Upload FlaxTPUGemma3ForCausalLM

README.md

@@ -0,0 +1,199 @@
+---
+library_name: transformers
+tags: []
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]

config.json

@@ -0,0 +1,97 @@
+{
+  "architectures": [
+    "TPUGemma3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "attn_logit_softcapping": null,
+  "auto_map": {
+    "FlaxAutoModelForCausalLM": "modelling_flax_tpu_gemma3.FlaxTPUGemma3ForCausalLM"
+  },
+  "bos_token_id": 2,
+  "cache_implementation": "hybrid",
+  "eos_token_id": 1,
+  "expand_input_ids": false,
+  "expand_input_ids_dict": null,
+  "expand_input_ids_maxlen": null,
+  "expand_input_ids_vocab_size": null,
+  "final_logit_softcapping": null,
+  "head_dim": 256,
+  "hidden_activation": "gelu_pytorch_tanh",
+  "hidden_size": 3840,
+  "initializer_range": 0.02,
+  "intermediate_size": 15360,
+  "layer_types": [
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 8192,
+  "model_type": "tpu_gemma3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 48,
+  "num_key_value_heads": 8,
+  "pad_token_id": 0,
+  "previous_hidden_size": null,
+  "project_mode": null,
+  "query_pre_attn_scalar": 256,
+  "rms_norm_eps": 1e-06,
+  "rope_local_base_freq": 10000.0,
+  "rope_scaling": {
+    "factor": 8.0,
+    "rope_type": "linear"
+  },
+  "rope_theta": 1000000.0,
+  "skip_out_norm": false,
+  "sliding_window": 1024,
+  "sliding_window_pattern": 6,
+  "torch_dtype": "float32",
+  "transformers_version": "4.53.1",
+  "use_cache": true,
+  "vocab_size": 262208
+}

configuration_tpu_gemma3.py

@@ -0,0 +1,91 @@
+"""TPU Gemma3 model configuration"""
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.modeling_rope_utils import rope_config_validation
+
+
+class TPUGemma3Config(PretrainedConfig):
+    model_type = "tpu_gemma3"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=262_208,
+        hidden_size=2304,
+        intermediate_size=9216,
+        num_hidden_layers=26,
+        num_attention_heads=8,
+        num_key_value_heads=4,
+        head_dim=256,
+        hidden_activation="gelu_pytorch_tanh",
+        max_position_embeddings=131_072,
+        initializer_range=0.02,
+        rms_norm_eps=1e-6,
+        use_cache=True,
+        pad_token_id=0,
+        eos_token_id=1,
+        bos_token_id=2,
+        tie_word_embeddings=True,
+        rope_theta=1_000_000.0,
+        attention_bias=False,
+        attention_dropout=0.0,
+        query_pre_attn_scalar=256,
+        sliding_window=4096,
+        final_logit_softcapping=None,
+        attn_logit_softcapping=None,
+        cache_implementation="hybrid",
+        rope_scaling=None,
+        rope_local_base_freq=10_000.0,
+        sliding_window_pattern=6,
+        expand_input_ids=False, # Transformers-native PyTorch generation support
+        expand_input_ids_maxlen=None,
+        expand_input_ids_vocab_size=None,
+        expand_input_ids_dict=None,
+        project_mode=None,  # latent projection args
+        previous_hidden_size=None,
+        skip_out_norm=False,
+        **kwargs,
+    ):
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.head_dim = head_dim
+        self.num_key_value_heads = num_key_value_heads
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.attention_bias = attention_bias
+        self.attention_dropout = attention_dropout
+        self.hidden_activation = hidden_activation
+        self.query_pre_attn_scalar = query_pre_attn_scalar
+        self.sliding_window = sliding_window
+        self.final_logit_softcapping = final_logit_softcapping
+        self.attn_logit_softcapping = attn_logit_softcapping
+        self.cache_implementation = cache_implementation
+
+        self.rope_local_base_freq = rope_local_base_freq
+        # For configuring HybridCache to work with 5:1 attention pattern
+        self.sliding_window_pattern = sliding_window_pattern
+        self.rope_scaling = rope_scaling
+        rope_config_validation(self)
+
+        self.expand_input_ids = expand_input_ids
+        self.expand_input_ids_maxlen = expand_input_ids_maxlen
+        self.expand_input_ids_vocab_size = expand_input_ids_vocab_size
+        self.expand_input_ids_dict = expand_input_ids_dict
+
+        self.project_mode = project_mode
+        self.previous_hidden_size = previous_hidden_size
+
+        self.skip_out_norm = skip_out_norm

