand_as(next_token_scores_processed)
+
+                # reshape for beam search
+                next_token_scores = next_token_scores.view(batch_size, group_size * vocab_size)
+
+                next_token_scores, next_tokens = torch.topk(
+                    next_token_scores, 2 * group_size, dim=1, largest=True, sorted=True
+                )
+
+                next_indices = torch.div(next_tokens, vocab_size, rounding_mode="floor")
+                next_tokens = next_tokens % vocab_size
+
+                # stateless
+                process_beam_indices = sum(beam_indices, ()) if beam_indices is not None else None
+                beam_outputs = beam_scorer.process(
+                    group_input_ids,
+                    next_token_scores,
+                    next_tokens,
+                    next_indices,
+                    pad_token_id=pad_token_id,
+                    eos_token_id=eos_token_id,
+                    beam_indices=process_beam_indices,
+                    group_index=beam_group_idx,
+                )
+                beam_scores[batch_group_indices] = beam_outputs["next_beam_scores"]
+                beam_next_tokens = beam_outputs["next_beam_tokens"]
+                beam_idx = beam_outputs["next_beam_indices"]
+
+                input_ids[batch_group_indices] = group_input_ids[beam_idx]
+                group_input_ids = torch.cat([group_input_ids[beam_idx, :], beam_next_tokens.unsqueeze(-1)], dim=-1)
+                current_tokens[batch_group_indices] = group_input_ids[:, -1]
+
+                # (beam_idx // group_size) -> batch_idx
+                # (beam_idx % group_size) -> offset of idx inside the group
+                reordering_indices[batch_group_indices] = (
+                    num_beams * torch.div(beam_idx, group_size, rounding_mode="floor") + group_start_idx + (beam_idx % group_size)
+                )
+
+            input_ids = torch.cat([input_ids, current_tokens.unsqueeze(-1)], dim=-1)
+
+            # increase cur_len
+            cur_len = cur_len + 1
+            if beam_scorer.is_done or all(stopping_criteria(input_ids, None)):
+                break
+
+        final_beam_indices = sum(beam_indices, ()) if beam_indices is not None else None
+        sequence_outputs = beam_scorer.finalize(
+            input_ids,
+            beam_scores,
+            next_tokens,
+            next_indices,
+            pad_token_id=pad_token_id,
+            eos_token_id=eos_token_id,
+            max_length=stopping_criteria.max_length,
+            beam_indices=final_beam_indices,
+        )
+        return sequence_outputs['sequences']
+
+
+def prepare_inputs_for_generation(input_ids, image_inputs, past=None, **kwargs):
+    if past:
+        input_ids = input_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:
+        # create position_ids on the fly for batch generation
+        position_ids = attention_mask.long().cumsum(-1) - 1
+        position_ids.masked_fill_(attention_mask == 0, 1)
+    else:
+        position_ids = None
+    return {
+        "text": input_ids,
+        "images": image_inputs,
+        "past_key_values": past,
+        "position_ids": position_ids,
+        "attention_mask": attention_mask,
+    }

evalkit_internvl/lib/python3.10/site-packages/open_clip/hf_model.py

@@ -0,0 +1,193 @@
+""" huggingface model adapter
+
+Wraps HuggingFace transformers (https://github.com/huggingface/transformers) models for use as a text tower in CLIP model.
+"""
+import re
+
+import torch
+import torch.nn as nn
+from torch import TensorType
+
+try:
+    import transformers
+    from transformers import AutoModel, AutoTokenizer, AutoConfig, PretrainedConfig
+    from transformers.modeling_outputs import BaseModelOutput, BaseModelOutputWithPooling, \
+        BaseModelOutputWithPoolingAndCrossAttentions
+except ImportError as e:
+    transformers = None
+
+
+    class BaseModelOutput:
+        pass
+
+
+    class PretrainedConfig:
+        pass
+
+from .hf_configs import arch_dict
+
+
+# utils
+def _camel2snake(s):
+    return re.sub(r'(?<!^)(?=[A-Z])', '_', s).lower()
+
+
+# TODO: ?last - for gpt-like models
+_POOLERS = {}
+
+
+def register_pooler(cls):
+    """Decorator registering pooler class"""
+    _POOLERS[_camel2snake(cls.__name__)] = cls
+    return cls
+
+
+@register_pooler
+class MeanPooler(nn.Module):
+    """Mean pooling"""
+
+    def forward(self, x: BaseModelOutput, attention_mask: TensorType):
+        masked_output = x.last_hidden_state * attention_mask.unsqueeze(-1)
+        return masked_output.sum(dim=1) / attention_mask.sum(-1, keepdim=True)
+
+
+@register_pooler
+class MaxPooler(nn.Module):
+    """Max pooling"""
+
+    def forward(self, x: BaseModelOutput, attention_mask: TensorType):
+        masked_output = x.last_hidden_state.masked_fill(attention_mask.unsqueeze(-1), -torch.inf)
+        return masked_output.max(1).values
+
+
+@register_pooler
+class ClsPooler(nn.Module):
+    """CLS token pooling"""
+
+    def __init__(self, use_pooler_output=True):
+        super().__init__()
+        self.cls_token_position = 0
+        self.use_pooler_output = use_pooler_output
+
+    def forward(self, x: BaseModelOutput, attention_mask: TensorType):
+        if (self.use_pooler_output and
+            isinstance(x, (BaseModelOutputWithPooling, BaseModelOutputWithPoolingAndCrossAttentions)) and
+            (x.pooler_output is not None)
+        ):
+            return x.pooler_output
+
+        return x.last_hidden_state[:, self.cls_token_position, :]
+
+
+@register_pooler
+class ClsLastHiddenStatePooler(nn.Module):
+    """CLS token pooling
+    NOTE: this is equivalent to ClsPooler above with use_pooler_output=False
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.cls_token_position = 0
+
+    def forward(self, x: BaseModelOutput, attention_mask: TensorType):
+        return x.last_hidden_state[:, self.cls_token_position, :]
+
+
+class HFTextEncoder(nn.Module):
+    """HuggingFace model adapter"""
+    output_tokens: torch.jit.Final[bool]
+
+    def __init__(
+            self,
+            model_name_or_path: str,
+            output_dim: int,
+            config: PretrainedConfig = None,
+            pooler_type: str = None,
+            proj_type: str = None,
+            pretrained: bool = True,
+            output_tokens: bool = False,
+    ):
+        super().__init__()
+        self.output_tokens = output_tokens
+        self.output_dim = output_dim
+
+        # TODO: find better way to get this information
+        uses_transformer_pooler = (pooler_type == "cls_pooler")
+
+        if transformers is None:
+            raise RuntimeError("Please `pip install transformers` to use pre-trained HuggingFace models")
+        if config is None:
+            self.config = AutoConfig.from_pretrained(model_name_or_path)
+            create_func, model_args = (AutoModel.from_pretrained, model_name_or_path) if pretrained else (
+                AutoModel.from_config, self.config)
+            # TODO: do all model configs have this attribute? PretrainedConfig does so yes??
+            if hasattr(self.config, "is_encoder_decoder") and self.config.is_encoder_decoder:
+                self.transformer = create_func(model_args)
+                self.transformer = self.transformer.encoder
+            else:
+                self.transformer = create_func(model_args, add_pooling_layer=uses_transformer_pooler)
+        else:
+            self.config = config
+            self.transformer = AutoModel.from_config(config)
+        if pooler_type is None:  # get default arch pooler
+            pooler_type = (arch_dict[self.config.model_type]["pooler"])
+
+        # FIXME downstream users of OpenCLIP models use these attr, need to verify valid across all models
+        self.vocab_size = getattr(self.config, 'vocab_size', 0)
+        self.context_length = getattr(self.config, 'max_position_embeddings', 0)
+
+        self.pooler = _POOLERS[pooler_type]()
+
+        d_model = getattr(self.config, arch_dict[self.config.model_type]["config_names"]["width"])
+        if (d_model == output_dim) and (proj_type is None):  # do we always need a proj?
+            self.proj = nn.Identity()
+        elif proj_type == 'linear':
+            self.proj = nn.Linear(d_model, output_dim, bias=False)
+        elif proj_type == 'mlp':
+            hidden_size = (d_model + output_dim) // 2
+            self.proj = nn.Sequential(
+                nn.Linear(d_model, hidden_size, bias=False),
+                nn.GELU(),
+                nn.Linear(hidden_size, output_dim, bias=False),
+            )
+
+    def forward(self, x: TensorType):
+        attn_mask = (x != self.config.pad_token_id).long()
+        out = self.transformer(input_ids=x, attention_mask=attn_mask)
+        pooled_out = self.pooler(out, attn_mask)
+        projected = self.proj(pooled_out)
+
+        seq_len = out.last_hidden_state.shape[1]
+        tokens = (
+            out.last_hidden_state[:, torch.arange(seq_len) != self.pooler.cls_token_position, :] 
+            if type(self.pooler) == ClsPooler 
+            else out.last_hidden_state
+        )
+        
+        if self.output_tokens:
+            return projected, tokens
+        return projected
+
+    def lock(self, unlocked_layers: int = 0, freeze_layer_norm: bool = True):
+        if not unlocked_layers:  # full freezing
+            for n, p in self.transformer.named_parameters():
+                p.requires_grad = (not freeze_layer_norm) if "LayerNorm" in n.split(".") else False
+            return
+
+        encoder = self.transformer.encoder if hasattr(self.transformer, 'encoder') else self.transformer
+        layer_list = getattr(encoder, arch_dict[self.config.model_type]["config_names"]["layer_attr"])
+        print(f"Unlocking {unlocked_layers}/{len(layer_list) + 1} layers of hf model")
+        embeddings = getattr(
+            self.transformer, arch_dict[self.config.model_type]["config_names"]["token_embeddings_attr"])
+        modules = [embeddings, *layer_list][:-unlocked_layers]
+        # freeze layers
+        for module in modules:
+            for n, p in module.named_parameters():
+                p.requires_grad = (not freeze_layer_norm) if "LayerNorm" in n.split(".") else False
+
+    @torch.jit.ignore
+    def set_grad_checkpointing(self, enable=True):
+        self.transformer.gradient_checkpointing_enable()
+
+    def init_parameters(self):
+        pass

evalkit_internvl/lib/python3.10/site-packages/open_clip/openai.py

@@ -0,0 +1,90 @@
+""" OpenAI pretrained model functions
+
+Adapted from https://github.com/openai/CLIP. Originally MIT License, Copyright (c) 2021 OpenAI.
+"""
+
+import os
+import warnings
+from typing import List, Optional, Union
+
+import torch
+
+from .constants import OPENAI_DATASET_MEAN, OPENAI_DATASET_STD
+from .model import build_model_from_openai_state_dict, convert_weights_to_lp, get_cast_dtype
+from .pretrained import get_pretrained_url, list_pretrained_models_by_tag, download_pretrained_from_url
+
+__all__ = ["list_openai_models", "load_openai_model"]
+
+
+def list_openai_models() -> List[str]:
+    """Returns the names of available CLIP models"""
+    return list_pretrained_models_by_tag('openai')
+
+
+def load_openai_model(
+        name: str,
+        precision: Optional[str] = None,
+        device: Optional[Union[str, torch.device]] = None,
+        cache_dir: Optional[str] = None,
+):
+    """Load a CLIP model
+
+    Parameters
+    ----------
+    name : str
+        A model name listed by `clip.available_models()`, or the path to a model checkpoint containing the state_dict
+    precision: str
+        Model precision, if None defaults to 'fp32' if device == 'cpu' else 'fp16'.
+    device : Union[str, torch.device]
+        The device to put the loaded model
+    cache_dir : Optional[str]
+        The directory to cache the downloaded model weights
+
+    Returns
+    -------
+    model : torch.nn.Module
+        The CLIP model
+    preprocess : Callable[[PIL.Image], torch.Tensor]
+        A torchvision transform that converts a PIL image into a tensor that the returned model can take as its input
+    """
+    if device is None:
+        device = "cuda" if torch.cuda.is_available() else "cpu"
+    if precision is None:
+        precision = 'fp32' if device == 'cpu' else 'fp16'
+
+    if get_pretrained_url(name, 'openai'):
+        model_path = download_pretrained_from_url(get_pretrained_url(name, 'openai'), cache_dir=cache_dir)
+    elif os.path.isfile(name):
+        model_path = name
+    else:
+        raise RuntimeError(f"Model {name} not found; available models = {list_openai_models()}")
+
+    try:
+        # loading JIT archive
+        model = torch.jit.load(model_path, map_location="cpu").eval()
+        state_dict = None
+    except RuntimeError:
+        # loading saved state dict
+        state_dict = torch.load(model_path, map_location="cpu")
+
+    # Build a non-jit model from the OpenAI jitted model state dict
+    cast_dtype = get_cast_dtype(precision)
+    try:
+        model = build_model_from_openai_state_dict(state_dict or model.state_dict(), cast_dtype=cast_dtype)
+    except KeyError:
+        sd = {k[7:]: v for k, v in state_dict["state_dict"].items()}
+        model = build_model_from_openai_state_dict(sd, cast_dtype=cast_dtype)
+
+    # model from OpenAI state dict is in manually cast fp16 mode, must be converted for AMP/fp32/bf16 use
+    model = model.to(device)
+    # FIXME support pure fp16/bf16 precision modes
+    if precision != 'fp16':
+        model.float()
+        if precision == 'bf16':
+            # for bf16, convert back to low-precision
+            convert_weights_to_lp(model, dtype=torch.bfloat16)
+
+    # add mean / std attributes for consistency with OpenCLIP models
+    model.visual.image_mean = OPENAI_DATASET_MEAN
+    model.visual.image_std = OPENAI_DATASET_STD
+    return model

evalkit_internvl/lib/python3.10/site-packages/open_clip/pos_embed.py

@@ -0,0 +1,96 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+# --------------------------------------------------------
+# Position embedding utils
+# --------------------------------------------------------
+
+import numpy as np
+
+import torch
+
+# --------------------------------------------------------
+# 2D sine-cosine position embedding
+# References:
+# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py
+# MoCo v3: https://github.com/facebookresearch/moco-v3
+# --------------------------------------------------------
+def get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):
+    """
+    grid_size: int of the grid height and width
+    return:
+    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)
+    """
+    grid_h = np.arange(grid_size, dtype=np.float32)
+    grid_w = np.arange(grid_size, dtype=np.float32)
+    grid = np.meshgrid(grid_w, grid_h)  # here w goes first
+    grid = np.stack(grid, axis=0)
+
+    grid = grid.reshape([2, 1, grid_size, grid_size])
+    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)
+    if cls_token:
+        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)
+    return pos_embed
+
+
+def get_2d_sincos_pos_embed_from_grid(embed_dim, grid):
+    assert embed_dim % 2 == 0
+
+    # use half of dimensions to encode grid_h
+    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)
+    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)
+
+    emb = np.concatenate([emb_h, emb_w], axis=1) # (H*W, D)
+    return emb
+
+
+def get_1d_sincos_pos_embed_from_grid(embed_dim, pos):
+    """
+    embed_dim: output dimension for each position
+    pos: a list of positions to be encoded: size (M,)
+    out: (M, D)
+    """
+    assert embed_dim % 2 == 0
+    omega = np.arange(embed_dim // 2, dtype=float)
+    omega /= embed_dim / 2.
+    omega = 1. / 10000**omega  # (D/2,)
+
+    pos = pos.reshape(-1)  # (M,)
+    out = np.einsum('m,d->md', pos, omega)  # (M, D/2), outer product
+
+    emb_sin = np.sin(out) # (M, D/2)
+    emb_cos = np.cos(out) # (M, D/2)
+
+    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)
+    return emb
+
+
+# --------------------------------------------------------
+# Interpolate position embeddings for high-resolution
+# References:
+# DeiT: https://github.com/facebookresearch/deit
+# --------------------------------------------------------
+def interpolate_pos_embed(model, checkpoint_model):
+    if 'pos_embed' in checkpoint_model:
+        pos_embed_checkpoint = checkpoint_model['pos_embed']
+        embedding_size = pos_embed_checkpoint.shape[-1]
+        num_patches = model.patch_embed.num_patches
+        num_extra_tokens = model.pos_embed.shape[-2] - num_patches
+        # height (== width) for the checkpoint position embedding
+        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)
+        # height (== width) for the new position embedding
+        new_size = int(num_patches ** 0.5)
+        # class_token and dist_token are kept unchanged
+        if orig_size != new_size:
+            print("Position interpolate from %dx%d to %dx%d" % (orig_size, orig_size, new_size, new_size))
+            extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]
+            # only the position tokens are interpolated
+            pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:]
+            pos_tokens = pos_tokens.reshape(-1, orig_size, orig_size, embedding_size).permute(0, 3, 1, 2)
+            pos_tokens = torch.nn.functional.interpolate(
+                pos_tokens, size=(new_size, new_size), mode='bicubic', align_corners=False)
+            pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)
+            new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)
+            checkpoint_model['pos_embed'] = new_pos_embed

evalkit_internvl/lib/python3.10/site-packages/open_clip/transformer.py

@@ -0,0 +1,912 @@
+from collections import OrderedDict
+import math
+from typing import Callable, List, Optional, Sequence, Tuple, Union
+from functools import partial
+
+import torch
+from torch import nn
+from torch.nn import functional as F
+from torch.utils.checkpoint import checkpoint
+
+from .utils import to_2tuple
+from .pos_embed import get_2d_sincos_pos_embed
+
+
+class LayerNormFp32(nn.LayerNorm):
+    """Subclass torch's LayerNorm to handle fp16 (by casting to float32 and back)."""
+
+    def forward(self, x: torch.Tensor):
+        orig_type = x.dtype
+        x = F.layer_norm(x.to(torch.float32), self.normalized_shape, self.weight, self.bias, self.eps)
+        return x.to(orig_type)
+
+
+class LayerNorm(nn.LayerNorm):
+    """Subclass torch's LayerNorm (with cast back to input dtype)."""
+
+    def forward(self, x: torch.Tensor):
+        orig_type = x.dtype
+        x = F.layer_norm(x, self.normalized_shape, self.weight, self.bias, self.eps)
+        return x.to(orig_type)
+
+
+class QuickGELU(nn.Module):
+    # NOTE This is slower than nn.GELU or nn.SiLU and uses more GPU memory
+    def forward(self, x: torch.Tensor):
+        return x * torch.sigmoid(1.702 * x)
+
+
+class LayerScale(nn.Module):
+    def __init__(self, dim, init_values=1e-5, inplace=False):
+        super().__init__()
+        self.inplace = inplace
+        self.gamma = nn.Parameter(init_values * torch.ones(dim))
+
+    def forward(self, x):
+        return x.mul_(self.gamma) if self.inplace else x * self.gamma
+
+
+class PatchDropout(nn.Module):
+    """
+    https://arxiv.org/abs/2212.00794
+    """
+
+    def __init__(self, prob, exclude_first_token=True):
+        super().__init__()
+        assert 0 <= prob < 1.
+        self.prob = prob
+        self.exclude_first_token = exclude_first_token  # exclude CLS token
+
+    def forward(self, x):
+        if not self.training or self.prob == 0.:
+            return x
+
+        if self.exclude_first_token:
+            cls_tokens, x = x[:, :1], x[:, 1:]
+        else:
+            cls_tokens = torch.jit.annotate(torch.Tensor, x[:, :1])
+
+        batch = x.size()[0]
+        num_tokens = x.size()[1]
+
+        batch_indices = torch.arange(batch)
+        batch_indices = batch_indices[..., None]
+
+        keep_prob = 1 - self.prob
+        num_patches_keep = max(1, int(num_tokens * keep_prob))
+
+        rand = torch.randn(batch, num_tokens)
+        patch_indices_keep = rand.topk(num_patches_keep, dim=-1).indices
+
+        x = x[batch_indices, patch_indices_keep]
+
+        if self.exclude_first_token:
+            x = torch.cat((cls_tokens, x), dim=1)
+
+        return x
+
+
+class Attention(nn.Module):
+    def __init__(
+            self,
+            dim: int,
+            num_heads: int = 8,
+            qkv_bias: bool = True,
+            scaled_cosine: bool = False,
+            scale_heads: bool = False,
+            logit_scale_max: float = math.log(1. / 0.01),
+            batch_first: bool = True,
+            attn_drop: float = 0.,
+            proj_drop: float = 0.
+    ):
+        super().__init__()
+        self.scaled_cosine = scaled_cosine
+        self.scale_heads = scale_heads
+        assert dim % num_heads == 0, 'dim should be divisible by num_heads'
+        self.num_heads = num_heads
+        self.head_dim = dim // num_heads
+        self.scale = self.head_dim ** -0.5
+        self.logit_scale_max = logit_scale_max
+        self.batch_first = batch_first
+        self.use_fsdpa = hasattr(nn.functional, 'scaled_dot_product_attention')
+
+        # keeping in_proj in this form (instead of nn.Linear) to match weight scheme of original
+        self.in_proj_weight = nn.Parameter(torch.randn((dim * 3, dim)) * self.scale)
+        if qkv_bias:
+            self.in_proj_bias = nn.Parameter(torch.zeros(dim * 3))
+        else:
+            self.in_proj_bias = None
+
+        if self.scaled_cosine:
+            self.logit_scale = nn.Parameter(torch.log(10 * torch.ones((num_heads, 1, 1))))
+        else:
+            self.logit_scale = None
+        self.attn_drop = nn.Dropout(attn_drop)
+        if self.scale_heads:
+            self.head_scale = nn.Parameter(torch.ones((num_heads, 1, 1)))
+        else:
+            self.head_scale = None
+        self.out_proj = nn.Linear(dim, dim)
+        self.out_drop = nn.Dropout(proj_drop)
+
+    def forward(self, x, attn_mask: Optional[torch.Tensor] = None):
+        if self.batch_first:
+            x = x.transpose(0, 1)
+
+        L, N, C = x.shape
+        q, k, v = F.linear(x, self.in_proj_weight, self.in_proj_bias).chunk(3, dim=-1)
+        q = q.reshape(L, N * self.num_heads, -1).transpose(0, 1)
+        k = k.reshape(L, N * self.num_heads, -1).transpose(0, 1)
+        v = v.reshape(L, N * self.num_heads, -1).transpose(0, 1)
+
+        if attn_mask is not None and attn_mask.dtype == torch.bool:
+            new_attn_mask = torch.zeros_like(attn_mask, dtype=q.dtype)
+            new_attn_mask.masked_fill_(attn_mask, float("-inf"))
+            attn_mask = new_attn_mask
+
+        if self.logit_scale is not None:
+            attn = torch.bmm(F.normalize(q, dim=-1), F.normalize(k, dim=-1).transpose(-1, -2))
+            logit_scale = torch.clamp(self.logit_scale, max=self.logit_scale_max).exp()
+            attn = attn.view(N, self.num_heads, L, L) * logit_scale
+            attn = attn.view(-1, L, L)
+            if attn_mask is not None:
+                attn = attn + attn_mask
+            attn = attn.softmax(dim=-1)
+            attn = self.attn_drop(attn)
+            x = torch.bmm(attn, v)
+        else:
+            if self.use_fsdpa:
+                x = F.scaled_dot_product_attention(
+                    q, k, v,
+                    attn_mask=attn_mask,
+                    dropout_p=self.attn_drop.p if self.training else 0.,
+                )
+            else:
+                q = q * self.scale
+                attn = torch.bmm(q, k.transpose(-1, -2))
+                if attn_mask is not None:
+                    attn += attn_mask
+                attn = attn.softmax(dim=-1)
+                attn = self.attn_drop(attn)
+                x = torch.bmm(attn, v)
+
+        if self.head_scale is not None:
+            x = x.view(N, self.num_heads, L, C) * self.head_scale
+            x = x.view(-1, L, C)
+
+        x = x.transpose(0, 1).reshape(L, N, C)
+
+        if self.batch_first:
+            x = x.transpose(0, 1)
+
+        x = self.out_proj(x)
+        x = self.out_drop(x)
+        return x
+
+
+class AttentionalPooler(nn.Module):
+    def __init__(
+            self,
+            d_model: int,
+            context_dim: int,
+            n_head: int = 8,
+            n_queries: int = 256,
+            norm_layer: Callable = LayerNorm,
+    ):
+        super().__init__()
+        self.query = nn.Parameter(torch.randn(n_queries, d_model))
+        self.attn = nn.MultiheadAttention(d_model, n_head, kdim=context_dim, vdim=context_dim, batch_first=True)
+        self.ln_q = norm_layer(d_model)
+        self.ln_k = norm_layer(context_dim)
+
+    def forward(self, x: torch.Tensor):
+        N = x.shape[0]
+        x = self.ln_k(x)
+        q = self.ln_q(self.query)
+        out = self.attn(q.unsqueeze(0).expand(N, -1, -1), x, x, need_weights=False)[0]
+        return out
+
+
+class ResidualAttentionBlock(nn.Module):
+    def __init__(
+            self,
+            d_model: int,
+            n_head: int,
+            mlp_ratio: float = 4.0,
+            ls_init_value: float = None,
+            act_layer: Callable = nn.GELU,
+            norm_layer: Callable = LayerNorm,
+            is_cross_attention: bool = False,
+            batch_first: bool = True,
+    ):
+        super().__init__()
+
+        self.ln_1 = norm_layer(d_model)
+        self.attn = nn.MultiheadAttention(d_model, n_head, batch_first=batch_first)
+        self.ls_1 = LayerScale(d_model, ls_init_value) if ls_init_value is not None else nn.Identity()
+        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.mlp = nn.Sequential(OrderedDict([
+            ("c_fc", nn.Linear(d_model, mlp_width)),
+            ("gelu", act_layer()),
+            ("c_proj", nn.Linear(mlp_width, d_model))
+        ]))
+        self.ls_2 = LayerScale(d_model, ls_init_value) if ls_init_value is not None else nn.Identity()
+
+    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, need_weights=False, attn_mask=attn_mask
+        )[0]
+
+    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.ls_1(self.attention(q_x=self.ln_1(q_x), k_x=k_x, v_x=v_x, attn_mask=attn_mask))
+        x = x + self.ls_2(self.mlp(self.ln_2(x)))
+        return x
+
+
+class CustomResidualAttentionBlock(nn.Module):
+    def __init__(
+            self,
+            d_model: int,
+            n_head: int,
+            mlp_ratio: float = 4.0,
+            ls_init_value: float = None,
+            act_layer: Callable = nn.GELU,
+            norm_layer: Callable = LayerNorm,
+            scale_cosine_attn: bool = False,
+            scale_heads: bool = False,
+            scale_attn: bool = False,
+            scale_fc: bool = False,
+            batch_first: bool = True,
+    ):
+        super().__init__()
+
+        self.ln_1 = norm_layer(d_model)
+        self.attn = Attention(
+            d_model,
+            n_head,
+            scaled_cosine=scale_cosine_attn,
+            scale_heads=scale_heads,
+            batch_first=batch_first,
+        )
+        self.ln_attn = norm_layer(d_model) if scale_attn else nn.Identity()
+        self.ls_1 = LayerScale(d_model, ls_init_value) if ls_init_value is not None else nn.Identity()
+
+        self.ln_2 = norm_layer(d_model)
+        mlp_width = int(d_model * mlp_ratio)
+        self.mlp = nn.Sequential(OrderedDict([
+            ("c_fc", nn.Linear(d_model, mlp_width)),
+            ("gelu", act_layer()),
+            ('ln', norm_layer(mlp_width) if scale_fc else nn.Identity()),
+            ("c_proj", nn.Linear(mlp_width, d_model))
+        ]))
+        self.ls_2 = LayerScale(d_model, ls_init_value) if ls_init_value is not None else nn.Identity()
+
+    def get_reference_weight(self):
+        return self.mlp.c_fc.weight
+
+    def forward(self, x: torch.Tensor, attn_mask: Optional[torch.Tensor] = None):
+        x = x + self.ls_1(self.ln_attn(self.attn(self.ln_1(x), attn_mask=attn_mask)))
+        x = x + self.ls_2(self.mlp(self.ln_2(x)))
+        return x
+
+
+def _expand_token(token, batch_size: int):
+    return token.view(1, 1, -1).expand(batch_size, -1, -1)
+
+
+class Transformer(nn.Module):
+    def __init__(
+            self,
+            width: int,
+            layers: int,
+            heads: int,
+            mlp_ratio: float = 4.0,
+            ls_init_value: float = None,
+            act_layer: Callable = nn.GELU,
+            norm_layer: Callable = LayerNorm,
+            batch_first: bool = True,
+    ):
+        super().__init__()
+        self.width = width
+        self.layers = layers
+        self.batch_first = batch_first
+        self.grad_checkpointing = False
+
+        self.resblocks = nn.ModuleList([
+            ResidualAttentionBlock(
+                width,
+                heads,
+                mlp_ratio,
+                ls_init_value=ls_init_value,
+                act_layer=act_layer,
+                norm_layer=norm_layer,
+                batch_first=batch_first,
+            )
+            for _ in range(layers)
+        ])
+
+    def get_cast_dtype(self) -> torch.dtype:
+        if hasattr(self.resblocks[0].mlp.c_fc, 'int8_original_dtype'):
+            return self.resblocks[0].mlp.c_fc.int8_original_dtype
+        return self.resblocks[0].mlp.c_fc.weight.dtype
+
+    def forward(self, x: torch.Tensor, attn_mask: Optional[torch.Tensor] = None):
+        if not self.batch_first:
+            x = x.transpose(0, 1).contiguous()    # NLD -> LND
+        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)
+        if not self.batch_first:
+            x = x.transpose(0, 1)    # LND -> NLD
+        return x
+
+
+class CustomTransformer(nn.Module):
+    """ A custom transformer that can use different block types. """
+    def __init__(
+            self,
+            width: int,
+            layers: int,
+            heads: int,
+            mlp_ratio: float = 4.0,
+            ls_init_value: float = None,
+            act_layer: Callable = nn.GELU,
+            norm_layer: Callable = LayerNorm,
+            batch_first: bool = True,
+            block_types: Union[str, List[str]] = 'CustomResidualAttentionBlock',
+    ):
+        super().__init__()
+        self.width = width
+        self.layers = layers
+        self.batch_first = batch_first  # run trasnformer stack in batch first (N, L, D)
+        self.grad_checkpointing = False
+
+        if isinstance(block_types, str):
+            block_types = [block_types] * layers
+        assert len(block_types) == layers
+
+        def _create_block(bt: str):
+            if bt == 'CustomResidualAttentionBlock':
+                return CustomResidualAttentionBlock(
+                    width,
+                    heads,
+                    mlp_ratio=mlp_ratio,
+                    ls_init_value=ls_init_value,
+                    act_layer=act_layer,
+                    norm_layer=norm_layer,
+                    batch_first=batch_first,
+                )
+            else:
+                assert False
+
+        self.resblocks = nn.ModuleList([
+            _create_block(bt)
+            for bt in block_types
+        ])
+
+    def get_cast_dtype(self) -> torch.dtype:
+        weight = self.resblocks[0].get_reference_weight()
+        if hasattr(weight, 'int8_original_dtype'):
+            return weight.int8_original_dtype
+        return weight.dtype
+
+    def forward(self, x: torch.Tensor, attn_mask: Optional[torch.Tensor] = None):
+        if not self.batch_first:
+            x = x.transpose(0, 1)  # NLD -> LND
+
+        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)
+
+        if not self.batch_first:
+            x = x.transpose(0, 1)  # NLD -> LND
+        return x
+
+
+class VisionTransformer(nn.Module):
+    output_tokens: torch.jit.Final[bool]
+
+    def __init__(
+            self,
+            image_size: int,
+            patch_size: int,
+            width: int,
+            layers: int,
+            heads: int,
+            mlp_ratio: float,
+            ls_init_value: float = None,
+            attentional_pool: bool = False,
+            attn_pooler_queries: int = 256,
+            attn_pooler_heads: int = 8,
+            output_dim: int = 512,
+            patch_dropout: float = 0.,
+            no_ln_pre: bool = False,
+            pos_embed_type: str = 'learnable',
+            pool_type: str = 'tok',
+            final_ln_after_pool: bool = False,
+            act_layer: Callable = nn.GELU,
+            norm_layer: Callable = LayerNorm,
+            output_tokens: bool = False,
+    ):
+        super().__init__()
+        assert pool_type in ('tok', 'avg', 'none')
+        self.output_tokens = output_tokens
+        image_height, image_width = self.image_size = to_2tuple(image_size)
+        patch_height, patch_width = self.patch_size = to_2tuple(patch_size)
+        self.grid_size = (image_height // patch_height, image_width // patch_width)
+        self.final_ln_after_pool = final_ln_after_pool  # currently ignored w/ attn pool enabled
+        self.output_dim = output_dim
+
+        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.class_embedding = nn.Parameter(scale * torch.randn(width))
+        if pos_embed_type == 'learnable':
+            self.positional_embedding = nn.Parameter(
+                scale * torch.randn(self.grid_size[0] * self.grid_size[1] + 1, width))
+        elif pos_embed_type == 'sin_cos_2d':
+            # fixed sin-cos embedding
+            assert self.grid_size[0] == self.grid_size[1],\
+                'currently sin cos 2d pos embedding only supports square input'
+            self.positional_embedding = nn.Parameter(
+                torch.zeros(self.grid_size[0] * self.grid_size[1] + 1, width), requires_grad=False)
+            pos_embed_type = get_2d_sincos_pos_embed(width, self.grid_size[0], cls_token=True)
+            self.positional_embedding.data.copy_(torch.from_numpy(pos_embed_type).float())
+        else:
+            raise ValueError
+
+        # setting a patch_dropout of 0. would mean it is disabled and this function would be the identity fn
+        self.patch_dropout = PatchDropout(patch_dropout) if patch_dropout > 0. else nn.Identity()
+
+        self.ln_pre = nn.Identity() if no_ln_pre else norm_layer(width)
+        self.transformer = Transformer(
+            width,
+            layers,
+            heads,
+            mlp_ratio,
+            ls_init_value=ls_init_value,
+            act_layer=act_layer,
+            norm_layer=norm_layer,
+        )
+
+        if attentional_pool:
+            if isinstance(attentional_pool, str):
+                self.attn_pool_type = attentional_pool
+                self.pool_type = 'none'
+                if attentional_pool in ('parallel', 'cascade'):
+                    self.attn_pool = AttentionalPooler(
+                        output_dim,
+                        width,
+                        n_head=attn_pooler_heads,
+                        n_queries=attn_pooler_queries,
+                    )
+                    self.attn_pool_contrastive = AttentionalPooler(
+                        output_dim,
+                        width,
+                        n_head=attn_pooler_heads,
+                        n_queries=1,
+                    )
+                else:
+                    assert False
+            else:
+                self.attn_pool_type = ''
+                self.pool_type = pool_type
+                self.attn_pool = AttentionalPooler(
+                    output_dim,
+                    width,
+                    n_head=attn_pooler_heads,
+                    n_queries=attn_pooler_queries,
+                )
+                self.attn_pool_contrastive = None
+            pool_dim = output_dim
+        else:
+            self.attn_pool = None
+            pool_dim = width
+            self.pool_type = pool_type
+
+        self.ln_post = norm_layer(pool_dim)
+        self.proj = nn.Parameter(scale * torch.randn(pool_dim, output_dim))
+
+        self.init_parameters()
+
+    def lock(self, unlocked_groups=0, freeze_bn_stats=False):
+        for param in self.parameters():
+            param.requires_grad = False
+
+        if unlocked_groups != 0:
+            groups = [
+                [
+                    self.conv1,
+                    self.class_embedding,
+                    self.positional_embedding,
+                    self.ln_pre,
+                ],
+                *self.transformer.resblocks[:-1],
+                [
+                    self.transformer.resblocks[-1],
+                    self.ln_post,
+                ],
+                self.proj,
+            ]
+
+            def _unlock(x):
+                if isinstance(x, Sequence):
+                    for g in x:
+                        _unlock(g)
+                else:
+                    if isinstance(x, torch.nn.Parameter):
+                        x.requires_grad = True
+                    else:
+                        for p in x.parameters():
+                            p.requires_grad = True
+
+            _unlock(groups[-unlocked_groups:])
+
+    def init_parameters(self):
+        # FIXME OpenAI CLIP did not define an init for the VisualTransformer
+        # TODO experiment if default PyTorch init, below, or alternate init is best.
+
+        # nn.init.normal_(self.class_embedding, std=self.scale)
+        # nn.init.normal_(self.positional_embedding, std=self.scale)
+        #
+        # proj_std = (self.transformer.width ** -0.5) * ((2 * self.transformer.layers) ** -0.5)
+        # attn_std = self.transformer.width ** -0.5
+        # fc_std = (2 * self.transformer.width) ** -0.5
+        # for block in self.transformer.resblocks:
+        #     nn.init.normal_(block.attn.in_proj_weight, std=attn_std)
+        #     nn.init.normal_(block.attn.out_proj.weight, std=proj_std)
+        #     nn.init.normal_(block.mlp.c_fc.weight, std=fc_std)
+        #     nn.init.normal_(block.mlp.c_proj.weight, std=proj_std)
+        #
+        # if self.text_projection is not None:
+        #     nn.init.normal_(self.text_projection, std=self.scale)
+        pass
+
+    @torch.jit.ignore
+    def set_grad_checkpointing(self, enable=True):
+        self.transformer.grad_checkpointing = enable
+
+    def _global_pool(self, x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
+        if self.pool_type == 'avg':
+            pooled, tokens = x[:, 1:].mean(dim=1), x[:, 1:]
+        elif self.pool_type == 'tok':
+            pooled, tokens = x[:, 0], x[:, 1:]
+        else:
+            pooled = tokens = x
+
+        return pooled, tokens
+
+    def forward(self, x: torch.Tensor):
+        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]
+
+        # class embeddings and positional embeddings
+        x = torch.cat([_expand_token(self.class_embedding, x.shape[0]).to(x.dtype), x], dim=1)
+        # shape = [*, grid ** 2 + 1, width]
+        x = x + self.positional_embedding.to(x.dtype)
+
+        x = self.patch_dropout(x)
+        x = self.ln_pre(x)
+        x = self.transformer(x)
+
+        if self.attn_pool is not None:
+            if self.attn_pool_contrastive is not None:
+                # This is untested, WIP pooling that should match paper
+                x = self.ln_post(x)  # TBD LN first or separate one after each pool?
+                tokens = self.attn_pool(x)
+                if self.attn_pool_type == 'parallel':
+                    pooled = self.attn_pool_contrastive(x)
+                else:
+                    assert self.attn_pool_type == 'cascade'
+                    pooled = self.attn_pool_contrastive(tokens)
+            else:
+                # this is the original OpenCLIP CoCa setup, does not match paper
+                x = self.attn_pool(x)
+                x = self.ln_post(x)
+                pooled, tokens = self._global_pool(x)
+        elif self.final_ln_after_pool:
+            pooled, tokens = self._global_pool(x)
+            pooled = self.ln_post(pooled)
+        else:
+            x = self.ln_post(x)
+            pooled, tokens = self._global_pool(x)
+
+        if self.proj is not None:
+            pooled = pooled @ self.proj
+
+        if self.output_tokens:
+            return pooled, tokens
+        
+        return pooled
+
+
+def text_global_pool(x, text: Optional[torch.Tensor] = None, pool_type: str = 'argmax'):
+    if pool_type == 'first':
+        pooled, tokens = x[:, 0], x[:, 1:]
+    elif pool_type == 'last':
+        pooled, tokens = x[:, -1], x[:, :-1]
+    elif pool_type == 'argmax':
+        # take features from the eot embedding (eot_token is the highest number in each sequence)
+        assert text is not None
+        pooled, tokens = x[torch.arange(x.shape[0]), text.argmax(dim=-1)], x
+    else:
+        pooled = tokens = x
+
+    return pooled, tokens
+
+
+class TextTransformer(nn.Module):
+    output_tokens: torch.jit.Final[bool]
+
+    def __init__(
+            self,
+            context_length: int = 77,
+            vocab_size: int = 49408,
+            width: int = 512,
+            heads: int = 8,
+            layers: int = 12,
+            mlp_ratio: float = 4.0,
+            ls_init_value: float = None,
+            output_dim: Optional[int] = 512,
+            embed_cls: bool = False,
+            no_causal_mask: bool = False,
+            pad_id: int = 0,
+            pool_type: str = 'argmax',
+            proj_type: str = 'linear',
+            proj_bias: bool = False,
+            act_layer: Callable = nn.GELU,
+            norm_layer: Callable = LayerNorm,
+            output_tokens: bool = False,
+    ):
+        super().__init__()
+        assert pool_type in ('first', 'last', 'argmax', 'none')
+        self.output_tokens = output_tokens
+        self.num_pos = self.context_length = context_length
+        self.vocab_size = vocab_size
+        self.width = width
+        self.output_dim = output_dim
+        self.heads = heads
+        self.pad_id = pad_id
+        self.pool_type = pool_type
+
+        self.token_embedding = nn.Embedding(vocab_size, width)
+        if embed_cls:
+            self.cls_emb = nn.Parameter(torch.empty(width))
+            self.num_pos += 1
+        else:
+            self.cls_emb = None
+        self.positional_embedding = nn.Parameter(torch.empty(self.num_pos, width))
+        self.transformer = Transformer(
+            width=width,
+            layers=layers,
+            heads=heads,
+            mlp_ratio=mlp_ratio,
+            ls_init_value=ls_init_value,
+            act_layer=act_layer,
+            norm_layer=norm_layer,
+        )
+        self.ln_final = norm_layer(width)
+
+        if no_causal_mask:
+            self.attn_mask = None
+        else:
+            self.register_buffer('attn_mask', self.build_causal_mask(), persistent=False)
+
+        if proj_type == 'none' or not output_dim:
+            self.text_projection = None
+        else:
+            if proj_bias:
+                self.text_projection = nn.Linear(width, output_dim)
+            else:
+                self.text_projection = nn.Parameter(torch.empty(width, output_dim))
+
+        self.init_parameters()
+
+    def init_parameters(self):
+        nn.init.normal_(self.token_embedding.weight, std=0.02)
+        nn.init.normal_(self.positional_embedding, std=0.01)
+        if self.cls_emb is not None:
+            nn.init.normal_(self.cls_emb, std=0.01)
+
+        proj_std = (self.transformer.width ** -0.5) * ((2 * self.transformer.layers) ** -0.5)
+        attn_std = self.transformer.width ** -0.5
+        fc_std = (2 * self.transformer.width) ** -0.5
+        for block in self.transformer.resblocks:
+            nn.init.normal_(block.attn.in_proj_weight, std=attn_std)
+            nn.init.normal_(block.attn.out_proj.weight, std=proj_std)
+            nn.init.normal_(block.mlp.c_fc.weight, std=fc_std)
+            nn.init.normal_(block.mlp.c_proj.weight, std=proj_std)
+
+        if self.text_projection is not None:
+            if isinstance(self.text_projection, nn.Linear):
+                nn.init.normal_(self.text_projection.weight, std=self.transformer.width ** -0.5)
+                if self.text_projection.bias is not None:
+                    nn.init.zeros_(self.text_projection.bias)
+            else:
+                nn.init.normal_(self.text_projection, std=self.transformer.width ** -0.5)
+
+    @torch.jit.ignore
+    def set_grad_checkpointing(self, enable=True):
+        self.transformer.grad_checkpointing = enable
+
+    def build_causal_mask(self):
+        # lazily create causal attention mask, with full attention between the tokens
+        # pytorch uses additive attention mask; fill with -inf
+        mask = torch.empty(self.num_pos, self.num_pos)
+        mask.fill_(float("-inf"))
+        mask.triu_(1)  # zero out the lower diagonal
+        return mask
+
+    def build_cls_mask(self, text, cast_dtype: torch.dtype):
+        cls_mask = (text != self.pad_id).unsqueeze(1)
+        cls_mask = F.pad(cls_mask, (1, 0, cls_mask.shape[2], 0), value=True)
+        additive_mask = torch.empty(cls_mask.shape, dtype=cast_dtype, device=cls_mask.device)
+        additive_mask.fill_(0)
+        additive_mask.masked_fill_(~cls_mask, float("-inf"))
+        additive_mask = torch.repeat_interleave(additive_mask, self.heads, 0)
+        return additive_mask
+
+    def forward(self, text):
+        cast_dtype = self.transformer.get_cast_dtype()
+        seq_len = text.shape[1]
+
+        x = self.token_embedding(text).to(cast_dtype)  # [batch_size, n_ctx, d_model]
+        attn_mask = self.attn_mask
+        if self.cls_emb is not None:
+            seq_len += 1
+            x = torch.cat([x, _expand_token(self.cls_emb, x.shape[0])], dim=1)
+            cls_mask = self.build_cls_mask(text, cast_dtype)
+            if attn_mask is not None:
+                attn_mask = attn_mask[None, :seq_len, :seq_len] + cls_mask[:, :seq_len, :seq_len]
+
+        x = x + self.positional_embedding[:seq_len].to(cast_dtype)
+        x = self.transformer(x, attn_mask=attn_mask)
+
+        # x.shape = [batch_size, n_ctx, transformer.width]
+        if self.cls_emb is not None:
+            # presence of appended cls embed (CoCa) overrides pool_type, always take last token
+            pooled, tokens = text_global_pool(x, pool_type='last')
+            pooled = self.ln_final(pooled)  # final LN applied after pooling in this case
+        else:
+            x = self.ln_final(x)
+            pooled, tokens = text_global_pool(x, text, pool_type=self.pool_type)
+
+        if self.text_projection is not None:
+            if isinstance(self.text_projection, nn.Linear):
+                pooled = self.text_projection(pooled)
+            else:
+                pooled = pooled @ self.text_projection
+
+        if self.output_tokens:
+            return pooled, tokens
+
+        return pooled
+
+
+class MultimodalTransformer(Transformer):
+    def __init__(
+            self,
+            width: int,
+            layers: int,
+            heads: int,
+            context_length: int = 77,
+            mlp_ratio: float = 4.0,
+            ls_init_value: float = None,
+            act_layer: Callable = nn.GELU,
+            norm_layer: Callable = LayerNorm,
+            output_dim: int = 512,
+            batch_first: bool = True,
+    ):
+        super().__init__(
+            width=width,
+            layers=layers,
+            heads=heads,
+            mlp_ratio=mlp_ratio,
+            ls_init_value=ls_init_value,
+            act_layer=act_layer,
+            norm_layer=norm_layer,
+            batch_first=batch_first,
+        )
+        self.context_length = context_length
+        self.cross_attn = nn.ModuleList([
+            ResidualAttentionBlock(
+                width,
+                heads,
+                mlp_ratio,
+                ls_init_value=ls_init_value,
+                act_layer=act_layer,
+                norm_layer=norm_layer,
+                is_cross_attention=True,
+                batch_first=batch_first,
+            )
+            for _ in range(layers)
+        ])
+
+        self.register_buffer('attn_mask', self.build_attention_mask(), persistent=False)
+
+        self.ln_final = norm_layer(width)
+        self.text_projection = nn.Parameter(torch.empty(width, output_dim))
+
+    def init_parameters(self):
+        proj_std = (self.transformer.width ** -0.5) * ((2 * self.transformer.layers) ** -0.5)
+        attn_std = self.transformer.width ** -0.5
+        fc_std = (2 * self.transformer.width) ** -0.5
+        for block in self.transformer.resblocks:
+            nn.init.normal_(block.attn.in_proj_weight, std=attn_std)
+            nn.init.normal_(block.attn.out_proj.weight, std=proj_std)
+            nn.init.normal_(block.mlp.c_fc.weight, std=fc_std)
+            nn.init.normal_(block.mlp.c_proj.weight, std=proj_std)
+        for block in self.transformer.cross_attn:
+            nn.init.normal_(block.attn.in_proj_weight, std=attn_std)
+            nn.init.normal_(block.attn.out_proj.weight, std=proj_std)
+            nn.init.normal_(block.mlp.c_fc.weight, std=fc_std)
+            nn.init.normal_(block.mlp.c_proj.weight, std=proj_std)
+
+        if self.text_projection is not None:
+            nn.init.normal_(self.text_projection, std=self.transformer.width ** -0.5)
+
+    def build_attention_mask(self):
+        # lazily create causal attention mask, with full attention between the tokens
+        # pytorch uses additive attention mask; fill with -inf
+        mask = torch.empty(self.context_length, self.context_length)
+        mask.fill_(float("-inf"))
+        mask.triu_(1)  # zero out the lower diagonal
+        return mask
+
+    def forward(self, image_embs, text_embs):
+        seq_len = text_embs.shape[1]
+        if not self.batch_first:
+            image_embs = image_embs.permute(1, 0, 2)  # NLD -> LND
+            text_embs = text_embs.permute(1, 0, 2)  # NLD -> LND
+
+        for resblock, cross_attn in zip(self.resblocks, self.cross_attn):
+            if self.grad_checkpointing and not torch.jit.is_scripting():
+                # TODO: handle kwargs https://github.com/pytorch/pytorch/issues/79887#issuecomment-1161758372
+                text_embs = checkpoint(resblock, text_embs, None, None, self.attn_mask[:seq_len, :seq_len])
+                text_embs = checkpoint(cross_attn, text_embs, image_embs, image_embs, None)
+            else:
+                text_embs = resblock(text_embs, attn_mask=self.attn_mask[:seq_len, :seq_len])
+                text_embs = cross_attn(text_embs, k_x=image_embs, v_x=image_embs)
+
+        if not self.batch_first:
+            text_embs = text_embs.permute(1, 0, 2)  # LND -> NLD
+
+        out = self.ln_final(text_embs)
+        if self.text_projection is not None:
+            out = out @ self.text_projection
+
+        return out
+
+    @torch.jit.ignore
+    def set_grad_checkpointing(self, enable=True):
+        self.grad_checkpointing = enable