flax_model-00001-of-00010.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20de0fb8831ed9017cf262af59a385cc92694f5a95ed6aad9a0a67b1387120ab
+size 4924127058

flax_model-00002-of-00010.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d549a0c784276c2a8b0ba4d653d6a2766b324f300a13235a99ffdd92058977a
+size 4954842139

flax_model-00003-of-00010.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0e8f4d34c5f4fda72d3fc86ce6241d1fc374b0b0c2479dea2213c3b1cc8df42
+size 4907720119

flax_model-00004-of-00010.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d85c1a386ef667b7c441252bd723ff7fec0328a18f726d94de577b6406c65f6
+size 4954842139

flax_model-00005-of-00010.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb4300f59a804d479207deb66066c7aae0043d954db336f081a557001418eab2
+size 4907720119

flax_model-00006-of-00010.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb69509d6bf85a9df8c4d71f4e2a32b2470e42f3faf6c57fb8b938d976766d27
+size 4954842139

flax_model-00007-of-00010.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73bd0c791f5e143e877750ced4f4394ab0042bf4fa5b5b5a7f2a4ff5141bbb9d
+size 4907720119

flax_model-00008-of-00010.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b30aa50c578af724476a22de6d40d34d706f44f047daeac1fd43048f7fd79a4e
+size 4954842139

flax_model-00009-of-00010.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc46f6938e85ff7ad910f67fe5f00078b90b32697aba4e08f3c14e94176f7268
+size 4907720117

flax_model-00010-of-00010.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc866c0a60f96d0de17b29eaee7041438e6507d59969a7d4c9e4d9f6f48fef08
+size 2689789697

flax_model.msgpack.index.json

@@ -0,0 +1,633 @@
+{
+  "metadata": {
+    "total_size": 47064136704
+  },
+  "weight_map": {
+    "model/embed_tokens/embedding": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/input_layernorm/weight": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/mlp/down_proj/kernel": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/mlp/gate_proj/kernel": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/mlp/up_proj/kernel": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/post_attention_layernorm/weight": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/post_feedforward_layernorm/weight": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/pre_feedforward_layernorm/weight": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/self_attn/k_norm/weight": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/self_attn/k_proj/kernel": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/self_attn/o_proj/kernel": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/self_attn/q_norm/weight": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/self_attn/q_proj/kernel": "flax_model-00001-of-00010.msgpack",
+    "model/layers/0/self_attn/v_proj/kernel": "flax_model-00001-of-00010.msgpack",
+    "model/layers/1/input_layernorm/weight": "flax_model-00001-of-00010.msgpack",
+    "model/layers/1/mlp/down_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/mlp/gate_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/mlp/up_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/post_attention_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/post_feedforward_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/pre_feedforward_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/self_attn/k_norm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/self_attn/k_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/self_attn/o_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/self_attn/q_norm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/self_attn/q_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/1/self_attn/v_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/input_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/mlp/down_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/mlp/gate_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/mlp/up_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/post_attention_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/post_feedforward_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/pre_feedforward_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/self_attn/k_norm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/self_attn/k_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/self_attn/o_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/self_attn/q_norm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/self_attn/q_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/10/self_attn/v_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/input_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/mlp/down_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/mlp/gate_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/mlp/up_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/post_attention_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/post_feedforward_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/pre_feedforward_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/self_attn/k_norm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/self_attn/k_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/self_attn/o_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/self_attn/q_norm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/self_attn/q_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/11/self_attn/v_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/input_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/mlp/down_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/mlp/gate_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/mlp/up_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/post_attention_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/post_feedforward_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/pre_feedforward_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/self_attn/k_norm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/self_attn/k_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/self_attn/o_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/self_attn/q_norm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/self_attn/q_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/12/self_attn/v_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/input_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/mlp/down_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/mlp/gate_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/mlp/up_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/post_attention_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/post_feedforward_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/pre_feedforward_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/self_attn/k_norm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/self_attn/k_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/self_attn/o_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/self_attn/q_norm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/self_attn/q_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/13/self_attn/v_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/14/input_layernorm/weight": "flax_model-00002-of-00010.msgpack",
+    "model/layers/14/mlp/down_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/14/mlp/gate_proj/kernel": "flax_model-00002-of-00010.msgpack",
+    "model/layers/14/mlp/up_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/14/post_attention_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/14/post_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/14/pre_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/14/self_attn/k_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/14/self_attn/k_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/14/self_attn/o_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/14/self_attn/q_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/14/self_attn/q_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/14/self_attn/v_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/input_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/mlp/down_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/mlp/gate_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/mlp/up_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/post_attention_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/post_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/pre_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/self_attn/k_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/self_attn/k_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/self_attn/o_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/self_attn/q_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/self_attn/q_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/15/self_attn/v_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/input_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/mlp/down_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/mlp/gate_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/mlp/up_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/post_attention_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/post_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/pre_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/self_attn/k_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/self_attn/k_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/self_attn/o_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/self_attn/q_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/self_attn/q_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/16/self_attn/v_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/input_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/mlp/down_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/mlp/gate_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/mlp/up_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/post_attention_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/post_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/pre_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/self_attn/k_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/self_attn/k_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/self_attn/o_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/self_attn/q_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/self_attn/q_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/17/self_attn/v_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/input_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/mlp/down_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/mlp/gate_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/mlp/up_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/post_attention_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/post_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/pre_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/self_attn/k_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/self_attn/k_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/self_attn/o_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/self_attn/q_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/self_attn/q_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/18/self_attn/v_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/input_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/mlp/down_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/mlp/gate_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/mlp/up_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/post_attention_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/post_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/pre_feedforward_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/self_attn/k_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/self_attn/k_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/self_attn/o_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/self_attn/q_norm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/self_attn/q_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/19/self_attn/v_proj/kernel": "flax_model-00003-of-00010.msgpack",
+    "model/layers/2/input_layernorm/weight": "flax_model-00003-of-00010.msgpack",
+    "model/layers/2/mlp/down_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/mlp/gate_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/mlp/up_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/post_attention_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/post_feedforward_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/pre_feedforward_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/self_attn/k_norm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/self_attn/k_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/self_attn/o_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/self_attn/q_norm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/self_attn/q_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/2/self_attn/v_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/input_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/mlp/down_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/mlp/gate_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/mlp/up_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/post_attention_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/post_feedforward_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/pre_feedforward_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/self_attn/k_norm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/self_attn/k_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/self_attn/o_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/self_attn/q_norm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/self_attn/q_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/20/self_attn/v_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/input_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/mlp/down_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/mlp/gate_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/mlp/up_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/post_attention_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/post_feedforward_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/pre_feedforward_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/self_attn/k_norm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/self_attn/k_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/self_attn/o_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/self_attn/q_norm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/self_attn/q_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/21/self_attn/v_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/input_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/mlp/down_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/mlp/gate_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/mlp/up_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/post_attention_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/post_feedforward_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/pre_feedforward_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/self_attn/k_norm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/self_attn/k_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/self_attn/o_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/self_attn/q_norm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/self_attn/q_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/22/self_attn/v_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/input_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/mlp/down_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/mlp/gate_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/mlp/up_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/post_attention_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/post_feedforward_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/pre_feedforward_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/self_attn/k_norm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/self_attn/k_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/self_attn/o_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/self_attn/q_norm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/self_attn/q_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/23/self_attn/v_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/24/input_layernorm/weight": "flax_model-00004-of-00010.msgpack",