evalkit_internvl/lib/python3.10/site-packages/open_clip/utils.py

@@ -0,0 +1,89 @@
+from itertools import repeat
+import collections.abc
+
+import torch
+from torch import nn as nn
+from torchvision.ops.misc import FrozenBatchNorm2d
+
+
+def freeze_batch_norm_2d(module, module_match={}, name=''):
+    """
+    Converts all `BatchNorm2d` and `SyncBatchNorm` layers of provided module into `FrozenBatchNorm2d`. If `module` is
+    itself an instance of either `BatchNorm2d` or `SyncBatchNorm`, it is converted into `FrozenBatchNorm2d` and
+    returned. Otherwise, the module is walked recursively and submodules are converted in place.
+
+    Args:
+        module (torch.nn.Module): Any PyTorch module.
+        module_match (dict): Dictionary of full module names to freeze (all if empty)
+        name (str): Full module name (prefix)
+
+    Returns:
+        torch.nn.Module: Resulting module
+
+    Inspired by https://github.com/pytorch/pytorch/blob/a5895f85be0f10212791145bfedc0261d364f103/torch/nn/modules/batchnorm.py#L762
+    """
+    res = module
+    is_match = True
+    if module_match:
+        is_match = name in module_match
+    if is_match and isinstance(module, (nn.modules.batchnorm.BatchNorm2d, nn.modules.batchnorm.SyncBatchNorm)):
+        res = FrozenBatchNorm2d(module.num_features)
+        res.num_features = module.num_features
+        res.affine = module.affine
+        if module.affine:
+            res.weight.data = module.weight.data.clone().detach()
+            res.bias.data = module.bias.data.clone().detach()
+        res.running_mean.data = module.running_mean.data
+        res.running_var.data = module.running_var.data
+        res.eps = module.eps
+    else:
+        for child_name, child in module.named_children():
+            full_child_name = '.'.join([name, child_name]) if name else child_name
+            new_child = freeze_batch_norm_2d(child, module_match, full_child_name)
+            if new_child is not child:
+                res.add_module(child_name, new_child)
+    return res
+
+
+# From PyTorch internals
+def _ntuple(n):
+    def parse(x):
+        if isinstance(x, collections.abc.Iterable):
+            return x
+        return tuple(repeat(x, n))
+    return parse
+
+
+to_1tuple = _ntuple(1)
+to_2tuple = _ntuple(2)
+to_3tuple = _ntuple(3)
+to_4tuple = _ntuple(4)
+to_ntuple = lambda n, x: _ntuple(n)(x)
+
+# Replaces all linear layers with linear_replacement
+# TODO: add int8 support for other linear layers including attn and convnets
+def replace_linear(model, linear_replacement, include_modules=['c_fc', 'c_proj'], copy_weights=True):
+    for name, module in model.named_children():
+        if len(list(module.children())) > 0:
+            replace_linear(module, linear_replacement, include_modules, copy_weights)
+
+        if isinstance(module, torch.nn.Linear) and name in include_modules:
+            old_module = model._modules[name]
+            model._modules[name] = linear_replacement(
+                module.in_features,
+                module.out_features,
+                module.bias is not None,
+            )
+            if copy_weights:
+                model._modules[name].weight.data.copy_(old_module.weight.data)
+                if model._modules[name].bias is not None:
+                    model._modules[name].bias.data.copy_(old_module.bias)
+
+    return model
+
+def convert_int8_model_to_inference_mode(model):
+    for m in model.modules():
+        if hasattr(m, 'prepare_for_eval'):
+            int8_original_dtype = m.weight.dtype
+            m.prepare_for_eval()
+            m.int8_original_dtype = int8_original_dtype

evalkit_internvl/lib/python3.10/site-packages/open_clip/version.py

@@ -0,0 +1 @@
+__version__ = '2.29.0'

evalkit_internvl/lib/python3.10/site-packages/open_clip/zero_shot_metadata.py

@@ -0,0 +1,266 @@
+
+OPENAI_IMAGENET_TEMPLATES = (
+    lambda c: f'a bad photo of a {c}.',
+    lambda c: f'a photo of many {c}.',
+    lambda c: f'a sculpture of a {c}.',
+    lambda c: f'a photo of the hard to see {c}.',
+    lambda c: f'a low resolution photo of the {c}.',
+    lambda c: f'a rendering of a {c}.',
+    lambda c: f'graffiti of a {c}.',
+    lambda c: f'a bad photo of the {c}.',
+    lambda c: f'a cropped photo of the {c}.',
+    lambda c: f'a tattoo of a {c}.',
+    lambda c: f'the embroidered {c}.',
+    lambda c: f'a photo of a hard to see {c}.',
+    lambda c: f'a bright photo of a {c}.',
+    lambda c: f'a photo of a clean {c}.',
+    lambda c: f'a photo of a dirty {c}.',
+    lambda c: f'a dark photo of the {c}.',
+    lambda c: f'a drawing of a {c}.',
+    lambda c: f'a photo of my {c}.',
+    lambda c: f'the plastic {c}.',
+    lambda c: f'a photo of the cool {c}.',
+    lambda c: f'a close-up photo of a {c}.',
+    lambda c: f'a black and white photo of the {c}.',
+    lambda c: f'a painting of the {c}.',
+    lambda c: f'a painting of a {c}.',
+    lambda c: f'a pixelated photo of the {c}.',
+    lambda c: f'a sculpture of the {c}.',
+    lambda c: f'a bright photo of the {c}.',
+    lambda c: f'a cropped photo of a {c}.',
+    lambda c: f'a plastic {c}.',
+    lambda c: f'a photo of the dirty {c}.',
+    lambda c: f'a jpeg corrupted photo of a {c}.',
+    lambda c: f'a blurry photo of the {c}.',
+    lambda c: f'a photo of the {c}.',
+    lambda c: f'a good photo of the {c}.',
+    lambda c: f'a rendering of the {c}.',
+    lambda c: f'a {c} in a video game.',
+    lambda c: f'a photo of one {c}.',
+    lambda c: f'a doodle of a {c}.',
+    lambda c: f'a close-up photo of the {c}.',
+    lambda c: f'a photo of a {c}.',
+    lambda c: f'the origami {c}.',
+    lambda c: f'the {c} in a video game.',
+    lambda c: f'a sketch of a {c}.',
+    lambda c: f'a doodle of the {c}.',
+    lambda c: f'a origami {c}.',
+    lambda c: f'a low resolution photo of a {c}.',
+    lambda c: f'the toy {c}.',
+    lambda c: f'a rendition of the {c}.',
+    lambda c: f'a photo of the clean {c}.',
+    lambda c: f'a photo of a large {c}.',
+    lambda c: f'a rendition of a {c}.',
+    lambda c: f'a photo of a nice {c}.',
+    lambda c: f'a photo of a weird {c}.',
+    lambda c: f'a blurry photo of a {c}.',
+    lambda c: f'a cartoon {c}.',
+    lambda c: f'art of a {c}.',
+    lambda c: f'a sketch of the {c}.',
+    lambda c: f'a embroidered {c}.',
+    lambda c: f'a pixelated photo of a {c}.',
+    lambda c: f'itap of the {c}.',
+    lambda c: f'a jpeg corrupted photo of the {c}.',
+    lambda c: f'a good photo of a {c}.',
+    lambda c: f'a plushie {c}.',
+    lambda c: f'a photo of the nice {c}.',
+    lambda c: f'a photo of the small {c}.',
+    lambda c: f'a photo of the weird {c}.',
+    lambda c: f'the cartoon {c}.',
+    lambda c: f'art of the {c}.',
+    lambda c: f'a drawing of the {c}.',
+    lambda c: f'a photo of the large {c}.',
+    lambda c: f'a black and white photo of a {c}.',
+    lambda c: f'the plushie {c}.',
+    lambda c: f'a dark photo of a {c}.',
+    lambda c: f'itap of a {c}.',
+    lambda c: f'graffiti of the {c}.',
+    lambda c: f'a toy {c}.',
+    lambda c: f'itap of my {c}.',
+    lambda c: f'a photo of a cool {c}.',
+    lambda c: f'a photo of a small {c}.',
+    lambda c: f'a tattoo of the {c}.',
+)
+
+
+# a much smaller subset of above prompts
+# from https://github.com/openai/CLIP/blob/main/notebooks/Prompt_Engineering_for_ImageNet.ipynb
+SIMPLE_IMAGENET_TEMPLATES = (
+    lambda c: f'itap of a {c}.',
+    lambda c: f'a bad photo of the {c}.',
+    lambda c: f'a origami {c}.',
+    lambda c: f'a photo of the large {c}.',
+    lambda c: f'a {c} in a video game.',
+    lambda c: f'art of the {c}.',
+    lambda c: f'a photo of the small {c}.',
+)
+
+
+IMAGENET_CLASSNAMES = (
+    "tench", "goldfish", "great white shark", "tiger shark", "hammerhead shark", "electric ray",
+    "stingray", "rooster", "hen", "ostrich", "brambling", "goldfinch", "house finch", "junco",
+    "indigo bunting", "American robin", "bulbul", "jay", "magpie", "chickadee", "American dipper",
+    "kite (bird of prey)", "bald eagle", "vulture", "great grey owl", "fire salamander",
+    "smooth newt", "newt", "spotted salamander", "axolotl", "American bullfrog", "tree frog",
+    "tailed frog", "loggerhead sea turtle", "leatherback sea turtle", "mud turtle", "terrapin",
+    "box turtle", "banded gecko", "green iguana", "Carolina anole",
+    "desert grassland whiptail lizard", "agama", "frilled-necked lizard", "alligator lizard",
+    "Gila monster", "European green lizard", "chameleon", "Komodo dragon", "Nile crocodile",
+    "American alligator", "triceratops", "worm snake", "ring-necked snake",
+    "eastern hog-nosed snake", "smooth green snake", "kingsnake", "garter snake", "water snake",
+    "vine snake", "night snake", "boa constrictor", "African rock python", "Indian cobra",
+    "green mamba", "sea snake", "Saharan horned viper", "eastern diamondback rattlesnake",
+    "sidewinder rattlesnake", "trilobite", "harvestman", "scorpion", "yellow garden spider",
+    "barn spider", "European garden spider", "southern black widow", "tarantula", "wolf spider",
+    "tick", "centipede", "black grouse", "ptarmigan", "ruffed grouse", "prairie grouse", "peafowl",
+    "quail", "partridge", "african grey parrot", "macaw", "sulphur-crested cockatoo", "lorikeet",
+    "coucal", "bee eater", "hornbill", "hummingbird", "jacamar", "toucan", "duck",
+    "red-breasted merganser", "goose", "black swan", "tusker", "echidna", "platypus", "wallaby",
+    "koala", "wombat", "jellyfish", "sea anemone", "brain coral", "flatworm", "nematode", "conch",
+    "snail", "slug", "sea slug", "chiton", "chambered nautilus", "Dungeness crab", "rock crab",
+    "fiddler crab", "red king crab", "American lobster", "spiny lobster", "crayfish", "hermit crab",
+    "isopod", "white stork", "black stork", "spoonbill", "flamingo", "little blue heron",
+    "great egret", "bittern bird", "crane bird", "limpkin", "common gallinule", "American coot",
+    "bustard", "ruddy turnstone", "dunlin", "common redshank", "dowitcher", "oystercatcher",
+    "pelican", "king penguin", "albatross", "grey whale", "killer whale", "dugong", "sea lion",
+    "Chihuahua", "Japanese Chin", "Maltese", "Pekingese", "Shih Tzu", "King Charles Spaniel",
+    "Papillon", "toy terrier", "Rhodesian Ridgeback", "Afghan Hound", "Basset Hound", "Beagle",
+    "Bloodhound", "Bluetick Coonhound", "Black and Tan Coonhound", "Treeing Walker Coonhound",
+    "English foxhound", "Redbone Coonhound", "borzoi", "Irish Wolfhound", "Italian Greyhound",
+    "Whippet", "Ibizan Hound", "Norwegian Elkhound", "Otterhound", "Saluki", "Scottish Deerhound",
+    "Weimaraner", "Staffordshire Bull Terrier", "American Staffordshire Terrier",
+    "Bedlington Terrier", "Border Terrier", "Kerry Blue Terrier", "Irish Terrier",
+    "Norfolk Terrier", "Norwich Terrier", "Yorkshire Terrier", "Wire Fox Terrier",
+    "Lakeland Terrier", "Sealyham Terrier", "Airedale Terrier", "Cairn Terrier",
+    "Australian Terrier", "Dandie Dinmont Terrier", "Boston Terrier", "Miniature Schnauzer",
+    "Giant Schnauzer", "Standard Schnauzer", "Scottish Terrier", "Tibetan Terrier",
+    "Australian Silky Terrier", "Soft-coated Wheaten Terrier", "West Highland White Terrier",
+    "Lhasa Apso", "Flat-Coated Retriever", "Curly-coated Retriever", "Golden Retriever",
+    "Labrador Retriever", "Chesapeake Bay Retriever", "German Shorthaired Pointer", "Vizsla",
+    "English Setter", "Irish Setter", "Gordon Setter", "Brittany dog", "Clumber Spaniel",
+    "English Springer Spaniel", "Welsh Springer Spaniel", "Cocker Spaniel", "Sussex Spaniel",
+    "Irish Water Spaniel", "Kuvasz", "Schipperke", "Groenendael dog", "Malinois", "Briard",
+    "Australian Kelpie", "Komondor", "Old English Sheepdog", "Shetland Sheepdog", "collie",
+    "Border Collie", "Bouvier des Flandres dog", "Rottweiler", "German Shepherd Dog", "Dobermann",
+    "Miniature Pinscher", "Greater Swiss Mountain Dog", "Bernese Mountain Dog",
+    "Appenzeller Sennenhund", "Entlebucher Sennenhund", "Boxer", "Bullmastiff", "Tibetan Mastiff",
+    "French Bulldog", "Great Dane", "St. Bernard", "husky", "Alaskan Malamute", "Siberian Husky",
+    "Dalmatian", "Affenpinscher", "Basenji", "pug", "Leonberger", "Newfoundland dog",
+    "Great Pyrenees dog", "Samoyed", "Pomeranian", "Chow Chow", "Keeshond", "brussels griffon",
+    "Pembroke Welsh Corgi", "Cardigan Welsh Corgi", "Toy Poodle", "Miniature Poodle",
+    "Standard Poodle", "Mexican hairless dog (xoloitzcuintli)", "grey wolf", "Alaskan tundra wolf",
+    "red wolf or maned wolf", "coyote", "dingo", "dhole", "African wild dog", "hyena", "red fox",
+    "kit fox", "Arctic fox", "grey fox", "tabby cat", "tiger cat", "Persian cat", "Siamese cat",
+    "Egyptian Mau", "cougar", "lynx", "leopard", "snow leopard", "jaguar", "lion", "tiger",
+    "cheetah", "brown bear", "American black bear", "polar bear", "sloth bear", "mongoose",
+    "meerkat", "tiger beetle", "ladybug", "ground beetle", "longhorn beetle", "leaf beetle",
+    "dung beetle", "rhinoceros beetle", "weevil", "fly", "bee", "ant", "grasshopper",
+    "cricket insect", "stick insect", "cockroach", "praying mantis", "cicada", "leafhopper",
+    "lacewing", "dragonfly", "damselfly", "red admiral butterfly", "ringlet butterfly",
+    "monarch butterfly", "small white butterfly", "sulphur butterfly", "gossamer-winged butterfly",
+    "starfish", "sea urchin", "sea cucumber", "cottontail rabbit", "hare", "Angora rabbit",
+    "hamster", "porcupine", "fox squirrel", "marmot", "beaver", "guinea pig", "common sorrel horse",
+    "zebra", "pig", "wild boar", "warthog", "hippopotamus", "ox", "water buffalo", "bison",
+    "ram (adult male sheep)", "bighorn sheep", "Alpine ibex", "hartebeest", "impala (antelope)",
+    "gazelle", "arabian camel", "llama", "weasel", "mink", "European polecat",
+    "black-footed ferret", "otter", "skunk", "badger", "armadillo", "three-toed sloth", "orangutan",
+    "gorilla", "chimpanzee", "gibbon", "siamang", "guenon", "patas monkey", "baboon", "macaque",
+    "langur", "black-and-white colobus", "proboscis monkey", "marmoset", "white-headed capuchin",
+    "howler monkey", "titi monkey", "Geoffroy's spider monkey", "common squirrel monkey",
+    "ring-tailed lemur", "indri", "Asian elephant", "African bush elephant", "red panda",
+    "giant panda", "snoek fish", "eel", "silver salmon", "rock beauty fish", "clownfish",
+    "sturgeon", "gar fish", "lionfish", "pufferfish", "abacus", "abaya", "academic gown",
+    "accordion", "acoustic guitar", "aircraft carrier", "airliner", "airship", "altar", "ambulance",
+    "amphibious vehicle", "analog clock", "apiary", "apron", "trash can", "assault rifle",
+    "backpack", "bakery", "balance beam", "balloon", "ballpoint pen", "Band-Aid", "banjo",
+    "baluster / handrail", "barbell", "barber chair", "barbershop", "barn", "barometer", "barrel",
+    "wheelbarrow", "baseball", "basketball", "bassinet", "bassoon", "swimming cap", "bath towel",
+    "bathtub", "station wagon", "lighthouse", "beaker", "military hat (bearskin or shako)",
+    "beer bottle", "beer glass", "bell tower", "baby bib", "tandem bicycle", "bikini",
+    "ring binder", "binoculars", "birdhouse", "boathouse", "bobsleigh", "bolo tie", "poke bonnet",
+    "bookcase", "bookstore", "bottle cap", "hunting bow", "bow tie", "brass memorial plaque", "bra",
+    "breakwater", "breastplate", "broom", "bucket", "buckle", "bulletproof vest",
+    "high-speed train", "butcher shop", "taxicab", "cauldron", "candle", "cannon", "canoe",
+    "can opener", "cardigan", "car mirror", "carousel", "tool kit", "cardboard box / carton",
+    "car wheel", "automated teller machine", "cassette", "cassette player", "castle", "catamaran",
+    "CD player", "cello", "mobile phone", "chain", "chain-link fence", "chain mail", "chainsaw",
+    "storage chest", "chiffonier", "bell or wind chime", "china cabinet", "Christmas stocking",
+    "church", "movie theater", "cleaver", "cliff dwelling", "cloak", "clogs", "cocktail shaker",
+    "coffee mug", "coffeemaker", "spiral or coil", "combination lock", "computer keyboard",
+    "candy store", "container ship", "convertible", "corkscrew", "cornet", "cowboy boot",
+    "cowboy hat", "cradle", "construction crane", "crash helmet", "crate", "infant bed",
+    "Crock Pot", "croquet ball", "crutch", "cuirass", "dam", "desk", "desktop computer",
+    "rotary dial telephone", "diaper", "digital clock", "digital watch", "dining table",
+    "dishcloth", "dishwasher", "disc brake", "dock", "dog sled", "dome", "doormat", "drilling rig",
+    "drum", "drumstick", "dumbbell", "Dutch oven", "electric fan", "electric guitar",
+    "electric locomotive", "entertainment center", "envelope", "espresso machine", "face powder",
+    "feather boa", "filing cabinet", "fireboat", "fire truck", "fire screen", "flagpole", "flute",
+    "folding chair", "football helmet", "forklift", "fountain", "fountain pen", "four-poster bed",
+    "freight car", "French horn", "frying pan", "fur coat", "garbage truck",
+    "gas mask or respirator", "gas pump", "goblet", "go-kart", "golf ball", "golf cart", "gondola",
+    "gong", "gown", "grand piano", "greenhouse", "radiator grille", "grocery store", "guillotine",
+    "hair clip", "hair spray", "half-track", "hammer", "hamper", "hair dryer", "hand-held computer",
+    "handkerchief", "hard disk drive", "harmonica", "harp", "combine harvester", "hatchet",
+    "holster", "home theater", "honeycomb", "hook", "hoop skirt", "gymnastic horizontal bar",
+    "horse-drawn vehicle", "hourglass", "iPod", "clothes iron", "carved pumpkin", "jeans", "jeep",
+    "T-shirt", "jigsaw puzzle", "rickshaw", "joystick", "kimono", "knee pad", "knot", "lab coat",
+    "ladle", "lampshade", "laptop computer", "lawn mower", "lens cap", "letter opener", "library",
+    "lifeboat", "lighter", "limousine", "ocean liner", "lipstick", "slip-on shoe", "lotion",
+    "music speaker", "loupe magnifying glass", "sawmill", "magnetic compass", "messenger bag",
+    "mailbox", "tights", "one-piece bathing suit", "manhole cover", "maraca", "marimba", "mask",
+    "matchstick", "maypole", "maze", "measuring cup", "medicine cabinet", "megalith", "microphone",
+    "microwave oven", "military uniform", "milk can", "minibus", "miniskirt", "minivan", "missile",
+    "mitten", "mixing bowl", "mobile home", "ford model t", "modem", "monastery", "monitor",
+    "moped", "mortar and pestle", "graduation cap", "mosque", "mosquito net", "vespa",
+    "mountain bike", "tent", "computer mouse", "mousetrap", "moving van", "muzzle", "metal nail",
+    "neck brace", "necklace", "baby pacifier", "notebook computer", "obelisk", "oboe", "ocarina",
+    "odometer", "oil filter", "pipe organ", "oscilloscope", "overskirt", "bullock cart",
+    "oxygen mask", "product packet / packaging", "paddle", "paddle wheel", "padlock", "paintbrush",
+    "pajamas", "palace", "pan flute", "paper towel", "parachute", "parallel bars", "park bench",
+    "parking meter", "railroad car", "patio", "payphone", "pedestal", "pencil case",
+    "pencil sharpener", "perfume", "Petri dish", "photocopier", "plectrum", "Pickelhaube",
+    "picket fence", "pickup truck", "pier", "piggy bank", "pill bottle", "pillow", "ping-pong ball",
+    "pinwheel", "pirate ship", "drink pitcher", "block plane", "planetarium", "plastic bag",
+    "plate rack", "farm plow", "plunger", "Polaroid camera", "pole", "police van", "poncho",
+    "pool table", "soda bottle", "plant pot", "potter's wheel", "power drill", "prayer rug",
+    "printer", "prison", "missile", "projector", "hockey puck", "punching bag", "purse", "quill",
+    "quilt", "race car", "racket", "radiator", "radio", "radio telescope", "rain barrel",
+    "recreational vehicle", "fishing casting reel", "reflex camera", "refrigerator",
+    "remote control", "restaurant", "revolver", "rifle", "rocking chair", "rotisserie", "eraser",
+    "rugby ball", "ruler measuring stick", "sneaker", "safe", "safety pin", "salt shaker", "sandal",
+    "sarong", "saxophone", "scabbard", "weighing scale", "school bus", "schooner", "scoreboard",
+    "CRT monitor", "screw", "screwdriver", "seat belt", "sewing machine", "shield", "shoe store",
+    "shoji screen / room divider", "shopping basket", "shopping cart", "shovel", "shower cap",
+    "shower curtain", "ski", "balaclava ski mask", "sleeping bag", "slide rule", "sliding door",
+    "slot machine", "snorkel", "snowmobile", "snowplow", "soap dispenser", "soccer ball", "sock",
+    "solar thermal collector", "sombrero", "soup bowl", "keyboard space bar", "space heater",
+    "space shuttle", "spatula", "motorboat", "spider web", "spindle", "sports car", "spotlight",
+    "stage", "steam locomotive", "through arch bridge", "steel drum", "stethoscope", "scarf",
+    "stone wall", "stopwatch", "stove", "strainer", "tram", "stretcher", "couch", "stupa",
+    "submarine", "suit", "sundial", "sunglasses", "sunglasses", "sunscreen", "suspension bridge",
+    "mop", "sweatshirt", "swim trunks / shorts", "swing", "electrical switch", "syringe",
+    "table lamp", "tank", "tape player", "teapot", "teddy bear", "television", "tennis ball",
+    "thatched roof", "front curtain", "thimble", "threshing machine", "throne", "tile roof",
+    "toaster", "tobacco shop", "toilet seat", "torch", "totem pole", "tow truck", "toy store",
+    "tractor", "semi-trailer truck", "tray", "trench coat", "tricycle", "trimaran", "tripod",
+    "triumphal arch", "trolleybus", "trombone", "hot tub", "turnstile", "typewriter keyboard",
+    "umbrella", "unicycle", "upright piano", "vacuum cleaner", "vase", "vaulted or arched ceiling",
+    "velvet fabric", "vending machine", "vestment", "viaduct", "violin", "volleyball",
+    "waffle iron", "wall clock", "wallet", "wardrobe", "military aircraft", "sink",
+    "washing machine", "water bottle", "water jug", "water tower", "whiskey jug", "whistle",
+    "hair wig", "window screen", "window shade", "Windsor tie", "wine bottle", "airplane wing",
+    "wok", "wooden spoon", "wool", "split-rail fence", "shipwreck", "sailboat", "yurt", "website",
+    "comic book", "crossword", "traffic or street sign", "traffic light", "dust jacket", "menu",
+    "plate", "guacamole", "consomme", "hot pot", "trifle", "ice cream", "popsicle", "baguette",
+    "bagel", "pretzel", "cheeseburger", "hot dog", "mashed potatoes", "cabbage", "broccoli",
+    "cauliflower", "zucchini", "spaghetti squash", "acorn squash", "butternut squash", "cucumber",
+    "artichoke", "bell pepper", "cardoon", "mushroom", "Granny Smith apple", "strawberry", "orange",
+    "lemon", "fig", "pineapple", "banana", "jackfruit", "cherimoya (custard apple)", "pomegranate",
+    "hay", "carbonara", "chocolate syrup", "dough", "meatloaf", "pizza", "pot pie", "burrito",
+    "red wine", "espresso", "tea cup", "eggnog", "mountain", "bubble", "cliff", "coral reef",
+    "geyser", "lakeshore", "promontory", "sandbar", "beach", "valley", "volcano", "baseball player",
+    "bridegroom", "scuba diver", "rapeseed", "daisy", "yellow lady's slipper", "corn", "acorn",
+    "rose hip", "horse chestnut seed", "coral fungus", "agaric", "gyromitra", "stinkhorn mushroom",
+    "earth star fungus", "hen of the woods mushroom", "bolete", "corn cob", "toilet paper"
+)
+

evalkit_internvl/lib/python3.10/site-packages/openai/types/fine_tuning/jobs/fine_tuning_job_checkpoint.py