+    "model/layers/24/mlp/down_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/24/mlp/gate_proj/kernel": "flax_model-00004-of-00010.msgpack",
+    "model/layers/24/mlp/up_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/24/post_attention_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/24/post_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/24/pre_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/24/self_attn/k_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/24/self_attn/k_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/24/self_attn/o_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/24/self_attn/q_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/24/self_attn/q_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/24/self_attn/v_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/input_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/mlp/down_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/mlp/gate_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/mlp/up_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/post_attention_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/post_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/pre_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/self_attn/k_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/self_attn/k_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/self_attn/o_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/self_attn/q_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/self_attn/q_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/25/self_attn/v_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/input_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/mlp/down_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/mlp/gate_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/mlp/up_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/post_attention_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/post_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/pre_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/self_attn/k_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/self_attn/k_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/self_attn/o_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/self_attn/q_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/self_attn/q_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/26/self_attn/v_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/input_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/mlp/down_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/mlp/gate_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/mlp/up_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/post_attention_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/post_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/pre_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/self_attn/k_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/self_attn/k_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/self_attn/o_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/self_attn/q_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/self_attn/q_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/27/self_attn/v_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/input_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/mlp/down_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/mlp/gate_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/mlp/up_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/post_attention_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/post_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/pre_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/self_attn/k_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/self_attn/k_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/self_attn/o_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/self_attn/q_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/self_attn/q_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/28/self_attn/v_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/input_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/mlp/down_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/mlp/gate_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/mlp/up_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/post_attention_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/post_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/pre_feedforward_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/self_attn/k_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/self_attn/k_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/self_attn/o_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/self_attn/q_norm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/self_attn/q_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/29/self_attn/v_proj/kernel": "flax_model-00005-of-00010.msgpack",
+    "model/layers/3/input_layernorm/weight": "flax_model-00005-of-00010.msgpack",
+    "model/layers/3/mlp/down_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/mlp/gate_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/mlp/up_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/post_attention_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/post_feedforward_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/pre_feedforward_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/self_attn/k_norm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/self_attn/k_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/self_attn/o_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/self_attn/q_norm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/self_attn/q_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/3/self_attn/v_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/input_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/mlp/down_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/mlp/gate_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/mlp/up_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/post_attention_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/post_feedforward_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/pre_feedforward_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/self_attn/k_norm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/self_attn/k_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/self_attn/o_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/self_attn/q_norm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/self_attn/q_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/30/self_attn/v_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/input_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/mlp/down_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/mlp/gate_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/mlp/up_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/post_attention_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/post_feedforward_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/pre_feedforward_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/self_attn/k_norm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/self_attn/k_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/self_attn/o_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/self_attn/q_norm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/self_attn/q_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/31/self_attn/v_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/input_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/mlp/down_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/mlp/gate_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/mlp/up_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/post_attention_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/post_feedforward_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/pre_feedforward_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/self_attn/k_norm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/self_attn/k_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/self_attn/o_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/self_attn/q_norm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/self_attn/q_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/32/self_attn/v_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/input_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/mlp/down_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/mlp/gate_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/mlp/up_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/post_attention_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/post_feedforward_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/pre_feedforward_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/self_attn/k_norm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/self_attn/k_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/self_attn/o_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/self_attn/q_norm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/self_attn/q_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/33/self_attn/v_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/34/input_layernorm/weight": "flax_model-00006-of-00010.msgpack",
+    "model/layers/34/mlp/down_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/34/mlp/gate_proj/kernel": "flax_model-00006-of-00010.msgpack",
+    "model/layers/34/mlp/up_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/34/post_attention_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/34/post_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/34/pre_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/34/self_attn/k_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/34/self_attn/k_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/34/self_attn/o_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/34/self_attn/q_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/34/self_attn/q_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/34/self_attn/v_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/input_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/mlp/down_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/mlp/gate_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/mlp/up_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/post_attention_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/post_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/pre_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/self_attn/k_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/self_attn/k_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/self_attn/o_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/self_attn/q_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/self_attn/q_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/35/self_attn/v_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/input_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/mlp/down_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/mlp/gate_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/mlp/up_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/post_attention_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/post_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/pre_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/self_attn/k_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/self_attn/k_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/self_attn/o_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/self_attn/q_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/self_attn/q_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/36/self_attn/v_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/input_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/mlp/down_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/mlp/gate_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/mlp/up_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/post_attention_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/post_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/pre_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/self_attn/k_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/self_attn/k_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/self_attn/o_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/self_attn/q_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/self_attn/q_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/37/self_attn/v_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/input_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/mlp/down_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/mlp/gate_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/mlp/up_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/post_attention_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/post_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/pre_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/self_attn/k_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/self_attn/k_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/self_attn/o_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/self_attn/q_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/self_attn/q_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/38/self_attn/v_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/input_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/mlp/down_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/mlp/gate_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/mlp/up_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/post_attention_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/post_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/pre_feedforward_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/self_attn/k_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/self_attn/k_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/self_attn/o_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/self_attn/q_norm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/self_attn/q_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/39/self_attn/v_proj/kernel": "flax_model-00007-of-00010.msgpack",