@@ -0,0 +1,47 @@
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Optional
+from typing_extensions import Literal
+
+from ...._models import BaseModel
+
+__all__ = ["FineTuningJobCheckpoint", "Metrics"]
+
+
+class Metrics(BaseModel):
+    full_valid_loss: Optional[float] = None
+
+    full_valid_mean_token_accuracy: Optional[float] = None
+
+    step: Optional[float] = None
+
+    train_loss: Optional[float] = None
+
+    train_mean_token_accuracy: Optional[float] = None
+
+    valid_loss: Optional[float] = None
+
+    valid_mean_token_accuracy: Optional[float] = None
+
+
+class FineTuningJobCheckpoint(BaseModel):
+    id: str
+    """The checkpoint identifier, which can be referenced in the API endpoints."""
+
+    created_at: int
+    """The Unix timestamp (in seconds) for when the checkpoint was created."""
+
+    fine_tuned_model_checkpoint: str
+    """The name of the fine-tuned checkpoint model that is created."""
+
+    fine_tuning_job_id: str
+    """The name of the fine-tuning job that this checkpoint was created from."""
+
+    metrics: Metrics
+    """Metrics at the step number during the fine-tuning job."""
+
+    object: Literal["fine_tuning.job.checkpoint"]
+    """The object type, which is always "fine_tuning.job.checkpoint"."""
+
+    step_number: int
+    """The step number that the checkpoint was created at."""

evalkit_tf437/lib/python3.10/site-packages/decord.libs/libavdevice-bf61e037.so.58.5.100

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06e50e5ae19def6f1c73ffa308361ed4e6665f5cd53056892874848540a2017c
+size 114320

evalkit_tf437/lib/python3.10/site-packages/decord.libs/libavfilter-1e2243e2.so.7.40.101

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79b37a526b50d6ebcd2255983198276718c29c0942d1fde96306e413041e01cb
+size 3075448

evalkit_tf437/lib/python3.10/site-packages/decord.libs/libavformat-8b46ea57.so.58.20.100

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a1eee8f823048e9059d3516d69c57b8a10604f7b722bfaf2327d021b4662518
+size 2399400

evalkit_tf437/lib/python3.10/site-packages/decord.libs/libavutil-2b26904a.so.56.22.100

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f257fbb5b251f0b8580b5d65296a4d391ae28b98a4ed715961db0df2aaa3c7b9
+size 413208

evalkit_tf437/lib/python3.10/site-packages/decord.libs/libpostproc-88b722f8.so.55.3.100

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94f8b2c0b25e7ecce3e6648c20fc25839c0000b07616d797d02983dc75efe9f8
+size 120064

evalkit_tf437/lib/python3.10/site-packages/decord.libs/libswscale-8e37dcfd.so.5.3.100

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc48193958276cf1ae7c1272554710a96271331f8d0b4f31061752caae8c5b13
+size 537808

evalkit_tf437/lib/python3.10/site-packages/decord.libs/libxcb-77222338.so.1.1.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:266a250225f2e7e62c3b21496090ff8bdf9c8ff7555c7fef82b2e06385d36dab
+size 194008

evalkit_tf437/lib/python3.10/site-packages/einops_exts/__init__.py

@@ -0,0 +1,3 @@
+from einops_exts.einops_exts import check_shape
+from einops_exts.einops_exts import rearrange_many, repeat_many, reduce_many
+from einops_exts.einops_exts import rearrange_with_anon_dims, repeat_with_anon_dims, reduce_with_anon_dims

evalkit_tf437/lib/python3.10/site-packages/einops_exts/einops_exts.py

@@ -0,0 +1,63 @@
+import re
+from torch import nn
+from functools import wraps, partial
+
+from einops import rearrange, reduce, repeat
+
+# checking shape
+# @nils-werner
+# https://github.com/arogozhnikov/einops/issues/168#issuecomment-1042933838
+
+def check_shape(tensor, pattern, **kwargs):
+    return rearrange(tensor, f"{pattern} -> {pattern}", **kwargs)
+
+# do same einops operations on a list of tensors
+
+def _many(fn):
+    @wraps(fn)
+    def inner(tensors, pattern, **kwargs):
+        return (fn(tensor, pattern, **kwargs) for tensor in tensors)
+    return inner
+
+# do einops with unflattening of anonymously named dimensions
+# (...flattened) ->  ...flattened
+
+def _with_anon_dims(fn):
+    @wraps(fn)
+    def inner(tensor, pattern, **kwargs):
+        regex = r'(\.\.\.[a-zA-Z]+)'
+        matches = re.findall(regex, pattern)
+        get_anon_dim_name = lambda t: t.lstrip('...')
+        dim_prefixes = tuple(map(get_anon_dim_name, set(matches)))
+
+        update_kwargs_dict = dict()
+
+        for prefix in dim_prefixes:
+            assert prefix in kwargs, f'dimension list "{prefix}" was not passed in'
+            dim_list = kwargs[prefix]
+            assert isinstance(dim_list, (list, tuple)), f'dimension list "{prefix}" needs to be a tuple of list of dimensions'
+            dim_names = list(map(lambda ind: f'{prefix}{ind}', range(len(dim_list))))
+            update_kwargs_dict[prefix] = dict(zip(dim_names, dim_list))
+
+        def sub_with_anonymous_dims(t):
+            dim_name_prefix = get_anon_dim_name(t.groups()[0])
+            return ' '.join(update_kwargs_dict[dim_name_prefix].keys())
+
+        pattern_new = re.sub(regex, sub_with_anonymous_dims, pattern)
+
+        for prefix, update_dict in update_kwargs_dict.items():
+            del kwargs[prefix]
+            kwargs.update(update_dict)
+
+        return fn(tensor, pattern_new, **kwargs)
+    return inner
+
+# generate all helper functions
+
+rearrange_many = _many(rearrange)
+repeat_many = _many(repeat)
+reduce_many = _many(reduce)
+
+rearrange_with_anon_dims = _with_anon_dims(rearrange)
+repeat_with_anon_dims = _with_anon_dims(repeat)
+reduce_with_anon_dims = _with_anon_dims(reduce)

evalkit_tf437/lib/python3.10/site-packages/parso/__init__.py

@@ -0,0 +1,58 @@
+r"""
+Parso is a Python parser that supports error recovery and round-trip parsing
+for different Python versions (in multiple Python versions). Parso is also able
+to list multiple syntax errors in your python file.
+
+Parso has been battle-tested by jedi_. It was pulled out of jedi to be useful
+for other projects as well.
+
+Parso consists of a small API to parse Python and analyse the syntax tree.
+
+.. _jedi: https://github.com/davidhalter/jedi
+
+A simple example:
+
+>>> import parso
+>>> module = parso.parse('hello + 1', version="3.9")
+>>> expr = module.children[0]
+>>> expr
+PythonNode(arith_expr, [<Name: hello@1,0>, <Operator: +>, <Number: 1>])
+>>> print(expr.get_code())
+hello + 1
+>>> name = expr.children[0]
+>>> name
+<Name: hello@1,0>
+>>> name.end_pos
+(1, 5)
+>>> expr.end_pos
+(1, 9)
+
+To list multiple issues:
+
+>>> grammar = parso.load_grammar()
+>>> module = grammar.parse('foo +\nbar\ncontinue')
+>>> error1, error2 = grammar.iter_errors(module)
+>>> error1.message
+'SyntaxError: invalid syntax'
+>>> error2.message
+"SyntaxError: 'continue' not properly in loop"
+"""
+
+from parso.parser import ParserSyntaxError
+from parso.grammar import Grammar, load_grammar
+from parso.utils import split_lines, python_bytes_to_unicode
+
+
+__version__ = '0.8.4'
+
+
+def parse(code=None, **kwargs):
+    """
+    A utility function to avoid loading grammars.
+    Params are documented in :py:meth:`parso.Grammar.parse`.
+
+    :param str version: The version used by :py:func:`parso.load_grammar`.
+    """
+    version = kwargs.pop('version', None)
+    grammar = load_grammar(version=version)
+    return grammar.parse(code, **kwargs)

evalkit_tf437/lib/python3.10/site-packages/parso/_compatibility.py

@@ -0,0 +1,3 @@
+import platform
+
+is_pypy = platform.python_implementation() == 'PyPy'

evalkit_tf437/lib/python3.10/site-packages/parso/cache.py

@@ -0,0 +1,275 @@
+import time
+import os
+import sys
+import hashlib
+import gc
+import shutil
+import platform
+import logging
+import warnings
+import pickle
+from pathlib import Path
+from typing import Dict, Any
+
+LOG = logging.getLogger(__name__)
+
+_CACHED_FILE_MINIMUM_SURVIVAL = 60 * 10  # 10 minutes
+"""
+Cached files should survive at least a few minutes.
+"""
+
+_CACHED_FILE_MAXIMUM_SURVIVAL = 60 * 60 * 24 * 30
+"""
+Maximum time for a cached file to survive if it is not
+accessed within.
+"""
+
+_CACHED_SIZE_TRIGGER = 600
+"""
+This setting limits the amount of cached files. It's basically a way to start
+garbage collection.
+
+The reasoning for this limit being as big as it is, is the following:
+
+Numpy, Pandas, Matplotlib and Tensorflow together use about 500 files. This
+makes Jedi use ~500mb of memory. Since we might want a bit more than those few
+libraries, we just increase it a bit.
+"""
+
+_PICKLE_VERSION = 33
+"""
+Version number (integer) for file system cache.
+
+Increment this number when there are any incompatible changes in
+the parser tree classes.  For example, the following changes
+are regarded as incompatible.
+
+- A class name is changed.
+- A class is moved to another module.
+- A __slot__ of a class is changed.
+"""
+
+_VERSION_TAG = '%s-%s%s-%s' % (
+    platform.python_implementation(),
+    sys.version_info[0],
+    sys.version_info[1],
+    _PICKLE_VERSION
+)
+"""
+Short name for distinguish Python implementations and versions.
+
+It's a bit similar to `sys.implementation.cache_tag`.
+See: http://docs.python.org/3/library/sys.html#sys.implementation
+"""
+
+
+def _get_default_cache_path():
+    if platform.system().lower() == 'windows':
+        dir_ = Path(os.getenv('LOCALAPPDATA') or '~', 'Parso', 'Parso')
+    elif platform.system().lower() == 'darwin':
+        dir_ = Path('~', 'Library', 'Caches', 'Parso')
+    else:
+        dir_ = Path(os.getenv('XDG_CACHE_HOME') or '~/.cache', 'parso')
+    return dir_.expanduser()
+
+
+_default_cache_path = _get_default_cache_path()
+"""
+The path where the cache is stored.
+
+On Linux, this defaults to ``~/.cache/parso/``, on OS X to
+``~/Library/Caches/Parso/`` and on Windows to ``%LOCALAPPDATA%\\Parso\\Parso\\``.
+On Linux, if environment variable ``$XDG_CACHE_HOME`` is set,
+``$XDG_CACHE_HOME/parso`` is used instead of the default one.
+"""
+
+_CACHE_CLEAR_THRESHOLD = 60 * 60 * 24
+
+
+def _get_cache_clear_lock_path(cache_path=None):
+    """
+    The path where the cache lock is stored.
+
+    Cache lock will prevent continous cache clearing and only allow garbage
+    collection once a day (can be configured in _CACHE_CLEAR_THRESHOLD).
+    """
+    cache_path = cache_path or _default_cache_path
+    return cache_path.joinpath("PARSO-CACHE-LOCK")
+
+
+parser_cache: Dict[str, Any] = {}
+
+
+class _NodeCacheItem:
+    def __init__(self, node, lines, change_time=None):
+        self.node = node
+        self.lines = lines
+        if change_time is None:
+            change_time = time.time()
+        self.change_time = change_time
+        self.last_used = change_time
+
+
+def load_module(hashed_grammar, file_io, cache_path=None):
+    """
+    Returns a module or None, if it fails.
+    """
+    p_time = file_io.get_last_modified()
+    if p_time is None:
+        return None
+
+    try:
+        module_cache_item = parser_cache[hashed_grammar][file_io.path]
+        if p_time <= module_cache_item.change_time:
+            module_cache_item.last_used = time.time()
+            return module_cache_item.node
+    except KeyError:
+        return _load_from_file_system(
+            hashed_grammar,
+            file_io.path,
+            p_time,
+            cache_path=cache_path
+        )
+
+
+def _load_from_file_system(hashed_grammar, path, p_time, cache_path=None):
+    cache_path = _get_hashed_path(hashed_grammar, path, cache_path=cache_path)
+    try:
+        if p_time > os.path.getmtime(cache_path):
+            # Cache is outdated
+            return None
+
+        with open(cache_path, 'rb') as f:
+            gc.disable()
+            try:
+                module_cache_item = pickle.load(f)
+            finally:
+                gc.enable()
+    except FileNotFoundError:
+        return None
+    else:
+        _set_cache_item(hashed_grammar, path, module_cache_item)
+        LOG.debug('pickle loaded: %s', path)
+        return module_cache_item.node
+
+
+def _set_cache_item(hashed_grammar, path, module_cache_item):
+    if sum(len(v) for v in parser_cache.values()) >= _CACHED_SIZE_TRIGGER:
+        # Garbage collection of old cache files.
+        # We are basically throwing everything away that hasn't been accessed
+        # in 10 minutes.
+        cutoff_time = time.time() - _CACHED_FILE_MINIMUM_SURVIVAL
+        for key, path_to_item_map in parser_cache.items():
+            parser_cache[key] = {
+                path: node_item
+                for path, node_item in path_to_item_map.items()
+                if node_item.last_used > cutoff_time
+            }
+
+    parser_cache.setdefault(hashed_grammar, {})[path] = module_cache_item
+
+
+def try_to_save_module(hashed_grammar, file_io, module, lines, pickling=True, cache_path=None):
+    path = file_io.path
+    try:
+        p_time = None if path is None else file_io.get_last_modified()
+    except OSError:
+        p_time = None
+        pickling = False
+
+    item = _NodeCacheItem(module, lines, p_time)
+    _set_cache_item(hashed_grammar, path, item)
+    if pickling and path is not None:
+        try:
+            _save_to_file_system(hashed_grammar, path, item, cache_path=cache_path)
+        except PermissionError:
+            # It's not really a big issue if the cache cannot be saved to the
+            # file system. It's still in RAM in that case. However we should
+            # still warn the user that this is happening.
+            warnings.warn(
+                'Tried to save a file to %s, but got permission denied.' % path,
+                Warning
+            )
+        else:
+            _remove_cache_and_update_lock(cache_path=cache_path)
+
+
+def _save_to_file_system(hashed_grammar, path, item, cache_path=None):
+    with open(_get_hashed_path(hashed_grammar, path, cache_path=cache_path), 'wb') as f:
+        pickle.dump(item, f, pickle.HIGHEST_PROTOCOL)
+
+
+def clear_cache(cache_path=None):
+    if cache_path is None:
+        cache_path = _default_cache_path
+    shutil.rmtree(cache_path)
+    parser_cache.clear()
+
+
+def clear_inactive_cache(
+    cache_path=None,
+    inactivity_threshold=_CACHED_FILE_MAXIMUM_SURVIVAL,
+):
+    if cache_path is None:
+        cache_path = _default_cache_path
+    if not cache_path.exists():
+        return False
+    for dirname in os.listdir(cache_path):
+        version_path = cache_path.joinpath(dirname)
+        if not version_path.is_dir():
+            continue
+        for file in os.scandir(version_path):
+            if file.stat().st_atime + _CACHED_FILE_MAXIMUM_SURVIVAL <= time.time():
+                try:
+                    os.remove(file.path)
+                except OSError:  # silently ignore all failures
+                    continue
+    else:
+        return True
+
+
+def _touch(path):
+    try:
+        os.utime(path, None)
+    except FileNotFoundError:
+        try:
+            file = open(path, 'a')
+            file.close()
+        except (OSError, IOError):  # TODO Maybe log this?
+            return False
+    return True
+
+
+def _remove_cache_and_update_lock(cache_path=None):
+    lock_path = _get_cache_clear_lock_path(cache_path=cache_path)
+    try:
+        clear_lock_time = os.path.getmtime(lock_path)
+    except FileNotFoundError:
+        clear_lock_time = None
+    if (
+        clear_lock_time is None  # first time
+        or clear_lock_time + _CACHE_CLEAR_THRESHOLD <= time.time()
+    ):
+        if not _touch(lock_path):
+            # First make sure that as few as possible other cleanup jobs also
+            # get started. There is still a race condition but it's probably
+            # not a big problem.
+            return False
+
+        clear_inactive_cache(cache_path=cache_path)
+
+
+def _get_hashed_path(hashed_grammar, path, cache_path=None):
+    directory = _get_cache_directory_path(cache_path=cache_path)
+
+    file_hash = hashlib.sha256(str(path).encode("utf-8")).hexdigest()
+    return os.path.join(directory, '%s-%s.pkl' % (hashed_grammar, file_hash))
+
+
+def _get_cache_directory_path(cache_path=None):
+    if cache_path is None:
+        cache_path = _default_cache_path
+    directory = cache_path.joinpath(_VERSION_TAG)
+    if not directory.exists():
+        os.makedirs(directory)
+    return directory

evalkit_tf437/lib/python3.10/site-packages/parso/file_io.py

@@ -0,0 +1,38 @@
+import os
+from pathlib import Path
+from typing import Union
+
+
+class FileIO:
+    def __init__(self, path: Union[os.PathLike, str]):
+        if isinstance(path, str):
+            path = Path(path)
+        self.path = path
+
+    def read(self):  # Returns bytes/str
+        # We would like to read unicode here, but we cannot, because we are not
+        # sure if it is a valid unicode file. Therefore just read whatever is
+        # here.
+        with open(self.path, 'rb') as f:
+            return f.read()
+
+    def get_last_modified(self):
+        """
+        Returns float - timestamp or None, if path doesn't exist.
+        """
+        try:
+            return os.path.getmtime(self.path)
+        except FileNotFoundError:
+            return None
+
+    def __repr__(self):
+        return '%s(%s)' % (self.__class__.__name__, self.path)
+
+
+class KnownContentFileIO(FileIO):
+    def __init__(self, path, content):
+        super().__init__(path)
+        self._content = content
+
+    def read(self):
+        return self._content

evalkit_tf437/lib/python3.10/site-packages/parso/grammar.py

@@ -0,0 +1,264 @@
+import hashlib
+import os
+from typing import Generic, TypeVar, Union, Dict, Optional, Any
+from pathlib import Path
+
+from parso._compatibility import is_pypy
+from parso.pgen2 import generate_grammar
+from parso.utils import split_lines, python_bytes_to_unicode, \
+    PythonVersionInfo, parse_version_string
+from parso.python.diff import DiffParser
+from parso.python.tokenize import tokenize_lines, tokenize
+from parso.python.token import PythonTokenTypes
+from parso.cache import parser_cache, load_module, try_to_save_module
+from parso.parser import BaseParser
+from parso.python.parser import Parser as PythonParser
+from parso.python.errors import ErrorFinderConfig
+from parso.python import pep8
+from parso.file_io import FileIO, KnownContentFileIO
+from parso.normalizer import RefactoringNormalizer, NormalizerConfig
+
+_loaded_grammars: Dict[str, 'Grammar'] = {}
+
+_NodeT = TypeVar("_NodeT")
+
+
+class Grammar(Generic[_NodeT]):
+    """
+    :py:func:`parso.load_grammar` returns instances of this class.
+
+    Creating custom none-python grammars by calling this is not supported, yet.
+
+    :param text: A BNF representation of your grammar.
+    """
+    _start_nonterminal: str
+    _error_normalizer_config: Optional[ErrorFinderConfig] = None
+    _token_namespace: Any = None
+    _default_normalizer_config: NormalizerConfig = pep8.PEP8NormalizerConfig()
+
+    def __init__(self, text: str, *, tokenizer, parser=BaseParser, diff_parser=None):
+        self._pgen_grammar = generate_grammar(
+            text,
+            token_namespace=self._get_token_namespace()
+        )
+        self._parser = parser
+        self._tokenizer = tokenizer
+        self._diff_parser = diff_parser
+        self._hashed = hashlib.sha256(text.encode("utf-8")).hexdigest()
+
+    def parse(self,
+              code: Union[str, bytes] = None,
+              *,
+              error_recovery=True,
+              path: Union[os.PathLike, str] = None,
+              start_symbol: str = None,
+              cache=False,
+              diff_cache=False,
+              cache_path: Union[os.PathLike, str] = None,
+              file_io: FileIO = None) -> _NodeT:
+        """
+        If you want to parse a Python file you want to start here, most likely.
+
+        If you need finer grained control over the parsed instance, there will be
+        other ways to access it.
+
+        :param str code: A unicode or bytes string. When it's not possible to
+            decode bytes to a string, returns a
+            :py:class:`UnicodeDecodeError`.
+        :param bool error_recovery: If enabled, any code will be returned. If
+            it is invalid, it will be returned as an error node. If disabled,
+            you will get a ParseError when encountering syntax errors in your
+            code.
+        :param str start_symbol: The grammar rule (nonterminal) that you want
+            to parse. Only allowed to be used when error_recovery is False.
+        :param str path: The path to the file you want to open. Only needed for caching.
+        :param bool cache: Keeps a copy of the parser tree in RAM and on disk
+            if a path is given. Returns the cached trees if the corresponding
+            files on disk have not changed. Note that this stores pickle files
+            on your file system (e.g. for Linux in ``~/.cache/parso/``).
+        :param bool diff_cache: Diffs the cached python module against the new
+            code and tries to parse only the parts that have changed. Returns
+            the same (changed) module that is found in cache. Using this option
+            requires you to not do anything anymore with the cached modules
+            under that path, because the contents of it might change. This
+            option is still somewhat experimental. If you want stability,
+            please don't use it.
+        :param bool cache_path: If given saves the parso cache in this
+            directory. If not given, defaults to the default cache places on
+            each platform.
+
+        :return: A subclass of :py:class:`parso.tree.NodeOrLeaf`. Typically a
+            :py:class:`parso.python.tree.Module`.
+        """
+        if code is None and path is None and file_io is None:
+            raise TypeError("Please provide either code or a path.")
+
+        if isinstance(path, str):
+            path = Path(path)
+        if isinstance(cache_path, str):
+            cache_path = Path(cache_path)
+
+        if start_symbol is None:
+            start_symbol = self._start_nonterminal
+
+        if error_recovery and start_symbol != 'file_input':
+            raise NotImplementedError("This is currently not implemented.")
+
+        if file_io is None:
+            if code is None:
+                file_io = FileIO(path)  # type: ignore[arg-type]
+            else:
+                file_io = KnownContentFileIO(path, code)
+
+        if cache and file_io.path is not None:
+            module_node = load_module(self._hashed, file_io, cache_path=cache_path)
+            if module_node is not None:
+                return module_node  # type: ignore[no-any-return]
+
+        if code is None:
+            code = file_io.read()
+        code = python_bytes_to_unicode(code)
+
+        lines = split_lines(code, keepends=True)
+        if diff_cache:
+            if self._diff_parser is None:
+                raise TypeError("You have to define a diff parser to be able "
+                                "to use this option.")
+            try:
+                module_cache_item = parser_cache[self._hashed][file_io.path]
+            except KeyError:
+                pass
+            else:
+                module_node = module_cache_item.node
+                old_lines = module_cache_item.lines
+                if old_lines == lines:
+                    return module_node  # type: ignore[no-any-return]
+
+                new_node = self._diff_parser(
+                    self._pgen_grammar, self._tokenizer, module_node
+                ).update(
+                    old_lines=old_lines,
+                    new_lines=lines
+                )
+                try_to_save_module(self._hashed, file_io, new_node, lines,
+                                   # Never pickle in pypy, it's slow as hell.
+                                   pickling=cache and not is_pypy,
+                                   cache_path=cache_path)
+                return new_node  # type: ignore[no-any-return]
+
+        tokens = self._tokenizer(lines)
+
+        p = self._parser(
+            self._pgen_grammar,
+            error_recovery=error_recovery,
+            start_nonterminal=start_symbol
+        )
+        root_node = p.parse(tokens=tokens)
+
+        if cache or diff_cache:
+            try_to_save_module(self._hashed, file_io, root_node, lines,
+                               # Never pickle in pypy, it's slow as hell.
+                               pickling=cache and not is_pypy,
+                               cache_path=cache_path)
+        return root_node  # type: ignore[no-any-return]
+
+    def _get_token_namespace(self):
+        ns = self._token_namespace
+        if ns is None:
+            raise ValueError("The token namespace should be set.")
+        return ns
+
+    def iter_errors(self, node):
+        """
+        Given a :py:class:`parso.tree.NodeOrLeaf` returns a generator of
+        :py:class:`parso.normalizer.Issue` objects. For Python this is
+        a list of syntax/indentation errors.
+        """
+        if self._error_normalizer_config is None:
+            raise ValueError("No error normalizer specified for this grammar.")
+
+        return self._get_normalizer_issues(node, self._error_normalizer_config)
+
+    def refactor(self, base_node, node_to_str_map):
+        return RefactoringNormalizer(node_to_str_map).walk(base_node)
+
+    def _get_normalizer(self, normalizer_config):
+        if normalizer_config is None:
+            normalizer_config = self._default_normalizer_config
+            if normalizer_config is None:
+                raise ValueError("You need to specify a normalizer, because "
+                                 "there's no default normalizer for this tree.")
+        return normalizer_config.create_normalizer(self)
+
+    def _normalize(self, node, normalizer_config=None):
+        """
+        TODO this is not public, yet.
+        The returned code will be normalized, e.g. PEP8 for Python.
+        """
+        normalizer = self._get_normalizer(normalizer_config)
+        return normalizer.walk(node)
+
+    def _get_normalizer_issues(self, node, normalizer_config=None):
+        normalizer = self._get_normalizer(normalizer_config)
+        normalizer.walk(node)
+        return normalizer.issues
+
+    def __repr__(self):
+        nonterminals = self._pgen_grammar.nonterminal_to_dfas.keys()
+        txt = ' '.join(list(nonterminals)[:3]) + ' ...'
+        return '<%s:%s>' % (self.__class__.__name__, txt)
+
+
+class PythonGrammar(Grammar):
+    _error_normalizer_config = ErrorFinderConfig()
+    _token_namespace = PythonTokenTypes
+    _start_nonterminal = 'file_input'
+
+    def __init__(self, version_info: PythonVersionInfo, bnf_text: str):
+        super().__init__(
+            bnf_text,
+            tokenizer=self._tokenize_lines,
+            parser=PythonParser,
+            diff_parser=DiffParser
+        )
+        self.version_info = version_info
+
+    def _tokenize_lines(self, lines, **kwargs):
+        return tokenize_lines(lines, version_info=self.version_info, **kwargs)
+
+    def _tokenize(self, code):
+        # Used by Jedi.
+        return tokenize(code, version_info=self.version_info)
+
+
+def load_grammar(*, version: str = None, path: str = None):
+    """
+    Loads a :py:class:`parso.Grammar`. The default version is the current Python
+    version.
+
+    :param str version: A python version string, e.g. ``version='3.8'``.
+    :param str path: A path to a grammar file
+    """
+    version_info = parse_version_string(version)
+
+    file = path or os.path.join(
+        'python',
+        'grammar%s%s.txt' % (version_info.major, version_info.minor)
+    )
+
+    global _loaded_grammars
+    path = os.path.join(os.path.dirname(__file__), file)
+    try:
+        return _loaded_grammars[path]
+    except KeyError:
+        try:
+            with open(path) as f:
+                bnf_text = f.read()
+
+            grammar = PythonGrammar(version_info, bnf_text)
+            return _loaded_grammars.setdefault(path, grammar)
+        except FileNotFoundError:
+            message = "Python version %s.%s is currently not supported." % (
+                version_info.major, version_info.minor
+            )
+            raise NotImplementedError(message)

evalkit_tf437/lib/python3.10/site-packages/parso/normalizer.py

@@ -0,0 +1,198 @@
+from contextlib import contextmanager
+from typing import Dict, List
+
+
+class _NormalizerMeta(type):
+    def __new__(cls, name, bases, dct):
+        new_cls = type.__new__(cls, name, bases, dct)
+        new_cls.rule_value_classes = {}
+        new_cls.rule_type_classes = {}
+        return new_cls
+
+
+class Normalizer(metaclass=_NormalizerMeta):
+    _rule_type_instances: Dict[str, List[type]] = {}
+    _rule_value_instances: Dict[str, List[type]] = {}
+
+    def __init__(self, grammar, config):
+        self.grammar = grammar
+        self._config = config
+        self.issues = []
+
+        self._rule_type_instances = self._instantiate_rules('rule_type_classes')
+        self._rule_value_instances = self._instantiate_rules('rule_value_classes')
+
+    def _instantiate_rules(self, attr):
+        dct = {}
+        for base in type(self).mro():
+            rules_map = getattr(base, attr, {})
+            for type_, rule_classes in rules_map.items():
+                new = [rule_cls(self) for rule_cls in rule_classes]
+                dct.setdefault(type_, []).extend(new)
+        return dct
+
+    def walk(self, node):
+        self.initialize(node)
+        value = self.visit(node)
+        self.finalize()
+        return value
+
+    def visit(self, node):
+        try:
+            children = node.children
+        except AttributeError:
+            return self.visit_leaf(node)
+        else:
+            with self.visit_node(node):
+                return ''.join(self.visit(child) for child in children)
+
+    @contextmanager
+    def visit_node(self, node):
+        self._check_type_rules(node)
+        yield
+
+    def _check_type_rules(self, node):
+        for rule in self._rule_type_instances.get(node.type, []):
+            rule.feed_node(node)
+
+    def visit_leaf(self, leaf):
+        self._check_type_rules(leaf)
+
+        for rule in self._rule_value_instances.get(leaf.value, []):
+            rule.feed_node(leaf)
+
+        return leaf.prefix + leaf.value
+
+    def initialize(self, node):
+        pass
+
+    def finalize(self):
+        pass
+
+    def add_issue(self, node, code, message):
+        issue = Issue(node, code, message)
+        if issue not in self.issues:
+            self.issues.append(issue)
+        return True
+
+    @classmethod
+    def register_rule(cls, *, value=None, values=(), type=None, types=()):
+        """
+        Use it as a class decorator::
+
+            normalizer = Normalizer('grammar', 'config')
+            @normalizer.register_rule(value='foo')
+            class MyRule(Rule):
+                error_code = 42
+        """
+        values = list(values)
+        types = list(types)
+        if value is not None:
+            values.append(value)
+        if type is not None:
+            types.append(type)
+
+        if not values and not types:
+            raise ValueError("You must register at least something.")
+
+        def decorator(rule_cls):
+            for v in values:
+                cls.rule_value_classes.setdefault(v, []).append(rule_cls)
+            for t in types:
+                cls.rule_type_classes.setdefault(t, []).append(rule_cls)
+            return rule_cls
+
+        return decorator
+
+
+class NormalizerConfig:
+    normalizer_class = Normalizer
+
+    def create_normalizer(self, grammar):
+        if self.normalizer_class is None:
+            return None
+
+        return self.normalizer_class(grammar, self)
+
+
+class Issue:
+    def __init__(self, node, code, message):
+        self.code = code
+        """
+        An integer code that stands for the type of error.
+        """
+        self.message = message
+        """
+        A message (string) for the issue.
+        """
+        self.start_pos = node.start_pos
+        """
+        The start position position of the error as a tuple (line, column). As
+        always in |parso| the first line is 1 and the first column 0.
+        """
+        self.end_pos = node.end_pos
+
+    def __eq__(self, other):
+        return self.start_pos == other.start_pos and self.code == other.code
+
+    def __ne__(self, other):
+        return not self.__eq__(other)
+
+    def __hash__(self):
+        return hash((self.code, self.start_pos))
+
+    def __repr__(self):
+        return '<%s: %s>' % (self.__class__.__name__, self.code)
+
+
+class Rule:
+    code: int
+    message: str
+
+    def __init__(self, normalizer):
+        self._normalizer = normalizer
+
+    def is_issue(self, node):
+        raise NotImplementedError()
+
+    def get_node(self, node):
+        return node
+
+    def _get_message(self, message, node):
+        if message is None:
+            message = self.message
+            if message is None:
+                raise ValueError("The message on the class is not set.")
+        return message
+
+    def add_issue(self, node, code=None, message=None):
+        if code is None:
+            code = self.code
+            if code is None:
+                raise ValueError("The error code on the class is not set.")
+
+        message = self._get_message(message, node)
+
+        self._normalizer.add_issue(node, code, message)
+
+    def feed_node(self, node):
+        if self.is_issue(node):
+            issue_node = self.get_node(node)
+            self.add_issue(issue_node)
+
+
+class RefactoringNormalizer(Normalizer):
+    def __init__(self, node_to_str_map):
+        self._node_to_str_map = node_to_str_map
+
+    def visit(self, node):
+        try:
+            return self._node_to_str_map[node]
+        except KeyError:
+            return super().visit(node)
+
+    def visit_leaf(self, leaf):
+        try:
+            return self._node_to_str_map[leaf]
+        except KeyError:
+            return super().visit_leaf(leaf)

evalkit_tf437/lib/python3.10/site-packages/parso/parser.py

@@ -0,0 +1,210 @@
+# Copyright 2004-2005 Elemental Security, Inc. All Rights Reserved.
+# Licensed to PSF under a Contributor Agreement.
+
+# Modifications:
+# Copyright David Halter and Contributors
+# Modifications are dual-licensed: MIT and PSF.
+# 99% of the code is different from pgen2, now.
+
+"""
+The ``Parser`` tries to convert the available Python code in an easy to read
+format, something like an abstract syntax tree. The classes who represent this
+tree, are sitting in the :mod:`parso.tree` module.
+
+The Python module ``tokenize`` is a very important part in the ``Parser``,
+because it splits the code into different words (tokens).  Sometimes it looks a
+bit messy. Sorry for that! You might ask now: "Why didn't you use the ``ast``
+module for this? Well, ``ast`` does a very good job understanding proper Python
+code, but fails to work as soon as there's a single line of broken code.
+
+There's one important optimization that needs to be known: Statements are not
+being parsed completely. ``Statement`` is just a representation of the tokens
+within the statement. This lowers memory usage and cpu time and reduces the
+complexity of the ``Parser`` (there's another parser sitting inside
+``Statement``, which produces ``Array`` and ``Call``).
+"""
+from typing import Dict, Type
+
+from parso import tree
+from parso.pgen2.generator import ReservedString
+
+
+class ParserSyntaxError(Exception):
+    """
+    Contains error information about the parser tree.
+
+    May be raised as an exception.
+    """
+    def __init__(self, message, error_leaf):
+        self.message = message
+        self.error_leaf = error_leaf
+
+
+class InternalParseError(Exception):
+    """
+    Exception to signal the parser is stuck and error recovery didn't help.
+    Basically this shouldn't happen. It's a sign that something is really
+    wrong.
+    """
+
+    def __init__(self, msg, type_, value, start_pos):
+        Exception.__init__(self, "%s: type=%r, value=%r, start_pos=%r" %
+                           (msg, type_.name, value, start_pos))
+        self.msg = msg
+        self.type = type
+        self.value = value
+        self.start_pos = start_pos
+
+
+class Stack(list):
+    def _allowed_transition_names_and_token_types(self):
+        def iterate():
+            # An API just for Jedi.
+            for stack_node in reversed(self):
+                for transition in stack_node.dfa.transitions:
+                    if isinstance(transition, ReservedString):
+                        yield transition.value
+                    else:
+                        yield transition  # A token type
+
+                if not stack_node.dfa.is_final:
+                    break
+
+        return list(iterate())
+
+
+class StackNode:
+    def __init__(self, dfa):
+        self.dfa = dfa
+        self.nodes = []
+
+    @property
+    def nonterminal(self):
+        return self.dfa.from_rule
+
+    def __repr__(self):
+        return '%s(%s, %s)' % (self.__class__.__name__, self.dfa, self.nodes)
+
+
+def _token_to_transition(grammar, type_, value):
+    # Map from token to label
+    if type_.value.contains_syntax:
+        # Check for reserved words (keywords)
+        try:
+            return grammar.reserved_syntax_strings[value]
+        except KeyError:
+            pass
+
+    return type_
+
+
+class BaseParser:
+    """Parser engine.
+
+    A Parser instance contains state pertaining to the current token
+    sequence, and should not be used concurrently by different threads
+    to parse separate token sequences.
+
+    See python/tokenize.py for how to get input tokens by a string.
+
+    When a syntax error occurs, error_recovery() is called.
+    """
+
+    node_map: Dict[str, Type[tree.BaseNode]] = {}
+    default_node = tree.Node
+
+    leaf_map: Dict[str, Type[tree.Leaf]] = {}
+    default_leaf = tree.Leaf
+
+    def __init__(self, pgen_grammar, start_nonterminal='file_input', error_recovery=False):
+        self._pgen_grammar = pgen_grammar
+        self._start_nonterminal = start_nonterminal
+        self._error_recovery = error_recovery
+
+    def parse(self, tokens):
+        first_dfa = self._pgen_grammar.nonterminal_to_dfas[self._start_nonterminal][0]
+        self.stack = Stack([StackNode(first_dfa)])
+
+        for token in tokens:
+            self._add_token(token)
+
+        while True:
+            tos = self.stack[-1]
+            if not tos.dfa.is_final:
+                # We never broke out -- EOF is too soon -- Unfinished statement.
+                # However, the error recovery might have added the token again, if
+                # the stack is empty, we're fine.
+                raise InternalParseError(
+                    "incomplete input", token.type, token.string, token.start_pos
+                )
+
+            if len(self.stack) > 1:
+                self._pop()
+            else:
+                return self.convert_node(tos.nonterminal, tos.nodes)
+
+    def error_recovery(self, token):
+        if self._error_recovery:
+            raise NotImplementedError("Error Recovery is not implemented")
+        else:
+            type_, value, start_pos, prefix = token
+            error_leaf = tree.ErrorLeaf(type_, value, start_pos, prefix)
+            raise ParserSyntaxError('SyntaxError: invalid syntax', error_leaf)
+
+    def convert_node(self, nonterminal, children):
+        try:
+            node = self.node_map[nonterminal](children)
+        except KeyError:
+            node = self.default_node(nonterminal, children)
+        return node
+
+    def convert_leaf(self, type_, value, prefix, start_pos):
+        try:
+            return self.leaf_map[type_](value, start_pos, prefix)
+        except KeyError:
+            return self.default_leaf(value, start_pos, prefix)
+
+    def _add_token(self, token):
+        """
+        This is the only core function for parsing. Here happens basically
+        everything. Everything is well prepared by the parser generator and we
+        only apply the necessary steps here.
+        """
+        grammar = self._pgen_grammar
+        stack = self.stack
+        type_, value, start_pos, prefix = token
+        transition = _token_to_transition(grammar, type_, value)
+
+        while True:
+            try:
+                plan = stack[-1].dfa.transitions[transition]
+                break
+            except KeyError:
+                if stack[-1].dfa.is_final:
+                    self._pop()
+                else:
+                    self.error_recovery(token)
+                    return
+            except IndexError:
+                raise InternalParseError("too much input", type_, value, start_pos)
+
+        stack[-1].dfa = plan.next_dfa
+
+        for push in plan.dfa_pushes:
+            stack.append(StackNode(push))
+
+        leaf = self.convert_leaf(type_, value, prefix, start_pos)
+        stack[-1].nodes.append(leaf)
+
+    def _pop(self):
+        tos = self.stack.pop()
+        # If there's exactly one child, return that child instead of
+        # creating a new node.  We still create expr_stmt and
+        # file_input though, because a lot of Jedi depends on its
+        # logic.
+        if len(tos.nodes) == 1:
+            new_node = tos.nodes[0]
+        else:
+            new_node = self.convert_node(tos.dfa.from_rule, tos.nodes)
+
+        self.stack[-1].nodes.append(new_node)