+    "model/layers/4/input_layernorm/weight": "flax_model-00007-of-00010.msgpack",
+    "model/layers/4/mlp/down_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/mlp/gate_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/mlp/up_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/post_attention_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/post_feedforward_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/pre_feedforward_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/self_attn/k_norm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/self_attn/k_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/self_attn/o_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/self_attn/q_norm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/self_attn/q_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/4/self_attn/v_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/input_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/mlp/down_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/mlp/gate_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/mlp/up_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/post_attention_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/post_feedforward_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/pre_feedforward_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/self_attn/k_norm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/self_attn/k_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/self_attn/o_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/self_attn/q_norm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/self_attn/q_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/40/self_attn/v_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/input_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/mlp/down_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/mlp/gate_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/mlp/up_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/post_attention_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/post_feedforward_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/pre_feedforward_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/self_attn/k_norm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/self_attn/k_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/self_attn/o_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/self_attn/q_norm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/self_attn/q_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/41/self_attn/v_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/input_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/mlp/down_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/mlp/gate_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/mlp/up_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/post_attention_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/post_feedforward_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/pre_feedforward_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/self_attn/k_norm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/self_attn/k_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/self_attn/o_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/self_attn/q_norm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/self_attn/q_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/42/self_attn/v_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/input_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/mlp/down_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/mlp/gate_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/mlp/up_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/post_attention_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/post_feedforward_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/pre_feedforward_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/self_attn/k_norm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/self_attn/k_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/self_attn/o_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/self_attn/q_norm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/self_attn/q_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/43/self_attn/v_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/44/input_layernorm/weight": "flax_model-00008-of-00010.msgpack",
+    "model/layers/44/mlp/down_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/44/mlp/gate_proj/kernel": "flax_model-00008-of-00010.msgpack",
+    "model/layers/44/mlp/up_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/44/post_attention_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/44/post_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/44/pre_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/44/self_attn/k_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/44/self_attn/k_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/44/self_attn/o_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/44/self_attn/q_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/44/self_attn/q_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/44/self_attn/v_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/input_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/mlp/down_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/mlp/gate_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/mlp/up_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/post_attention_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/post_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/pre_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/self_attn/k_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/self_attn/k_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/self_attn/o_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/self_attn/q_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/self_attn/q_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/45/self_attn/v_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/input_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/mlp/down_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/mlp/gate_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/mlp/up_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/post_attention_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/post_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/pre_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/self_attn/k_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/self_attn/k_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/self_attn/o_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/self_attn/q_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/self_attn/q_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/46/self_attn/v_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/input_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/mlp/down_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/mlp/gate_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/mlp/up_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/post_attention_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/post_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/pre_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/self_attn/k_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/self_attn/k_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/self_attn/o_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/self_attn/q_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/self_attn/q_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/47/self_attn/v_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/input_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/mlp/down_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/mlp/gate_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/mlp/up_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/post_attention_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/post_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/pre_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/self_attn/k_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/self_attn/k_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/self_attn/o_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/self_attn/q_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/self_attn/q_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/5/self_attn/v_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/input_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/mlp/down_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/mlp/gate_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/mlp/up_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/post_attention_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/post_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/pre_feedforward_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/self_attn/k_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/self_attn/k_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/self_attn/o_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/self_attn/q_norm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/self_attn/q_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/6/self_attn/v_proj/kernel": "flax_model-00009-of-00010.msgpack",
+    "model/layers/7/input_layernorm/weight": "flax_model-00009-of-00010.msgpack",
+    "model/layers/7/mlp/down_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/mlp/gate_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/mlp/up_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/post_attention_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/post_feedforward_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/pre_feedforward_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/self_attn/k_norm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/self_attn/k_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/self_attn/o_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/self_attn/q_norm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/self_attn/q_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/7/self_attn/v_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/input_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/mlp/down_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/mlp/gate_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/mlp/up_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/post_attention_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/post_feedforward_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/pre_feedforward_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/self_attn/k_norm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/self_attn/k_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/self_attn/o_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/self_attn/q_norm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/self_attn/q_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/8/self_attn/v_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/input_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/mlp/down_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/mlp/gate_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/mlp/up_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/post_attention_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/post_feedforward_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/pre_feedforward_layernorm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/self_attn/k_norm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/self_attn/k_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/self_attn/o_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/self_attn/q_norm/weight": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/self_attn/q_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/layers/9/self_attn/v_proj/kernel": "flax_model-00010-of-00010.msgpack",
+    "model/norm/weight": "flax_model-00010-of-00010.msgpack"
+  }
+}