evalkit_tf437/lib/python3.10/site-packages/parso/python/diff.py

@@ -0,0 +1,886 @@
+"""
+The diff parser is trying to be a faster version of the normal parser by trying
+to reuse the nodes of a previous pass over the same file. This is also called
+incremental parsing in parser literature. The difference is mostly that with
+incremental parsing you get a range that needs to be reparsed. Here we
+calculate that range ourselves by using difflib. After that it's essentially
+incremental parsing.
+
+The biggest issue of this approach is that we reuse nodes in a mutable way. The
+intial design and idea is quite problematic for this parser, but it is also
+pretty fast. Measurements showed that just copying nodes in Python is simply
+quite a bit slower (especially for big files >3 kLOC). Therefore we did not
+want to get rid of the mutable nodes, since this is usually not an issue.
+
+This is by far the hardest software I ever wrote, exactly because the initial
+design is crappy. When you have to account for a lot of mutable state, it
+creates a ton of issues that you would otherwise not have. This file took
+probably 3-6 months to write, which is insane for a parser.
+
+There is a fuzzer in that helps test this whole thing. Please use it if you
+make changes here. If you run the fuzzer like::
+
+    test/fuzz_diff_parser.py random -n 100000
+
+you can be pretty sure that everything is still fine. I sometimes run the
+fuzzer up to 24h to make sure everything is still ok.
+"""
+import re
+import difflib
+from collections import namedtuple
+import logging
+
+from parso.utils import split_lines
+from parso.python.parser import Parser
+from parso.python.tree import EndMarker
+from parso.python.tokenize import PythonToken, BOM_UTF8_STRING
+from parso.python.token import PythonTokenTypes
+
+LOG = logging.getLogger(__name__)
+DEBUG_DIFF_PARSER = False
+
+_INDENTATION_TOKENS = 'INDENT', 'ERROR_DEDENT', 'DEDENT'
+
+NEWLINE = PythonTokenTypes.NEWLINE
+DEDENT = PythonTokenTypes.DEDENT
+NAME = PythonTokenTypes.NAME
+ERROR_DEDENT = PythonTokenTypes.ERROR_DEDENT
+ENDMARKER = PythonTokenTypes.ENDMARKER
+
+
+def _is_indentation_error_leaf(node):
+    return node.type == 'error_leaf' and node.token_type in _INDENTATION_TOKENS
+
+
+def _get_previous_leaf_if_indentation(leaf):
+    while leaf and _is_indentation_error_leaf(leaf):
+        leaf = leaf.get_previous_leaf()
+    return leaf
+
+
+def _get_next_leaf_if_indentation(leaf):
+    while leaf and _is_indentation_error_leaf(leaf):
+        leaf = leaf.get_next_leaf()
+    return leaf
+
+
+def _get_suite_indentation(tree_node):
+    return _get_indentation(tree_node.children[1])
+
+
+def _get_indentation(tree_node):
+    return tree_node.start_pos[1]
+
+
+def _assert_valid_graph(node):
+    """
+    Checks if the parent/children relationship is correct.
+
+    This is a check that only runs during debugging/testing.
+    """
+    try:
+        children = node.children
+    except AttributeError:
+        # Ignore INDENT is necessary, because indent/dedent tokens don't
+        # contain value/prefix and are just around, because of the tokenizer.
+        if node.type == 'error_leaf' and node.token_type in _INDENTATION_TOKENS:
+            assert not node.value
+            assert not node.prefix
+            return
+
+        # Calculate the content between two start positions.
+        previous_leaf = _get_previous_leaf_if_indentation(node.get_previous_leaf())
+        if previous_leaf is None:
+            content = node.prefix
+            previous_start_pos = 1, 0
+        else:
+            assert previous_leaf.end_pos <= node.start_pos, \
+                (previous_leaf, node)
+
+            content = previous_leaf.value + node.prefix
+            previous_start_pos = previous_leaf.start_pos
+
+        if '\n' in content or '\r' in content:
+            splitted = split_lines(content)
+            line = previous_start_pos[0] + len(splitted) - 1
+            actual = line, len(splitted[-1])
+        else:
+            actual = previous_start_pos[0], previous_start_pos[1] + len(content)
+            if content.startswith(BOM_UTF8_STRING) \
+                    and node.get_start_pos_of_prefix() == (1, 0):
+                # Remove the byte order mark
+                actual = actual[0], actual[1] - 1
+
+        assert node.start_pos == actual, (node.start_pos, actual)
+    else:
+        for child in children:
+            assert child.parent == node, (node, child)
+            _assert_valid_graph(child)
+
+
+def _assert_nodes_are_equal(node1, node2):
+    try:
+        children1 = node1.children
+    except AttributeError:
+        assert not hasattr(node2, 'children'), (node1, node2)
+        assert node1.value == node2.value, (node1, node2)
+        assert node1.type == node2.type, (node1, node2)
+        assert node1.prefix == node2.prefix, (node1, node2)
+        assert node1.start_pos == node2.start_pos, (node1, node2)
+        return
+    else:
+        try:
+            children2 = node2.children
+        except AttributeError:
+            assert False, (node1, node2)
+    for n1, n2 in zip(children1, children2):
+        _assert_nodes_are_equal(n1, n2)
+    assert len(children1) == len(children2), '\n' + repr(children1) + '\n' + repr(children2)
+
+
+def _get_debug_error_message(module, old_lines, new_lines):
+    current_lines = split_lines(module.get_code(), keepends=True)
+    current_diff = difflib.unified_diff(new_lines, current_lines)
+    old_new_diff = difflib.unified_diff(old_lines, new_lines)
+    import parso
+    return (
+        "There's an issue with the diff parser. Please "
+        "report (parso v%s) - Old/New:\n%s\nActual Diff (May be empty):\n%s"
+        % (parso.__version__, ''.join(old_new_diff), ''.join(current_diff))
+    )
+
+
+def _get_last_line(node_or_leaf):
+    last_leaf = node_or_leaf.get_last_leaf()
+    if _ends_with_newline(last_leaf):
+        return last_leaf.start_pos[0]
+    else:
+        n = last_leaf.get_next_leaf()
+        if n.type == 'endmarker' and '\n' in n.prefix:
+            # This is a very special case and has to do with error recovery in
+            # Parso. The problem is basically that there's no newline leaf at
+            # the end sometimes (it's required in the grammar, but not needed
+            # actually before endmarker, CPython just adds a newline to make
+            # source code pass the parser, to account for that Parso error
+            # recovery allows small_stmt instead of simple_stmt).
+            return last_leaf.end_pos[0] + 1
+        return last_leaf.end_pos[0]
+
+
+def _skip_dedent_error_leaves(leaf):
+    while leaf is not None and leaf.type == 'error_leaf' and leaf.token_type == 'DEDENT':
+        leaf = leaf.get_previous_leaf()
+    return leaf
+
+
+def _ends_with_newline(leaf, suffix=''):
+    leaf = _skip_dedent_error_leaves(leaf)
+
+    if leaf.type == 'error_leaf':
+        typ = leaf.token_type.lower()
+    else:
+        typ = leaf.type
+
+    return typ == 'newline' or suffix.endswith('\n') or suffix.endswith('\r')
+
+
+def _flows_finished(pgen_grammar, stack):
+    """
+    if, while, for and try might not be finished, because another part might
+    still be parsed.
+    """
+    for stack_node in stack:
+        if stack_node.nonterminal in ('if_stmt', 'while_stmt', 'for_stmt', 'try_stmt'):
+            return False
+    return True
+
+
+def _func_or_class_has_suite(node):
+    if node.type == 'decorated':
+        node = node.children[-1]
+    if node.type in ('async_funcdef', 'async_stmt'):
+        node = node.children[-1]
+    return node.type in ('classdef', 'funcdef') and node.children[-1].type == 'suite'
+
+
+def _suite_or_file_input_is_valid(pgen_grammar, stack):
+    if not _flows_finished(pgen_grammar, stack):
+        return False
+
+    for stack_node in reversed(stack):
+        if stack_node.nonterminal == 'decorator':
+            # A decorator is only valid with the upcoming function.
+            return False
+
+        if stack_node.nonterminal == 'suite':
+            # If only newline is in the suite, the suite is not valid, yet.
+            return len(stack_node.nodes) > 1
+    # Not reaching a suite means that we're dealing with file_input levels
+    # where there's no need for a valid statement in it. It can also be empty.
+    return True
+
+
+def _is_flow_node(node):
+    if node.type == 'async_stmt':
+        node = node.children[1]
+    try:
+        value = node.children[0].value
+    except AttributeError:
+        return False
+    return value in ('if', 'for', 'while', 'try', 'with')
+
+
+class _PositionUpdatingFinished(Exception):
+    pass
+
+
+def _update_positions(nodes, line_offset, last_leaf):
+    for node in nodes:
+        try:
+            children = node.children
+        except AttributeError:
+            # Is a leaf
+            node.line += line_offset
+            if node is last_leaf:
+                raise _PositionUpdatingFinished
+        else:
+            _update_positions(children, line_offset, last_leaf)
+
+
+class DiffParser:
+    """
+    An advanced form of parsing a file faster. Unfortunately comes with huge
+    side effects. It changes the given module.
+    """
+    def __init__(self, pgen_grammar, tokenizer, module):
+        self._pgen_grammar = pgen_grammar
+        self._tokenizer = tokenizer
+        self._module = module
+
+    def _reset(self):
+        self._copy_count = 0
+        self._parser_count = 0
+
+        self._nodes_tree = _NodesTree(self._module)
+
+    def update(self, old_lines, new_lines):
+        '''
+        The algorithm works as follows:
+
+        Equal:
+            - Assure that the start is a newline, otherwise parse until we get
+              one.
+            - Copy from parsed_until_line + 1 to max(i2 + 1)
+            - Make sure that the indentation is correct (e.g. add DEDENT)
+            - Add old and change positions
+        Insert:
+            - Parse from parsed_until_line + 1 to min(j2 + 1), hopefully not
+              much more.
+
+        Returns the new module node.
+        '''
+        LOG.debug('diff parser start')
+        # Reset the used names cache so they get regenerated.
+        self._module._used_names = None
+
+        self._parser_lines_new = new_lines
+
+        self._reset()
+
+        line_length = len(new_lines)
+        sm = difflib.SequenceMatcher(None, old_lines, self._parser_lines_new)
+        opcodes = sm.get_opcodes()
+        LOG.debug('line_lengths old: %s; new: %s' % (len(old_lines), line_length))
+
+        for operation, i1, i2, j1, j2 in opcodes:
+            LOG.debug('-> code[%s] old[%s:%s] new[%s:%s]',
+                      operation, i1 + 1, i2, j1 + 1, j2)
+
+            if j2 == line_length and new_lines[-1] == '':
+                # The empty part after the last newline is not relevant.
+                j2 -= 1
+
+            if operation == 'equal':
+                line_offset = j1 - i1
+                self._copy_from_old_parser(line_offset, i1 + 1, i2, j2)
+            elif operation == 'replace':
+                self._parse(until_line=j2)
+            elif operation == 'insert':
+                self._parse(until_line=j2)
+            else:
+                assert operation == 'delete'
+
+        # With this action all change will finally be applied and we have a
+        # changed module.
+        self._nodes_tree.close()
+
+        if DEBUG_DIFF_PARSER:
+            # If there is reasonable suspicion that the diff parser is not
+            # behaving well, this should be enabled.
+            try:
+                code = ''.join(new_lines)
+                assert self._module.get_code() == code
+                _assert_valid_graph(self._module)
+                without_diff_parser_module = Parser(
+                    self._pgen_grammar,
+                    error_recovery=True
+                ).parse(self._tokenizer(new_lines))
+                _assert_nodes_are_equal(self._module, without_diff_parser_module)
+            except AssertionError:
+                print(_get_debug_error_message(self._module, old_lines, new_lines))
+                raise
+
+        last_pos = self._module.end_pos[0]
+        if last_pos != line_length:
+            raise Exception(
+                ('(%s != %s) ' % (last_pos, line_length))
+                + _get_debug_error_message(self._module, old_lines, new_lines)
+            )
+        LOG.debug('diff parser end')
+        return self._module
+
+    def _enabled_debugging(self, old_lines, lines_new):
+        if self._module.get_code() != ''.join(lines_new):
+            LOG.warning('parser issue:\n%s\n%s', ''.join(old_lines), ''.join(lines_new))
+
+    def _copy_from_old_parser(self, line_offset, start_line_old, until_line_old, until_line_new):
+        last_until_line = -1
+        while until_line_new > self._nodes_tree.parsed_until_line:
+            parsed_until_line_old = self._nodes_tree.parsed_until_line - line_offset
+            line_stmt = self._get_old_line_stmt(parsed_until_line_old + 1)
+            if line_stmt is None:
+                # Parse 1 line at least. We don't need more, because we just
+                # want to get into a state where the old parser has statements
+                # again that can be copied (e.g. not lines within parentheses).
+                self._parse(self._nodes_tree.parsed_until_line + 1)
+            else:
+                p_children = line_stmt.parent.children
+                index = p_children.index(line_stmt)
+
+                if start_line_old == 1 \
+                        and p_children[0].get_first_leaf().prefix.startswith(BOM_UTF8_STRING):
+                    # If there's a BOM in the beginning, just reparse. It's too
+                    # complicated to account for it otherwise.
+                    copied_nodes = []
+                else:
+                    from_ = self._nodes_tree.parsed_until_line + 1
+                    copied_nodes = self._nodes_tree.copy_nodes(
+                        p_children[index:],
+                        until_line_old,
+                        line_offset
+                    )
+                # Match all the nodes that are in the wanted range.
+                if copied_nodes:
+                    self._copy_count += 1
+
+                    to = self._nodes_tree.parsed_until_line
+
+                    LOG.debug('copy old[%s:%s] new[%s:%s]',
+                              copied_nodes[0].start_pos[0],
+                              copied_nodes[-1].end_pos[0] - 1, from_, to)
+                else:
+                    # We have copied as much as possible (but definitely not too
+                    # much). Therefore we just parse a bit more.
+                    self._parse(self._nodes_tree.parsed_until_line + 1)
+            # Since there are potential bugs that might loop here endlessly, we
+            # just stop here.
+            assert last_until_line != self._nodes_tree.parsed_until_line, last_until_line
+            last_until_line = self._nodes_tree.parsed_until_line
+
+    def _get_old_line_stmt(self, old_line):
+        leaf = self._module.get_leaf_for_position((old_line, 0), include_prefixes=True)
+
+        if _ends_with_newline(leaf):
+            leaf = leaf.get_next_leaf()
+        if leaf.get_start_pos_of_prefix()[0] == old_line:
+            node = leaf
+            while node.parent.type not in ('file_input', 'suite'):
+                node = node.parent
+
+            # Make sure that if only the `else:` line of an if statement is
+            # copied that not the whole thing is going to be copied.
+            if node.start_pos[0] >= old_line:
+                return node
+        # Must be on the same line. Otherwise we need to parse that bit.
+        return None
+
+    def _parse(self, until_line):
+        """
+        Parses at least until the given line, but might just parse more until a
+        valid state is reached.
+        """
+        last_until_line = 0
+        while until_line > self._nodes_tree.parsed_until_line:
+            node = self._try_parse_part(until_line)
+            nodes = node.children
+
+            self._nodes_tree.add_parsed_nodes(nodes, self._keyword_token_indents)
+            if self._replace_tos_indent is not None:
+                self._nodes_tree.indents[-1] = self._replace_tos_indent
+
+            LOG.debug(
+                'parse_part from %s to %s (to %s in part parser)',
+                nodes[0].get_start_pos_of_prefix()[0],
+                self._nodes_tree.parsed_until_line,
+                node.end_pos[0] - 1
+            )
+            # Since the tokenizer sometimes has bugs, we cannot be sure that
+            # this loop terminates. Therefore assert that there's always a
+            # change.
+            assert last_until_line != self._nodes_tree.parsed_until_line, last_until_line
+            last_until_line = self._nodes_tree.parsed_until_line
+
+    def _try_parse_part(self, until_line):
+        """
+        Sets up a normal parser that uses a spezialized tokenizer to only parse
+        until a certain position (or a bit longer if the statement hasn't
+        ended.
+        """
+        self._parser_count += 1
+        # TODO speed up, shouldn't copy the whole list all the time.
+        # memoryview?
+        parsed_until_line = self._nodes_tree.parsed_until_line
+        lines_after = self._parser_lines_new[parsed_until_line:]
+        tokens = self._diff_tokenize(
+            lines_after,
+            until_line,
+            line_offset=parsed_until_line
+        )
+        self._active_parser = Parser(
+            self._pgen_grammar,
+            error_recovery=True
+        )
+        return self._active_parser.parse(tokens=tokens)
+
+    def _diff_tokenize(self, lines, until_line, line_offset=0):
+        was_newline = False
+        indents = self._nodes_tree.indents
+        initial_indentation_count = len(indents)
+
+        tokens = self._tokenizer(
+            lines,
+            start_pos=(line_offset + 1, 0),
+            indents=indents,
+            is_first_token=line_offset == 0,
+        )
+        stack = self._active_parser.stack
+        self._replace_tos_indent = None
+        self._keyword_token_indents = {}
+        # print('start', line_offset + 1, indents)
+        for token in tokens:
+            # print(token, indents)
+            typ = token.type
+            if typ == DEDENT:
+                if len(indents) < initial_indentation_count:
+                    # We are done here, only thing that can come now is an
+                    # endmarker or another dedented code block.
+                    while True:
+                        typ, string, start_pos, prefix = token = next(tokens)
+                        if typ in (DEDENT, ERROR_DEDENT):
+                            if typ == ERROR_DEDENT:
+                                # We want to force an error dedent in the next
+                                # parser/pass. To make this possible we just
+                                # increase the location by one.
+                                self._replace_tos_indent = start_pos[1] + 1
+                                pass
+                        else:
+                            break
+
+                    if '\n' in prefix or '\r' in prefix:
+                        prefix = re.sub(r'[^\n\r]+\Z', '', prefix)
+                    else:
+                        assert start_pos[1] >= len(prefix), repr(prefix)
+                        if start_pos[1] - len(prefix) == 0:
+                            prefix = ''
+                    yield PythonToken(
+                        ENDMARKER, '',
+                        start_pos,
+                        prefix
+                    )
+                    break
+            elif typ == NEWLINE and token.start_pos[0] >= until_line:
+                was_newline = True
+            elif was_newline:
+                was_newline = False
+                if len(indents) == initial_indentation_count:
+                    # Check if the parser is actually in a valid suite state.
+                    if _suite_or_file_input_is_valid(self._pgen_grammar, stack):
+                        yield PythonToken(ENDMARKER, '', token.start_pos, '')
+                        break
+
+            if typ == NAME and token.string in ('class', 'def'):
+                self._keyword_token_indents[token.start_pos] = list(indents)
+
+            yield token
+
+
+class _NodesTreeNode:
+    _ChildrenGroup = namedtuple(
+        '_ChildrenGroup',
+        'prefix children line_offset last_line_offset_leaf')
+
+    def __init__(self, tree_node, parent=None, indentation=0):
+        self.tree_node = tree_node
+        self._children_groups = []
+        self.parent = parent
+        self._node_children = []
+        self.indentation = indentation
+
+    def finish(self):
+        children = []
+        for prefix, children_part, line_offset, last_line_offset_leaf in self._children_groups:
+            first_leaf = _get_next_leaf_if_indentation(
+                children_part[0].get_first_leaf()
+            )
+
+            first_leaf.prefix = prefix + first_leaf.prefix
+            if line_offset != 0:
+                try:
+                    _update_positions(
+                        children_part, line_offset, last_line_offset_leaf)
+                except _PositionUpdatingFinished:
+                    pass
+            children += children_part
+        self.tree_node.children = children
+        # Reset the parents
+        for node in children:
+            node.parent = self.tree_node
+
+        for node_child in self._node_children:
+            node_child.finish()
+
+    def add_child_node(self, child_node):
+        self._node_children.append(child_node)
+
+    def add_tree_nodes(self, prefix, children, line_offset=0,
+                       last_line_offset_leaf=None):
+        if last_line_offset_leaf is None:
+            last_line_offset_leaf = children[-1].get_last_leaf()
+        group = self._ChildrenGroup(
+            prefix, children, line_offset, last_line_offset_leaf
+        )
+        self._children_groups.append(group)
+
+    def get_last_line(self, suffix):
+        line = 0
+        if self._children_groups:
+            children_group = self._children_groups[-1]
+            last_leaf = _get_previous_leaf_if_indentation(
+                children_group.last_line_offset_leaf
+            )
+
+            line = last_leaf.end_pos[0] + children_group.line_offset
+
+            # Newlines end on the next line, which means that they would cover
+            # the next line. That line is not fully parsed at this point.
+            if _ends_with_newline(last_leaf, suffix):
+                line -= 1
+        line += len(split_lines(suffix)) - 1
+
+        if suffix and not suffix.endswith('\n') and not suffix.endswith('\r'):
+            # This is the end of a file (that doesn't end with a newline).
+            line += 1
+
+        if self._node_children:
+            return max(line, self._node_children[-1].get_last_line(suffix))
+        return line
+
+    def __repr__(self):
+        return '<%s: %s>' % (self.__class__.__name__, self.tree_node)
+
+
+class _NodesTree:
+    def __init__(self, module):
+        self._base_node = _NodesTreeNode(module)
+        self._working_stack = [self._base_node]
+        self._module = module
+        self._prefix_remainder = ''
+        self.prefix = ''
+        self.indents = [0]
+
+    @property
+    def parsed_until_line(self):
+        return self._working_stack[-1].get_last_line(self.prefix)
+
+    def _update_insertion_node(self, indentation):
+        for node in reversed(list(self._working_stack)):
+            if node.indentation < indentation or node is self._working_stack[0]:
+                return node
+            self._working_stack.pop()
+
+    def add_parsed_nodes(self, tree_nodes, keyword_token_indents):
+        old_prefix = self.prefix
+        tree_nodes = self._remove_endmarker(tree_nodes)
+        if not tree_nodes:
+            self.prefix = old_prefix + self.prefix
+            return
+
+        assert tree_nodes[0].type != 'newline'
+
+        node = self._update_insertion_node(tree_nodes[0].start_pos[1])
+        assert node.tree_node.type in ('suite', 'file_input')
+        node.add_tree_nodes(old_prefix, tree_nodes)
+        # tos = Top of stack
+        self._update_parsed_node_tos(tree_nodes[-1], keyword_token_indents)
+
+    def _update_parsed_node_tos(self, tree_node, keyword_token_indents):
+        if tree_node.type == 'suite':
+            def_leaf = tree_node.parent.children[0]
+            new_tos = _NodesTreeNode(
+                tree_node,
+                indentation=keyword_token_indents[def_leaf.start_pos][-1],
+            )
+            new_tos.add_tree_nodes('', list(tree_node.children))
+
+            self._working_stack[-1].add_child_node(new_tos)
+            self._working_stack.append(new_tos)
+
+            self._update_parsed_node_tos(tree_node.children[-1], keyword_token_indents)
+        elif _func_or_class_has_suite(tree_node):
+            self._update_parsed_node_tos(tree_node.children[-1], keyword_token_indents)
+
+    def _remove_endmarker(self, tree_nodes):
+        """
+        Helps cleaning up the tree nodes that get inserted.
+        """
+        last_leaf = tree_nodes[-1].get_last_leaf()
+        is_endmarker = last_leaf.type == 'endmarker'
+        self._prefix_remainder = ''
+        if is_endmarker:
+            prefix = last_leaf.prefix
+            separation = max(prefix.rfind('\n'), prefix.rfind('\r'))
+            if separation > -1:
+                # Remove the whitespace part of the prefix after a newline.
+                # That is not relevant if parentheses were opened. Always parse
+                # until the end of a line.
+                last_leaf.prefix, self._prefix_remainder = \
+                    last_leaf.prefix[:separation + 1], last_leaf.prefix[separation + 1:]
+
+        self.prefix = ''
+
+        if is_endmarker:
+            self.prefix = last_leaf.prefix
+
+            tree_nodes = tree_nodes[:-1]
+        return tree_nodes
+
+    def _get_matching_indent_nodes(self, tree_nodes, is_new_suite):
+        # There might be a random dedent where we have to stop copying.
+        # Invalid indents are ok, because the parser handled that
+        # properly before. An invalid dedent can happen, because a few
+        # lines above there was an invalid indent.
+        node_iterator = iter(tree_nodes)
+        if is_new_suite:
+            yield next(node_iterator)
+
+        first_node = next(node_iterator)
+        indent = _get_indentation(first_node)
+        if not is_new_suite and indent not in self.indents:
+            return
+        yield first_node
+
+        for n in node_iterator:
+            if _get_indentation(n) != indent:
+                return
+            yield n
+
+    def copy_nodes(self, tree_nodes, until_line, line_offset):
+        """
+        Copies tree nodes from the old parser tree.
+
+        Returns the number of tree nodes that were copied.
+        """
+        if tree_nodes[0].type in ('error_leaf', 'error_node'):
+            # Avoid copying errors in the beginning. Can lead to a lot of
+            # issues.
+            return []
+
+        indentation = _get_indentation(tree_nodes[0])
+        old_working_stack = list(self._working_stack)
+        old_prefix = self.prefix
+        old_indents = self.indents
+        self.indents = [i for i in self.indents if i <= indentation]
+
+        self._update_insertion_node(indentation)
+
+        new_nodes, self._working_stack, self.prefix, added_indents = self._copy_nodes(
+            list(self._working_stack),
+            tree_nodes,
+            until_line,
+            line_offset,
+            self.prefix,
+        )
+        if new_nodes:
+            self.indents += added_indents
+        else:
+            self._working_stack = old_working_stack
+            self.prefix = old_prefix
+            self.indents = old_indents
+        return new_nodes
+
+    def _copy_nodes(self, working_stack, nodes, until_line, line_offset,
+                    prefix='', is_nested=False):
+        new_nodes = []
+        added_indents = []
+
+        nodes = list(self._get_matching_indent_nodes(
+            nodes,
+            is_new_suite=is_nested,
+        ))
+
+        new_prefix = ''
+        for node in nodes:
+            if node.start_pos[0] > until_line:
+                break
+
+            if node.type == 'endmarker':
+                break
+
+            if node.type == 'error_leaf' and node.token_type in ('DEDENT', 'ERROR_DEDENT'):
+                break
+            # TODO this check might take a bit of time for large files. We
+            # might want to change this to do more intelligent guessing or
+            # binary search.
+            if _get_last_line(node) > until_line:
+                # We can split up functions and classes later.
+                if _func_or_class_has_suite(node):
+                    new_nodes.append(node)
+                break
+            try:
+                c = node.children
+            except AttributeError:
+                pass
+            else:
+                # This case basically appears with error recovery of one line
+                # suites like `def foo(): bar.-`. In this case we might not
+                # include a newline in the statement and we need to take care
+                # of that.
+                n = node
+                if n.type == 'decorated':
+                    n = n.children[-1]
+                if n.type in ('async_funcdef', 'async_stmt'):
+                    n = n.children[-1]
+                if n.type in ('classdef', 'funcdef'):
+                    suite_node = n.children[-1]
+                else:
+                    suite_node = c[-1]
+
+                if suite_node.type in ('error_leaf', 'error_node'):
+                    break
+
+            new_nodes.append(node)
+
+        # Pop error nodes at the end from the list
+        if new_nodes:
+            while new_nodes:
+                last_node = new_nodes[-1]
+                if (last_node.type in ('error_leaf', 'error_node')
+                        or _is_flow_node(new_nodes[-1])):
+                    # Error leafs/nodes don't have a defined start/end. Error
+                    # nodes might not end with a newline (e.g. if there's an
+                    # open `(`). Therefore ignore all of them unless they are
+                    # succeeded with valid parser state.
+                    # If we copy flows at the end, they might be continued
+                    # after the copy limit (in the new parser).
+                    # In this while loop we try to remove until we find a newline.
+                    new_prefix = ''
+                    new_nodes.pop()
+                    while new_nodes:
+                        last_node = new_nodes[-1]
+                        if last_node.get_last_leaf().type == 'newline':
+                            break
+                        new_nodes.pop()
+                    continue
+                if len(new_nodes) > 1 and new_nodes[-2].type == 'error_node':
+                    # The problem here is that Parso error recovery sometimes
+                    # influences nodes before this node.
+                    # Since the new last node is an error node this will get
+                    # cleaned up in the next while iteration.
+                    new_nodes.pop()
+                    continue
+                break
+
+        if not new_nodes:
+            return [], working_stack, prefix, added_indents
+
+        tos = working_stack[-1]
+        last_node = new_nodes[-1]
+        had_valid_suite_last = False
+        # Pop incomplete suites from the list
+        if _func_or_class_has_suite(last_node):
+            suite = last_node
+            while suite.type != 'suite':
+                suite = suite.children[-1]
+
+            indent = _get_suite_indentation(suite)
+            added_indents.append(indent)
+
+            suite_tos = _NodesTreeNode(suite, indentation=_get_indentation(last_node))
+            # Don't need to pass line_offset here, it's already done by the
+            # parent.
+            suite_nodes, new_working_stack, new_prefix, ai = self._copy_nodes(
+                working_stack + [suite_tos], suite.children, until_line, line_offset,
+                is_nested=True,
+            )
+            added_indents += ai
+            if len(suite_nodes) < 2:
+                # A suite only with newline is not valid.
+                new_nodes.pop()
+                new_prefix = ''
+            else:
+                assert new_nodes
+                tos.add_child_node(suite_tos)
+                working_stack = new_working_stack
+                had_valid_suite_last = True
+
+        if new_nodes:
+            if not _ends_with_newline(new_nodes[-1].get_last_leaf()) and not had_valid_suite_last:
+                p = new_nodes[-1].get_next_leaf().prefix
+                # We are not allowed to remove the newline at the end of the
+                # line, otherwise it's going to be missing. This happens e.g.
+                # if a bracket is around before that moves newlines to
+                # prefixes.
+                new_prefix = split_lines(p, keepends=True)[0]
+
+            if had_valid_suite_last:
+                last = new_nodes[-1]
+                if last.type == 'decorated':
+                    last = last.children[-1]
+                if last.type in ('async_funcdef', 'async_stmt'):
+                    last = last.children[-1]
+                last_line_offset_leaf = last.children[-2].get_last_leaf()
+                assert last_line_offset_leaf == ':'
+            else:
+                last_line_offset_leaf = new_nodes[-1].get_last_leaf()
+            tos.add_tree_nodes(
+                prefix, new_nodes, line_offset, last_line_offset_leaf,
+            )
+            prefix = new_prefix
+            self._prefix_remainder = ''
+
+        return new_nodes, working_stack, prefix, added_indents
+
+    def close(self):
+        self._base_node.finish()
+
+        # Add an endmarker.
+        try:
+            last_leaf = self._module.get_last_leaf()
+        except IndexError:
+            end_pos = [1, 0]
+        else:
+            last_leaf = _skip_dedent_error_leaves(last_leaf)
+            end_pos = list(last_leaf.end_pos)
+        lines = split_lines(self.prefix)
+        assert len(lines) > 0
+        if len(lines) == 1:
+            if lines[0].startswith(BOM_UTF8_STRING) and end_pos == [1, 0]:
+                end_pos[1] -= 1
+            end_pos[1] += len(lines[0])
+        else:
+            end_pos[0] += len(lines) - 1
+            end_pos[1] = len(lines[-1])
+
+        endmarker = EndMarker('', tuple(end_pos), self.prefix + self._prefix_remainder)
+        endmarker.parent = self._module
+        self._module.children.append(endmarker)

evalkit_tf437/lib/python3.10/site-packages/parso/python/errors.py

@@ -0,0 +1,1326 @@
+# -*- coding: utf-8 -*-
+import codecs
+import sys
+import warnings
+import re
+from contextlib import contextmanager
+
+from parso.normalizer import Normalizer, NormalizerConfig, Issue, Rule
+from parso.python.tokenize import _get_token_collection
+
+_BLOCK_STMTS = ('if_stmt', 'while_stmt', 'for_stmt', 'try_stmt', 'with_stmt')
+_STAR_EXPR_PARENTS = ('testlist_star_expr', 'testlist_comp', 'exprlist')
+# This is the maximal block size given by python.
+_MAX_BLOCK_SIZE = 20
+_MAX_INDENT_COUNT = 100
+ALLOWED_FUTURES = (
+    'nested_scopes', 'generators', 'division', 'absolute_import',
+    'with_statement', 'print_function', 'unicode_literals', 'generator_stop',
+)
+_COMP_FOR_TYPES = ('comp_for', 'sync_comp_for')
+
+
+def _get_rhs_name(node, version):
+    type_ = node.type
+    if type_ == "lambdef":
+        return "lambda"
+    elif type_ == "atom":
+        comprehension = _get_comprehension_type(node)
+        first, second = node.children[:2]
+        if comprehension is not None:
+            return comprehension
+        elif second.type == "dictorsetmaker":
+            if version < (3, 8):
+                return "literal"
+            else:
+                if second.children[1] == ":" or second.children[0] == "**":
+                    if version < (3, 10):
+                        return "dict display"
+                    else:
+                        return "dict literal"
+                else:
+                    return "set display"
+        elif (
+            first == "("
+            and (second == ")"
+                 or (len(node.children) == 3 and node.children[1].type == "testlist_comp"))
+        ):
+            return "tuple"
+        elif first == "(":
+            return _get_rhs_name(_remove_parens(node), version=version)
+        elif first == "[":
+            return "list"
+        elif first == "{" and second == "}":
+            if version < (3, 10):
+                return "dict display"
+            else:
+                return "dict literal"
+        elif first == "{" and len(node.children) > 2:
+            return "set display"
+    elif type_ == "keyword":
+        if "yield" in node.value:
+            return "yield expression"
+        if version < (3, 8):
+            return "keyword"
+        else:
+            return str(node.value)
+    elif type_ == "operator" and node.value == "...":
+        if version < (3, 10):
+            return "Ellipsis"
+        else:
+            return "ellipsis"
+    elif type_ == "comparison":
+        return "comparison"
+    elif type_ in ("string", "number", "strings"):
+        return "literal"
+    elif type_ == "yield_expr":
+        return "yield expression"
+    elif type_ == "test":
+        return "conditional expression"
+    elif type_ in ("atom_expr", "power"):
+        if node.children[0] == "await":
+            return "await expression"
+        elif node.children[-1].type == "trailer":
+            trailer = node.children[-1]
+            if trailer.children[0] == "(":
+                return "function call"
+            elif trailer.children[0] == "[":
+                return "subscript"
+            elif trailer.children[0] == ".":
+                return "attribute"
+    elif (
+        ("expr" in type_ and "star_expr" not in type_)  # is a substring
+        or "_test" in type_
+        or type_ in ("term", "factor")
+    ):
+        if version < (3, 10):
+            return "operator"
+        else:
+            return "expression"
+    elif type_ == "star_expr":
+        return "starred"
+    elif type_ == "testlist_star_expr":
+        return "tuple"
+    elif type_ == "fstring":
+        return "f-string expression"
+    return type_  # shouldn't reach here
+
+
+def _iter_stmts(scope):
+    """
+    Iterates over all statements and splits up  simple_stmt.
+    """
+    for child in scope.children:
+        if child.type == 'simple_stmt':
+            for child2 in child.children:
+                if child2.type == 'newline' or child2 == ';':
+                    continue
+                yield child2
+        else:
+            yield child
+
+
+def _get_comprehension_type(atom):
+    first, second = atom.children[:2]
+    if second.type == 'testlist_comp' and second.children[1].type in _COMP_FOR_TYPES:
+        if first == '[':
+            return 'list comprehension'
+        else:
+            return 'generator expression'
+    elif second.type == 'dictorsetmaker' and second.children[-1].type in _COMP_FOR_TYPES:
+        if second.children[1] == ':':
+            return 'dict comprehension'
+        else:
+            return 'set comprehension'
+    return None
+
+
+def _is_future_import(import_from):
+    # It looks like a __future__ import that is relative is still a future
+    # import. That feels kind of odd, but whatever.
+    # if import_from.level != 0:
+    #     return False
+    from_names = import_from.get_from_names()
+    return [n.value for n in from_names] == ['__future__']
+
+
+def _remove_parens(atom):
+    """
+    Returns the inner part of an expression like `(foo)`. Also removes nested
+    parens.
+    """
+    try:
+        children = atom.children
+    except AttributeError:
+        pass
+    else:
+        if len(children) == 3 and children[0] == '(':
+            return _remove_parens(atom.children[1])
+    return atom
+
+
+def _skip_parens_bottom_up(node):
+    """
+    Returns an ancestor node of an expression, skipping all levels of parens
+    bottom-up.
+    """
+    while node.parent is not None:
+        node = node.parent
+        if node.type != 'atom' or node.children[0] != '(':
+            return node
+    return None
+
+
+def _iter_params(parent_node):
+    return (n for n in parent_node.children if n.type == 'param' or n.type == 'operator')
+
+
+def _is_future_import_first(import_from):
+    """
+    Checks if the import is the first statement of a file.
+    """
+    found_docstring = False
+    for stmt in _iter_stmts(import_from.get_root_node()):
+        if stmt.type == 'string' and not found_docstring:
+            continue
+        found_docstring = True
+
+        if stmt == import_from:
+            return True
+        if stmt.type == 'import_from' and _is_future_import(stmt):
+            continue
+        return False
+
+
+def _iter_definition_exprs_from_lists(exprlist):
+    def check_expr(child):
+        if child.type == 'atom':
+            if child.children[0] == '(':
+                testlist_comp = child.children[1]
+                if testlist_comp.type == 'testlist_comp':
+                    yield from _iter_definition_exprs_from_lists(testlist_comp)
+                    return
+                else:
+                    # It's a paren that doesn't do anything, like 1 + (1)
+                    yield from check_expr(testlist_comp)
+                    return
+            elif child.children[0] == '[':
+                yield testlist_comp
+                return
+        yield child
+
+    if exprlist.type in _STAR_EXPR_PARENTS:
+        for child in exprlist.children[::2]:
+            yield from check_expr(child)
+    else:
+        yield from check_expr(exprlist)
+
+
+def _get_expr_stmt_definition_exprs(expr_stmt):
+    exprs = []
+    for list_ in expr_stmt.children[:-2:2]:
+        if list_.type in ('testlist_star_expr', 'testlist'):
+            exprs += _iter_definition_exprs_from_lists(list_)
+        else:
+            exprs.append(list_)
+    return exprs
+
+
+def _get_for_stmt_definition_exprs(for_stmt):
+    exprlist = for_stmt.children[1]
+    return list(_iter_definition_exprs_from_lists(exprlist))
+
+
+def _is_argument_comprehension(argument):
+    return argument.children[1].type in _COMP_FOR_TYPES
+
+
+def _any_fstring_error(version, node):
+    if version < (3, 9) or node is None:
+        return False
+    if node.type == "error_node":
+        return any(child.type == "fstring_start" for child in node.children)
+    elif node.type == "fstring":
+        return True
+    else:
+        return node.search_ancestor("fstring")
+
+
+class _Context:
+    def __init__(self, node, add_syntax_error, parent_context=None):
+        self.node = node
+        self.blocks = []
+        self.parent_context = parent_context
+        self._used_name_dict = {}
+        self._global_names = []
+        self._local_params_names = []
+        self._nonlocal_names = []
+        self._nonlocal_names_in_subscopes = []
+        self._add_syntax_error = add_syntax_error
+
+    def is_async_funcdef(self):
+        # Stupidly enough async funcdefs can have two different forms,
+        # depending if a decorator is used or not.
+        return self.is_function() \
+            and self.node.parent.type in ('async_funcdef', 'async_stmt')
+
+    def is_function(self):
+        return self.node.type == 'funcdef'
+
+    def add_name(self, name):
+        parent_type = name.parent.type
+        if parent_type == 'trailer':
+            # We are only interested in first level names.
+            return
+
+        if parent_type == 'global_stmt':
+            self._global_names.append(name)
+        elif parent_type == 'nonlocal_stmt':
+            self._nonlocal_names.append(name)
+        elif parent_type == 'funcdef':
+            self._local_params_names.extend(
+                [param.name.value for param in name.parent.get_params()]
+            )
+        else:
+            self._used_name_dict.setdefault(name.value, []).append(name)
+
+    def finalize(self):
+        """
+        Returns a list of nonlocal names that need to be part of that scope.
+        """
+        self._analyze_names(self._global_names, 'global')
+        self._analyze_names(self._nonlocal_names, 'nonlocal')
+
+        global_name_strs = {n.value: n for n in self._global_names}
+        for nonlocal_name in self._nonlocal_names:
+            try:
+                global_name = global_name_strs[nonlocal_name.value]
+            except KeyError:
+                continue
+
+            message = "name '%s' is nonlocal and global" % global_name.value
+            if global_name.start_pos < nonlocal_name.start_pos:
+                error_name = global_name
+            else:
+                error_name = nonlocal_name
+            self._add_syntax_error(error_name, message)
+
+        nonlocals_not_handled = []
+        for nonlocal_name in self._nonlocal_names_in_subscopes:
+            search = nonlocal_name.value
+            if search in self._local_params_names:
+                continue
+            if search in global_name_strs or self.parent_context is None:
+                message = "no binding for nonlocal '%s' found" % nonlocal_name.value
+                self._add_syntax_error(nonlocal_name, message)
+            elif not self.is_function() or \
+                    nonlocal_name.value not in self._used_name_dict:
+                nonlocals_not_handled.append(nonlocal_name)
+        return self._nonlocal_names + nonlocals_not_handled
+
+    def _analyze_names(self, globals_or_nonlocals, type_):
+        def raise_(message):
+            self._add_syntax_error(base_name, message % (base_name.value, type_))
+
+        params = []
+        if self.node.type == 'funcdef':
+            params = self.node.get_params()
+
+        for base_name in globals_or_nonlocals:
+            found_global_or_nonlocal = False
+            # Somehow Python does it the reversed way.
+            for name in reversed(self._used_name_dict.get(base_name.value, [])):
+                if name.start_pos > base_name.start_pos:
+                    # All following names don't have to be checked.
+                    found_global_or_nonlocal = True
+
+                parent = name.parent
+                if parent.type == 'param' and parent.name == name:
+                    # Skip those here, these definitions belong to the next
+                    # scope.
+                    continue
+
+                if name.is_definition():
+                    if parent.type == 'expr_stmt' \
+                            and parent.children[1].type == 'annassign':
+                        if found_global_or_nonlocal:
+                            # If it's after the global the error seems to be
+                            # placed there.
+                            base_name = name
+                        raise_("annotated name '%s' can't be %s")
+                        break
+                    else:
+                        message = "name '%s' is assigned to before %s declaration"
+                else:
+                    message = "name '%s' is used prior to %s declaration"
+
+                if not found_global_or_nonlocal:
+                    raise_(message)
+                    # Only add an error for the first occurence.
+                    break
+
+            for param in params:
+                if param.name.value == base_name.value:
+                    raise_("name '%s' is parameter and %s"),
+
+    @contextmanager
+    def add_block(self, node):
+        self.blocks.append(node)
+        yield
+        self.blocks.pop()
+
+    def add_context(self, node):
+        return _Context(node, self._add_syntax_error, parent_context=self)
+
+    def close_child_context(self, child_context):
+        self._nonlocal_names_in_subscopes += child_context.finalize()
+
+
+class ErrorFinder(Normalizer):
+    """
+    Searches for errors in the syntax tree.
+    """
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self._error_dict = {}
+        self.version = self.grammar.version_info
+
+    def initialize(self, node):
+        def create_context(node):
+            if node is None:
+                return None
+
+            parent_context = create_context(node.parent)
+            if node.type in ('classdef', 'funcdef', 'file_input'):
+                return _Context(node, self._add_syntax_error, parent_context)
+            return parent_context
+
+        self.context = create_context(node) or _Context(node, self._add_syntax_error)
+        self._indentation_count = 0
+
+    def visit(self, node):
+        if node.type == 'error_node':
+            with self.visit_node(node):
+                # Don't need to investigate the inners of an error node. We
+                # might find errors in there that should be ignored, because
+                # the error node itself already shows that there's an issue.
+                return ''
+        return super().visit(node)
+
+    @contextmanager
+    def visit_node(self, node):
+        self._check_type_rules(node)
+
+        if node.type in _BLOCK_STMTS:
+            with self.context.add_block(node):
+                if len(self.context.blocks) == _MAX_BLOCK_SIZE:
+                    self._add_syntax_error(node, "too many statically nested blocks")
+                yield
+            return
+        elif node.type == 'suite':
+            self._indentation_count += 1
+            if self._indentation_count == _MAX_INDENT_COUNT:
+                self._add_indentation_error(node.children[1], "too many levels of indentation")
+
+        yield
+
+        if node.type == 'suite':
+            self._indentation_count -= 1
+        elif node.type in ('classdef', 'funcdef'):
+            context = self.context
+            self.context = context.parent_context
+            self.context.close_child_context(context)
+
+    def visit_leaf(self, leaf):
+        if leaf.type == 'error_leaf':
+            if leaf.token_type in ('INDENT', 'ERROR_DEDENT'):
+                # Indents/Dedents itself never have a prefix. They are just
+                # "pseudo" tokens that get removed by the syntax tree later.
+                # Therefore in case of an error we also have to check for this.
+                spacing = list(leaf.get_next_leaf()._split_prefix())[-1]
+                if leaf.token_type == 'INDENT':
+                    message = 'unexpected indent'
+                else:
+                    message = 'unindent does not match any outer indentation level'
+                self._add_indentation_error(spacing, message)
+            else:
+                if leaf.value.startswith('\\'):
+                    message = 'unexpected character after line continuation character'
+                else:
+                    match = re.match('\\w{,2}("{1,3}|\'{1,3})', leaf.value)
+                    if match is None:
+                        message = 'invalid syntax'
+                        if (
+                            self.version >= (3, 9)
+                            and leaf.value in _get_token_collection(
+                                self.version
+                            ).always_break_tokens
+                        ):
+                            message = "f-string: " + message
+                    else:
+                        if len(match.group(1)) == 1:
+                            message = 'EOL while scanning string literal'
+                        else:
+                            message = 'EOF while scanning triple-quoted string literal'
+                self._add_syntax_error(leaf, message)
+            return ''
+        elif leaf.value == ':':
+            parent = leaf.parent
+            if parent.type in ('classdef', 'funcdef'):
+                self.context = self.context.add_context(parent)
+
+        # The rest is rule based.
+        return super().visit_leaf(leaf)
+
+    def _add_indentation_error(self, spacing, message):
+        self.add_issue(spacing, 903, "IndentationError: " + message)
+
+    def _add_syntax_error(self, node, message):
+        self.add_issue(node, 901, "SyntaxError: " + message)
+
+    def add_issue(self, node, code, message):
+        # Overwrite the default behavior.
+        # Check if the issues are on the same line.
+        line = node.start_pos[0]
+        args = (code, message, node)
+        self._error_dict.setdefault(line, args)
+
+    def finalize(self):
+        self.context.finalize()
+
+        for code, message, node in self._error_dict.values():
+            self.issues.append(Issue(node, code, message))
+
+
+class IndentationRule(Rule):
+    code = 903
+
+    def _get_message(self, message, node):
+        message = super()._get_message(message, node)
+        return "IndentationError: " + message
+
+
+@ErrorFinder.register_rule(type='error_node')
+class _ExpectIndentedBlock(IndentationRule):
+    message = 'expected an indented block'
+
+    def get_node(self, node):
+        leaf = node.get_next_leaf()
+        return list(leaf._split_prefix())[-1]
+
+    def is_issue(self, node):
+        # This is the beginning of a suite that is not indented.
+        return node.children[-1].type == 'newline'
+
+
+class ErrorFinderConfig(NormalizerConfig):
+    normalizer_class = ErrorFinder
+
+
+class SyntaxRule(Rule):
+    code = 901
+
+    def _get_message(self, message, node):
+        message = super()._get_message(message, node)
+        if (
+            "f-string" not in message
+            and _any_fstring_error(self._normalizer.version, node)
+        ):
+            message = "f-string: " + message
+        return "SyntaxError: " + message
+
+
+@ErrorFinder.register_rule(type='error_node')
+class _InvalidSyntaxRule(SyntaxRule):
+    message = "invalid syntax"
+    fstring_message = "f-string: invalid syntax"
+
+    def get_node(self, node):
+        return node.get_next_leaf()
+
+    def is_issue(self, node):
+        error = node.get_next_leaf().type != 'error_leaf'
+        if (
+            error
+            and _any_fstring_error(self._normalizer.version, node)
+        ):
+            self.add_issue(node, message=self.fstring_message)
+        else:
+            # Error leafs will be added later as an error.
+            return error
+
+
+@ErrorFinder.register_rule(value='await')
+class _AwaitOutsideAsync(SyntaxRule):
+    message = "'await' outside async function"
+
+    def is_issue(self, leaf):
+        return not self._normalizer.context.is_async_funcdef()
+
+    def get_error_node(self, node):
+        # Return the whole await statement.
+        return node.parent
+
+
+@ErrorFinder.register_rule(value='break')
+class _BreakOutsideLoop(SyntaxRule):
+    message = "'break' outside loop"
+
+    def is_issue(self, leaf):
+        in_loop = False
+        for block in self._normalizer.context.blocks:
+            if block.type in ('for_stmt', 'while_stmt'):
+                in_loop = True
+        return not in_loop
+
+
+@ErrorFinder.register_rule(value='continue')
+class _ContinueChecks(SyntaxRule):
+    message = "'continue' not properly in loop"
+    message_in_finally = "'continue' not supported inside 'finally' clause"
+
+    def is_issue(self, leaf):
+        in_loop = False
+        for block in self._normalizer.context.blocks:
+            if block.type in ('for_stmt', 'while_stmt'):
+                in_loop = True
+            if block.type == 'try_stmt':
+                last_block = block.children[-3]
+                if (
+                    last_block == "finally"
+                    and leaf.start_pos > last_block.start_pos
+                    and self._normalizer.version < (3, 8)
+                ):
+                    self.add_issue(leaf, message=self.message_in_finally)
+                    return False  # Error already added
+        if not in_loop:
+            return True
+
+
+@ErrorFinder.register_rule(value='from')
+class _YieldFromCheck(SyntaxRule):
+    message = "'yield from' inside async function"
+
+    def get_node(self, leaf):
+        return leaf.parent.parent  # This is the actual yield statement.
+
+    def is_issue(self, leaf):
+        return leaf.parent.type == 'yield_arg' \
+            and self._normalizer.context.is_async_funcdef()
+
+
+@ErrorFinder.register_rule(type='name')
+class _NameChecks(SyntaxRule):
+    message = 'cannot assign to __debug__'
+    message_none = 'cannot assign to None'
+
+    def is_issue(self, leaf):
+        self._normalizer.context.add_name(leaf)
+
+        if leaf.value == '__debug__' and leaf.is_definition():
+            return True
+
+
+@ErrorFinder.register_rule(type='string')
+class _StringChecks(SyntaxRule):
+    if sys.version_info < (3, 10):
+        message = "bytes can only contain ASCII literal characters."
+    else:
+        message = "bytes can only contain ASCII literal characters"
+
+    def is_issue(self, leaf):
+        string_prefix = leaf.string_prefix.lower()
+        if 'b' in string_prefix \
+                and any(c for c in leaf.value if ord(c) > 127):
+            # b'ä'
+            return True
+
+        if 'r' not in string_prefix:
+            # Raw strings don't need to be checked if they have proper
+            # escaping.
+
+            payload = leaf._get_payload()
+            if 'b' in string_prefix:
+                payload = payload.encode('utf-8')
+                func = codecs.escape_decode
+            else:
+                func = codecs.unicode_escape_decode
+
+            try:
+                with warnings.catch_warnings():
+                    # The warnings from parsing strings are not relevant.
+                    warnings.filterwarnings('ignore')
+                    func(payload)
+            except UnicodeDecodeError as e:
+                self.add_issue(leaf, message='(unicode error) ' + str(e))
+            except ValueError as e:
+                self.add_issue(leaf, message='(value error) ' + str(e))
+
+
+@ErrorFinder.register_rule(value='*')
+class _StarCheck(SyntaxRule):
+    message = "named arguments must follow bare *"
+
+    def is_issue(self, leaf):
+        params = leaf.parent
+        if params.type == 'parameters' and params:
+            after = params.children[params.children.index(leaf) + 1:]
+            after = [child for child in after
+                     if child not in (',', ')') and not child.star_count]
+            return len(after) == 0
+
+
+@ErrorFinder.register_rule(value='**')
+class _StarStarCheck(SyntaxRule):
+    # e.g. {**{} for a in [1]}
+    # TODO this should probably get a better end_pos including
+    #      the next sibling of leaf.
+    message = "dict unpacking cannot be used in dict comprehension"
+
+    def is_issue(self, leaf):
+        if leaf.parent.type == 'dictorsetmaker':
+            comp_for = leaf.get_next_sibling().get_next_sibling()
+            return comp_for is not None and comp_for.type in _COMP_FOR_TYPES
+
+
+@ErrorFinder.register_rule(value='yield')
+@ErrorFinder.register_rule(value='return')
+class _ReturnAndYieldChecks(SyntaxRule):
+    message = "'return' with value in async generator"
+    message_async_yield = "'yield' inside async function"
+
+    def get_node(self, leaf):
+        return leaf.parent
+
+    def is_issue(self, leaf):
+        if self._normalizer.context.node.type != 'funcdef':
+            self.add_issue(self.get_node(leaf), message="'%s' outside function" % leaf.value)
+        elif self._normalizer.context.is_async_funcdef() \
+                and any(self._normalizer.context.node.iter_yield_exprs()):
+            if leaf.value == 'return' and leaf.parent.type == 'return_stmt':
+                return True
+
+
+@ErrorFinder.register_rule(type='strings')
+class _BytesAndStringMix(SyntaxRule):
+    # e.g. 's' b''
+    message = "cannot mix bytes and nonbytes literals"
+
+    def _is_bytes_literal(self, string):
+        if string.type == 'fstring':
+            return False
+        return 'b' in string.string_prefix.lower()
+
+    def is_issue(self, node):
+        first = node.children[0]
+        first_is_bytes = self._is_bytes_literal(first)
+        for string in node.children[1:]:
+            if first_is_bytes != self._is_bytes_literal(string):
+                return True
+
+
+@ErrorFinder.register_rule(type='import_as_names')
+class _TrailingImportComma(SyntaxRule):
+    # e.g. from foo import a,
+    message = "trailing comma not allowed without surrounding parentheses"
+
+    def is_issue(self, node):
+        if node.children[-1] == ',' and node.parent.children[-1] != ')':
+            return True
+
+
+@ErrorFinder.register_rule(type='import_from')
+class _ImportStarInFunction(SyntaxRule):
+    message = "import * only allowed at module level"
+
+    def is_issue(self, node):
+        return node.is_star_import() and self._normalizer.context.parent_context is not None
+
+
+@ErrorFinder.register_rule(type='import_from')
+class _FutureImportRule(SyntaxRule):
+    message = "from __future__ imports must occur at the beginning of the file"
+
+    def is_issue(self, node):
+        if _is_future_import(node):
+            if not _is_future_import_first(node):
+                return True
+
+            for from_name, future_name in node.get_paths():
+                name = future_name.value
+                allowed_futures = list(ALLOWED_FUTURES)
+                if self._normalizer.version >= (3, 7):
+                    allowed_futures.append('annotations')
+                if name == 'braces':
+                    self.add_issue(node, message="not a chance")
+                elif name == 'barry_as_FLUFL':
+                    m = "Seriously I'm not implementing this :) ~ Dave"
+                    self.add_issue(node, message=m)
+                elif name not in allowed_futures:
+                    message = "future feature %s is not defined" % name
+                    self.add_issue(node, message=message)
+
+
+@ErrorFinder.register_rule(type='star_expr')
+class _StarExprRule(SyntaxRule):
+    message_iterable_unpacking = "iterable unpacking cannot be used in comprehension"
+
+    def is_issue(self, node):
+        def check_delete_starred(node):
+            while node.parent is not None:
+                node = node.parent
+                if node.type == 'del_stmt':
+                    return True
+                if node.type not in (*_STAR_EXPR_PARENTS, 'atom'):
+                    return False
+            return False
+
+        if self._normalizer.version >= (3, 9):
+            ancestor = node.parent
+        else:
+            ancestor = _skip_parens_bottom_up(node)
+        # starred expression not in tuple/list/set
+        if ancestor.type not in (*_STAR_EXPR_PARENTS, 'dictorsetmaker') \
+                and not (ancestor.type == 'atom' and ancestor.children[0] != '('):
+            self.add_issue(node, message="can't use starred expression here")
+            return
+
+        if check_delete_starred(node):
+            if self._normalizer.version >= (3, 9):
+                self.add_issue(node, message="cannot delete starred")
+            else:
+                self.add_issue(node, message="can't use starred expression here")
+            return
+
+        if node.parent.type == 'testlist_comp':
+            # [*[] for a in [1]]
+            if node.parent.children[1].type in _COMP_FOR_TYPES:
+                self.add_issue(node, message=self.message_iterable_unpacking)
+
+
+@ErrorFinder.register_rule(types=_STAR_EXPR_PARENTS)
+class _StarExprParentRule(SyntaxRule):
+    def is_issue(self, node):
+        def is_definition(node, ancestor):
+            if ancestor is None:
+                return False
+
+            type_ = ancestor.type
+            if type_ == 'trailer':
+                return False
+
+            if type_ == 'expr_stmt':
+                return node.start_pos < ancestor.children[-1].start_pos
+
+            return is_definition(node, ancestor.parent)
+
+        if is_definition(node, node.parent):
+            args = [c for c in node.children if c != ',']
+            starred = [c for c in args if c.type == 'star_expr']
+            if len(starred) > 1:
+                if self._normalizer.version < (3, 9):
+                    message = "two starred expressions in assignment"
+                else:
+                    message = "multiple starred expressions in assignment"
+                self.add_issue(starred[1], message=message)
+            elif starred:
+                count = args.index(starred[0])
+                if count >= 256:
+                    message = "too many expressions in star-unpacking assignment"
+                    self.add_issue(starred[0], message=message)
+
+
+@ErrorFinder.register_rule(type='annassign')
+class _AnnotatorRule(SyntaxRule):
+    # True: int
+    # {}: float
+    message = "illegal target for annotation"
+
+    def get_node(self, node):
+        return node.parent
+
+    def is_issue(self, node):
+        type_ = None
+        lhs = node.parent.children[0]
+        lhs = _remove_parens(lhs)
+        try:
+            children = lhs.children
+        except AttributeError:
+            pass
+        else:
+            if ',' in children or lhs.type == 'atom' and children[0] == '(':
+                type_ = 'tuple'
+            elif lhs.type == 'atom' and children[0] == '[':
+                type_ = 'list'
+            trailer = children[-1]
+
+        if type_ is None:
+            if not (lhs.type == 'name'
+                    # subscript/attributes are allowed
+                    or lhs.type in ('atom_expr', 'power')
+                    and trailer.type == 'trailer'
+                    and trailer.children[0] != '('):
+                return True
+        else:
+            # x, y: str
+            message = "only single target (not %s) can be annotated"
+            self.add_issue(lhs.parent, message=message % type_)
+
+
+@ErrorFinder.register_rule(type='argument')
+class _ArgumentRule(SyntaxRule):
+    def is_issue(self, node):
+        first = node.children[0]
+        if self._normalizer.version < (3, 8):
+            # a((b)=c) is valid in <3.8
+            first = _remove_parens(first)
+        if node.children[1] == '=' and first.type != 'name':
+            if first.type == 'lambdef':
+                # f(lambda: 1=1)
+                if self._normalizer.version < (3, 8):
+                    message = "lambda cannot contain assignment"
+                else:
+                    message = 'expression cannot contain assignment, perhaps you meant "=="?'
+            else:
+                # f(+x=1)
+                if self._normalizer.version < (3, 8):
+                    message = "keyword can't be an expression"
+                else:
+                    message = 'expression cannot contain assignment, perhaps you meant "=="?'
+            self.add_issue(first, message=message)
+
+        if _is_argument_comprehension(node) and node.parent.type == 'classdef':
+            self.add_issue(node, message='invalid syntax')
+
+
+@ErrorFinder.register_rule(type='nonlocal_stmt')
+class _NonlocalModuleLevelRule(SyntaxRule):
+    message = "nonlocal declaration not allowed at module level"
+
+    def is_issue(self, node):
+        return self._normalizer.context.parent_context is None
+
+
+@ErrorFinder.register_rule(type='arglist')
+class _ArglistRule(SyntaxRule):
+    @property
+    def message(self):
+        if self._normalizer.version < (3, 7):
+            return "Generator expression must be parenthesized if not sole argument"
+        else:
+            return "Generator expression must be parenthesized"
+
+    def is_issue(self, node):
+        arg_set = set()
+        kw_only = False
+        kw_unpacking_only = False
+        for argument in node.children:
+            if argument == ',':
+                continue
+
+            if argument.type == 'argument':
+                first = argument.children[0]
+                if _is_argument_comprehension(argument) and len(node.children) >= 2:
+                    # a(a, b for b in c)
+                    return True
+
+                if first in ('*', '**'):
+                    if first == '*':
+                        if kw_unpacking_only:
+                            # foo(**kwargs, *args)
+                            message = "iterable argument unpacking " \
+                                      "follows keyword argument unpacking"
+                            self.add_issue(argument, message=message)
+                    else:
+                        kw_unpacking_only = True
+                else:  # Is a keyword argument.
+                    kw_only = True
+                    if first.type == 'name':
+                        if first.value in arg_set:
+                            # f(x=1, x=2)
+                            message = "keyword argument repeated"
+                            if self._normalizer.version >= (3, 9):
+                                message += ": {}".format(first.value)
+                            self.add_issue(first, message=message)
+                        else:
+                            arg_set.add(first.value)
+            else:
+                if kw_unpacking_only:
+                    # f(**x, y)
+                    message = "positional argument follows keyword argument unpacking"
+                    self.add_issue(argument, message=message)
+                elif kw_only:
+                    # f(x=2, y)
+                    message = "positional argument follows keyword argument"
+                    self.add_issue(argument, message=message)
+
+
+@ErrorFinder.register_rule(type='parameters')
+@ErrorFinder.register_rule(type='lambdef')
+class _ParameterRule(SyntaxRule):
+    # def f(x=3, y): pass
+    message = "non-default argument follows default argument"
+
+    def is_issue(self, node):
+        param_names = set()
+        default_only = False
+        star_seen = False
+        for p in _iter_params(node):
+            if p.type == 'operator':
+                if p.value == '*':
+                    star_seen = True
+                    default_only = False
+                continue
+
+            if p.name.value in param_names:
+                message = "duplicate argument '%s' in function definition"
+                self.add_issue(p.name, message=message % p.name.value)
+            param_names.add(p.name.value)
+
+            if not star_seen:
+                if p.default is None and not p.star_count:
+                    if default_only:
+                        return True
+                elif p.star_count:
+                    star_seen = True
+                    default_only = False
+                else:
+                    default_only = True
+
+
+@ErrorFinder.register_rule(type='try_stmt')
+class _TryStmtRule(SyntaxRule):
+    message = "default 'except:' must be last"
+
+    def is_issue(self, try_stmt):
+        default_except = None
+        for except_clause in try_stmt.children[3::3]:
+            if except_clause in ('else', 'finally'):
+                break
+            if except_clause == 'except':
+                default_except = except_clause
+            elif default_except is not None:
+                self.add_issue(default_except, message=self.message)
+
+
+@ErrorFinder.register_rule(type='fstring')
+class _FStringRule(SyntaxRule):
+    _fstring_grammar = None
+    message_expr = "f-string expression part cannot include a backslash"
+    message_nested = "f-string: expressions nested too deeply"
+    message_conversion = "f-string: invalid conversion character: expected 's', 'r', or 'a'"
+
+    def _check_format_spec(self, format_spec, depth):
+        self._check_fstring_contents(format_spec.children[1:], depth)
+
+    def _check_fstring_expr(self, fstring_expr, depth):
+        if depth >= 2:
+            self.add_issue(fstring_expr, message=self.message_nested)
+
+        expr = fstring_expr.children[1]
+        if '\\' in expr.get_code():
+            self.add_issue(expr, message=self.message_expr)
+
+        children_2 = fstring_expr.children[2]
+        if children_2.type == 'operator' and children_2.value == '=':
+            conversion = fstring_expr.children[3]
+        else:
+            conversion = children_2
+        if conversion.type == 'fstring_conversion':
+            name = conversion.children[1]
+            if name.value not in ('s', 'r', 'a'):
+                self.add_issue(name, message=self.message_conversion)
+
+        format_spec = fstring_expr.children[-2]
+        if format_spec.type == 'fstring_format_spec':
+            self._check_format_spec(format_spec, depth + 1)
+
+    def is_issue(self, fstring):
+        self._check_fstring_contents(fstring.children[1:-1])
+
+    def _check_fstring_contents(self, children, depth=0):
+        for fstring_content in children:
+            if fstring_content.type == 'fstring_expr':
+                self._check_fstring_expr(fstring_content, depth)
+
+
+class _CheckAssignmentRule(SyntaxRule):
+    def _check_assignment(self, node, is_deletion=False, is_namedexpr=False, is_aug_assign=False):
+        error = None
+        type_ = node.type
+        if type_ == 'lambdef':
+            error = 'lambda'
+        elif type_ == 'atom':
+            first, second = node.children[:2]
+            error = _get_comprehension_type(node)
+            if error is None:
+                if second.type == 'dictorsetmaker':
+                    if self._normalizer.version < (3, 8):
+                        error = 'literal'
+                    else:
+                        if second.children[1] == ':':
+                            if self._normalizer.version < (3, 10):
+                                error = 'dict display'
+                            else:
+                                error = 'dict literal'
+                        else:
+                            error = 'set display'
+                elif first == "{" and second == "}":
+                    if self._normalizer.version < (3, 8):
+                        error = 'literal'
+                    else:
+                        if self._normalizer.version < (3, 10):
+                            error = "dict display"
+                        else:
+                            error = "dict literal"
+                elif first == "{" and len(node.children) > 2:
+                    if self._normalizer.version < (3, 8):
+                        error = 'literal'
+                    else:
+                        error = "set display"
+                elif first in ('(', '['):
+                    if second.type == 'yield_expr':
+                        error = 'yield expression'
+                    elif second.type == 'testlist_comp':
+                        # ([a, b] := [1, 2])
+                        # ((a, b) := [1, 2])
+                        if is_namedexpr:
+                            if first == '(':
+                                error = 'tuple'
+                            elif first == '[':
+                                error = 'list'
+
+                        # This is not a comprehension, they were handled
+                        # further above.
+                        for child in second.children[::2]:
+                            self._check_assignment(child, is_deletion, is_namedexpr, is_aug_assign)
+                    else:  # Everything handled, must be useless brackets.
+                        self._check_assignment(second, is_deletion, is_namedexpr, is_aug_assign)
+        elif type_ == 'keyword':
+            if node.value == "yield":
+                error = "yield expression"
+            elif self._normalizer.version < (3, 8):
+                error = 'keyword'
+            else:
+                error = str(node.value)
+        elif type_ == 'operator':
+            if node.value == '...':
+                if self._normalizer.version < (3, 10):
+                    error = 'Ellipsis'
+                else:
+                    error = 'ellipsis'
+        elif type_ == 'comparison':
+            error = 'comparison'
+        elif type_ in ('string', 'number', 'strings'):
+            error = 'literal'
+        elif type_ == 'yield_expr':
+            # This one seems to be a slightly different warning in Python.
+            message = 'assignment to yield expression not possible'
+            self.add_issue(node, message=message)
+        elif type_ == 'test':
+            error = 'conditional expression'
+        elif type_ in ('atom_expr', 'power'):
+            if node.children[0] == 'await':
+                error = 'await expression'
+            elif node.children[-2] == '**':
+                if self._normalizer.version < (3, 10):
+                    error = 'operator'
+                else:
+                    error = 'expression'
+            else:
+                # Has a trailer
+                trailer = node.children[-1]
+                assert trailer.type == 'trailer'
+                if trailer.children[0] == '(':
+                    error = 'function call'
+                elif is_namedexpr and trailer.children[0] == '[':
+                    error = 'subscript'
+                elif is_namedexpr and trailer.children[0] == '.':
+                    error = 'attribute'
+        elif type_ == "fstring":
+            if self._normalizer.version < (3, 8):
+                error = 'literal'
+            else:
+                error = "f-string expression"
+        elif type_ in ('testlist_star_expr', 'exprlist', 'testlist'):
+            for child in node.children[::2]:
+                self._check_assignment(child, is_deletion, is_namedexpr, is_aug_assign)
+        elif ('expr' in type_ and type_ != 'star_expr'  # is a substring
+              or '_test' in type_
+              or type_ in ('term', 'factor')):
+            if self._normalizer.version < (3, 10):
+                error = 'operator'
+            else:
+                error = 'expression'
+        elif type_ == "star_expr":
+            if is_deletion:
+                if self._normalizer.version >= (3, 9):
+                    error = "starred"
+                else:
+                    self.add_issue(node, message="can't use starred expression here")
+            else:
+                if self._normalizer.version >= (3, 9):
+                    ancestor = node.parent
+                else:
+                    ancestor = _skip_parens_bottom_up(node)
+                if ancestor.type not in _STAR_EXPR_PARENTS and not is_aug_assign \
+                        and not (ancestor.type == 'atom' and ancestor.children[0] == '['):
+                    message = "starred assignment target must be in a list or tuple"
+                    self.add_issue(node, message=message)
+
+            self._check_assignment(node.children[1])
+
+        if error is not None:
+            if is_namedexpr:
+                message = 'cannot use assignment expressions with %s' % error
+            else:
+                cannot = "can't" if self._normalizer.version < (3, 8) else "cannot"
+                message = ' '.join([cannot, "delete" if is_deletion else "assign to", error])
+            self.add_issue(node, message=message)
+
+
+@ErrorFinder.register_rule(type='sync_comp_for')
+class _CompForRule(_CheckAssignmentRule):
+    message = "asynchronous comprehension outside of an asynchronous function"
+
+    def is_issue(self, node):
+        expr_list = node.children[1]
+        if expr_list.type != 'expr_list':  # Already handled.
+            self._check_assignment(expr_list)
+
+        return node.parent.children[0] == 'async' \
+            and not self._normalizer.context.is_async_funcdef()
+
+
+@ErrorFinder.register_rule(type='expr_stmt')
+class _ExprStmtRule(_CheckAssignmentRule):
+    message = "illegal expression for augmented assignment"
+    extended_message = "'{target}' is an " + message
+
+    def is_issue(self, node):
+        augassign = node.children[1]
+        is_aug_assign = augassign != '=' and augassign.type != 'annassign'
+
+        if self._normalizer.version <= (3, 8) or not is_aug_assign:
+            for before_equal in node.children[:-2:2]:
+                self._check_assignment(before_equal, is_aug_assign=is_aug_assign)
+
+        if is_aug_assign:
+            target = _remove_parens(node.children[0])
+            # a, a[b], a.b
+
+            if target.type == "name" or (
+                target.type in ("atom_expr", "power")
+                and target.children[1].type == "trailer"
+                and target.children[-1].children[0] != "("
+            ):
+                return False
+
+            if self._normalizer.version <= (3, 8):
+                return True
+            else:
+                self.add_issue(
+                    node,
+                    message=self.extended_message.format(
+                        target=_get_rhs_name(node.children[0], self._normalizer.version)
+                    ),
+                )
+
+
+@ErrorFinder.register_rule(type='with_item')
+class _WithItemRule(_CheckAssignmentRule):
+    def is_issue(self, with_item):
+        self._check_assignment(with_item.children[2])
+
+
+@ErrorFinder.register_rule(type='del_stmt')
+class _DelStmtRule(_CheckAssignmentRule):
+    def is_issue(self, del_stmt):
+        child = del_stmt.children[1]
+
+        if child.type != 'expr_list':  # Already handled.
+            self._check_assignment(child, is_deletion=True)
+
+
+@ErrorFinder.register_rule(type='expr_list')
+class _ExprListRule(_CheckAssignmentRule):
+    def is_issue(self, expr_list):
+        for expr in expr_list.children[::2]:
+            self._check_assignment(expr)
+
+
+@ErrorFinder.register_rule(type='for_stmt')
+class _ForStmtRule(_CheckAssignmentRule):
+    def is_issue(self, for_stmt):
+        # Some of the nodes here are already used, so no else if
+        expr_list = for_stmt.children[1]
+        if expr_list.type != 'expr_list':  # Already handled.
+            self._check_assignment(expr_list)
+
+
+@ErrorFinder.register_rule(type='namedexpr_test')
+class _NamedExprRule(_CheckAssignmentRule):
+    # namedexpr_test: test [':=' test]
+
+    def is_issue(self, namedexpr_test):
+        # assigned name
+        first = namedexpr_test.children[0]
+
+        def search_namedexpr_in_comp_for(node):
+            while True:
+                parent = node.parent
+                if parent is None:
+                    return parent
+                if parent.type == 'sync_comp_for' and parent.children[3] == node:
+                    return parent
+                node = parent
+
+        if search_namedexpr_in_comp_for(namedexpr_test):
+            # [i+1 for i in (i := range(5))]
+            # [i+1 for i in (j := range(5))]
+            # [i+1 for i in (lambda: (j := range(5)))()]
+            message = 'assignment expression cannot be used in a comprehension iterable expression'
+            self.add_issue(namedexpr_test, message=message)
+
+        # defined names
+        exprlist = list()
+
+        def process_comp_for(comp_for):
+            if comp_for.type == 'sync_comp_for':
+                comp = comp_for
+            elif comp_for.type == 'comp_for':
+                comp = comp_for.children[1]
+            exprlist.extend(_get_for_stmt_definition_exprs(comp))
+
+        def search_all_comp_ancestors(node):
+            has_ancestors = False
+            while True:
+                node = node.search_ancestor('testlist_comp', 'dictorsetmaker')
+                if node is None:
+                    break
+                for child in node.children:
+                    if child.type in _COMP_FOR_TYPES:
+                        process_comp_for(child)
+                        has_ancestors = True
+                        break
+            return has_ancestors
+
+        # check assignment expressions in comprehensions
+        search_all = search_all_comp_ancestors(namedexpr_test)
+        if search_all:
+            if self._normalizer.context.node.type == 'classdef':
+                message = 'assignment expression within a comprehension ' \
+                          'cannot be used in a class body'
+                self.add_issue(namedexpr_test, message=message)
+
+            namelist = [expr.value for expr in exprlist if expr.type == 'name']
+            if first.type == 'name' and first.value in namelist:
+                # [i := 0 for i, j in range(5)]
+                # [[(i := i) for j in range(5)] for i in range(5)]
+                # [i for i, j in range(5) if True or (i := 1)]
+                # [False and (i := 0) for i, j in range(5)]
+                message = 'assignment expression cannot rebind ' \
+                          'comprehension iteration variable %r' % first.value
+                self.add_issue(namedexpr_test, message=message)
+
+        self._check_assignment(first, is_namedexpr=True)

evalkit_tf437/lib/python3.10/site-packages/parso/python/grammar311.txt

@@ -0,0 +1,169 @@
+# Grammar for Python
+
+# NOTE WELL: You should also follow all the steps listed at
+# https://devguide.python.org/grammar/
+
+# Start symbols for the grammar:
+#       single_input is a single interactive statement;
+#       file_input is a module or sequence of commands read from an input file;
+#       eval_input is the input for the eval() functions.
+# NB: compound_stmt in single_input is followed by extra NEWLINE!
+single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE
+file_input: stmt* ENDMARKER
+eval_input: testlist NEWLINE* ENDMARKER
+
+decorator: '@' namedexpr_test NEWLINE
+decorators: decorator+
+decorated: decorators (classdef | funcdef | async_funcdef)
+
+async_funcdef: 'async' funcdef
+funcdef: 'def' NAME parameters ['->' test] ':' suite
+
+parameters: '(' [typedargslist] ')'
+typedargslist: (
+  (tfpdef ['=' test] (',' tfpdef ['=' test])* ',' '/' [',' [ tfpdef ['=' test] (
+        ',' tfpdef ['=' test])* ([',' [
+        '*' [tfpdef] (',' tfpdef ['=' test])* [',' ['**' tfpdef [',']]]
+      | '**' tfpdef [',']]])
+  | '*' [tfpdef] (',' tfpdef ['=' test])* ([',' ['**' tfpdef [',']]])
+  | '**' tfpdef [',']]] )
+|  (tfpdef ['=' test] (',' tfpdef ['=' test])* [',' [
+        '*' [tfpdef] (',' tfpdef ['=' test])* [',' ['**' tfpdef [',']]]
+      | '**' tfpdef [',']]]
+  | '*' [tfpdef] (',' tfpdef ['=' test])* [',' ['**' tfpdef [',']]]
+  | '**' tfpdef [','])
+)
+tfpdef: NAME [':' test]
+varargslist: vfpdef ['=' test ](',' vfpdef ['=' test])* ',' '/' [',' [ (vfpdef ['=' test] (',' vfpdef ['=' test])* [',' [
+        '*' [vfpdef] (',' vfpdef ['=' test])* [',' ['**' vfpdef [',']]]
+      | '**' vfpdef [',']]]
+  | '*' [vfpdef] (',' vfpdef ['=' test])* [',' ['**' vfpdef [',']]]
+  | '**' vfpdef [',']) ]] | (vfpdef ['=' test] (',' vfpdef ['=' test])* [',' [
+        '*' [vfpdef] (',' vfpdef ['=' test])* [',' ['**' vfpdef [',']]]
+      | '**' vfpdef [',']]]
+  | '*' [vfpdef] (',' vfpdef ['=' test])* [',' ['**' vfpdef [',']]]
+  | '**' vfpdef [',']
+)
+vfpdef: NAME
+
+stmt: simple_stmt | compound_stmt | NEWLINE
+simple_stmt: small_stmt (';' small_stmt)* [';'] NEWLINE
+small_stmt: (expr_stmt | del_stmt | pass_stmt | flow_stmt |
+             import_stmt | global_stmt | nonlocal_stmt | assert_stmt)
+expr_stmt: testlist_star_expr (annassign | augassign (yield_expr|testlist) |
+                     ('=' (yield_expr|testlist_star_expr))*)
+annassign: ':' test ['=' (yield_expr|testlist_star_expr)]
+testlist_star_expr: (test|star_expr) (',' (test|star_expr))* [',']
+augassign: ('+=' | '-=' | '*=' | '@=' | '/=' | '%=' | '&=' | '|=' | '^=' |
+            '<<=' | '>>=' | '**=' | '//=')
+# For normal and annotated assignments, additional restrictions enforced by the interpreter
+del_stmt: 'del' exprlist
+pass_stmt: 'pass'
+flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt | yield_stmt
+break_stmt: 'break'
+continue_stmt: 'continue'
+return_stmt: 'return' [testlist_star_expr]
+yield_stmt: yield_expr
+raise_stmt: 'raise' [test ['from' test]]
+import_stmt: import_name | import_from
+import_name: 'import' dotted_as_names
+# note below: the ('.' | '...') is necessary because '...' is tokenized as ELLIPSIS
+import_from: ('from' (('.' | '...')* dotted_name | ('.' | '...')+)
+              'import' ('*' | '(' import_as_names ')' | import_as_names))
+import_as_name: NAME ['as' NAME]
+dotted_as_name: dotted_name ['as' NAME]
+import_as_names: import_as_name (',' import_as_name)* [',']
+dotted_as_names: dotted_as_name (',' dotted_as_name)*
+dotted_name: NAME ('.' NAME)*
+global_stmt: 'global' NAME (',' NAME)*
+nonlocal_stmt: 'nonlocal' NAME (',' NAME)*
+assert_stmt: 'assert' test [',' test]
+
+compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | with_stmt | funcdef | classdef | decorated | async_stmt
+async_stmt: 'async' (funcdef | with_stmt | for_stmt)
+if_stmt: 'if' namedexpr_test ':' suite ('elif' namedexpr_test ':' suite)* ['else' ':' suite]
+while_stmt: 'while' namedexpr_test ':' suite ['else' ':' suite]
+for_stmt: 'for' exprlist 'in' testlist ':' suite ['else' ':' suite]
+try_stmt: ('try' ':' suite
+           ((except_clause ':' suite)+
+            ['else' ':' suite]
+            ['finally' ':' suite] |
+           'finally' ':' suite))
+with_stmt: 'with' with_item (',' with_item)*  ':' suite
+with_item: test ['as' expr]
+# NB compile.c makes sure that the default except clause is last
+except_clause: 'except' [test ['as' NAME]]
+suite: simple_stmt | NEWLINE INDENT stmt+ DEDENT
+
+namedexpr_test: test [':=' test]
+test: or_test ['if' or_test 'else' test] | lambdef
+lambdef: 'lambda' [varargslist] ':' test
+or_test: and_test ('or' and_test)*
+and_test: not_test ('and' not_test)*
+not_test: 'not' not_test | comparison
+comparison: expr (comp_op expr)*
+# <> isn't actually a valid comparison operator in Python. It's here for the
+# sake of a __future__ import described in PEP 401 (which really works :-)
+comp_op: '<'|'>'|'=='|'>='|'<='|'<>'|'!='|'in'|'not' 'in'|'is'|'is' 'not'
+star_expr: '*' expr
+expr: xor_expr ('|' xor_expr)*
+xor_expr: and_expr ('^' and_expr)*
+and_expr: shift_expr ('&' shift_expr)*
+shift_expr: arith_expr (('<<'|'>>') arith_expr)*
+arith_expr: term (('+'|'-') term)*
+term: factor (('*'|'@'|'/'|'%'|'//') factor)*
+factor: ('+'|'-'|'~') factor | power
+power: atom_expr ['**' factor]
+atom_expr: ['await'] atom trailer*
+atom: ('(' [yield_expr|testlist_comp] ')' |
+       '[' [testlist_comp] ']' |
+       '{' [dictorsetmaker] '}' |
+       NAME | NUMBER | strings | '...' | 'None' | 'True' | 'False')
+testlist_comp: (namedexpr_test|star_expr) ( comp_for | (',' (namedexpr_test|star_expr))* [','] )
+trailer: '(' [arglist] ')' | '[' subscriptlist ']' | '.' NAME
+subscriptlist: subscript (',' subscript)* [',']
+subscript: test [':=' test] | [test] ':' [test] [sliceop]
+sliceop: ':' [test]
+exprlist: (expr|star_expr) (',' (expr|star_expr))* [',']
+testlist: test (',' test)* [',']
+dictorsetmaker: ( ((test ':' test | '**' expr)
+                   (comp_for | (',' (test ':' test | '**' expr))* [','])) |
+                  ((test [':=' test] | star_expr)
+                   (comp_for | (',' (test [':=' test] | star_expr))* [','])) )
+
+classdef: 'class' NAME ['(' [arglist] ')'] ':' suite
+
+arglist: argument (',' argument)*  [',']
+
+# The reason that keywords are test nodes instead of NAME is that using NAME
+# results in an ambiguity. ast.c makes sure it's a NAME.
+# "test '=' test" is really "keyword '=' test", but we have no such token.
+# These need to be in a single rule to avoid grammar that is ambiguous
+# to our LL(1) parser. Even though 'test' includes '*expr' in star_expr,
+# we explicitly match '*' here, too, to give it proper precedence.
+# Illegal combinations and orderings are blocked in ast.c:
+# multiple (test comp_for) arguments are blocked; keyword unpackings
+# that precede iterable unpackings are blocked; etc.
+argument: ( test [comp_for] |
+            test ':=' test |
+            test '=' test |
+            '**' test |
+            '*' test )
+
+comp_iter: comp_for | comp_if
+sync_comp_for: 'for' exprlist 'in' or_test [comp_iter]
+comp_for: ['async'] sync_comp_for
+comp_if: 'if' or_test [comp_iter]
+
+# not used in grammar, but may appear in "node" passed from Parser to Compiler
+encoding_decl: NAME
+
+yield_expr: 'yield' [yield_arg]
+yield_arg: 'from' test | testlist_star_expr
+
+strings: (STRING | fstring)+
+fstring: FSTRING_START fstring_content* FSTRING_END
+fstring_content: FSTRING_STRING | fstring_expr
+fstring_conversion: '!' NAME
+fstring_expr: '{' (testlist_comp | yield_expr) ['='] [ fstring_conversion ] [ fstring_format_spec ] '}'
+fstring_format_spec: ':' fstring_content*