modelling_flax_tpu_gemma3.py

@@ -0,0 +1,1018 @@
+"""Flax TPU Gemma3 model."""
+
+from typing import Optional, Tuple
+import copy
+
+import flax.linen as nn
+import jax
+import jax.numpy as jnp
+import numpy as np
+from flax.core.frozen_dict import FrozenDict, freeze, unfreeze
+from flax.linen import combine_masks, make_causal_mask
+from flax.linen.attention import dot_product_attention_weights
+from flax.linen import partitioning as nn_partitioning
+from flax.traverse_util import flatten_dict, unflatten_dict
+from jax import lax
+from jax.sharding import PartitionSpec as P
+
+from transformers.modeling_flax_outputs import FlaxBaseModelOutput, FlaxCausalLMOutput
+from transformers.modeling_flax_utils import ACT2FN, FlaxPreTrainedModel, append_call_sample_docstring
+from transformers.utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging
+from .configuration_tpu_gemma3 import TPUGemma3Config
+
+
+logger = logging.get_logger(__name__)
+
+_CONFIG_FOR_DOC = "TPUGemma3Config"
+_CHECKPOINT_FOR_DOC = "google/gemma-2-2b"
+_REAL_CHECKPOINT_FOR_DOC = "openlm-research/open_llama_3b_v2"
+
+TPU_GEMMA3_START_DOCSTRING = r"""
+
+    This model inherits from [`FlaxPreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+
+    This model is also a Flax Linen
+    [flax.nn.Module](https://flax.readthedocs.io/en/latest/_autosummary/flax.nn.module.html) subclass. Use it as a
+    regular Flax Module and refer to the Flax documentation for all matter related to general usage and behavior.
+
+    Finally, this model supports inherent JAX features such as:
+
+    - [Just-In-Time (JIT) compilation](https://jax.readthedocs.io/en/latest/jax.html#just-in-time-compilation-jit)
+    - [Automatic Differentiation](https://jax.readthedocs.io/en/latest/jax.html#automatic-differentiation)
+    - [Vectorization](https://jax.readthedocs.io/en/latest/jax.html#vectorization-vmap)
+    - [Parallelization](https://jax.readthedocs.io/en/latest/jax.html#parallelization-pmap)
+
+    Parameters:
+        config ([`GemmaConfig`]): Model configuration class with all the parameters of the model.
+            Initializing with a config file does not load the weights associated with the model, only the
+            configuration. Check out the [`~FlaxPreTrainedModel.from_pretrained`] method to load the model weights.
+        dtype (`jax.numpy.dtype`, *optional*, defaults to `jax.numpy.float32`):
+            The data type of the computation. Can be one of `jax.numpy.float32`, `jax.numpy.float16`, or
+            `jax.numpy.bfloat16`.
+
+            This can be used to enable mixed-precision training or half-precision inference on GPUs or TPUs. If
+            specified all the computation will be performed with the given `dtype`.
+
+            **Note that this only specifies the dtype of the computation and does not influence the dtype of model
+            parameters.**
+
+            If you wish to change the dtype of the model parameters, see [`~FlaxPreTrainedModel.to_fp16`] and
+            [`~FlaxPreTrainedModel.to_bf16`].
+"""
+
+TPU_GEMMA3_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`numpy.ndarray` of shape `(batch_size, input_ids_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            [What are attention masks?](../glossary#attention-mask)
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see
+            `past_key_values`).
+
+            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
+            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
+            information on the default strategy.
+
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+        position_ids (`numpy.ndarray` of shape `(batch_size, sequence_length)`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.n_positions - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        past_key_values (`Dict[str, np.ndarray]`, *optional*, returned by `init_cache` or when passing previous `past_key_values`):
+            Dictionary of pre-computed hidden-states (key and values in the attention blocks) that can be used for fast
+            auto-regressive decoding. Pre-computed key and value hidden-states are of shape *[batch_size, max_length]*.
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+"""
+
+remat = nn_partitioning.remat
+
+# adapted from modeling_rope_utils
+def _compute_default_rope_parameters(
+    config=None,
+    seq_len: Optional[int] = None,
+    **rope_kwargs,
+):
+    if config is not None and len(rope_kwargs) > 0:
+        raise ValueError(
+            "Unexpected arguments: `**rope_kwargs` and `config` are mutually exclusive in "
+            f"`_compute_default_rope_parameters`, got `rope_kwargs`={rope_kwargs} and `config`={config}"
+        )
+    if len(rope_kwargs) > 0:
+        base = rope_kwargs["base"]
+        dim = rope_kwargs["dim"]
+    elif config is not None:
+        base = config.rope_theta
+        partial_rotary_factor = config.partial_rotary_factor if hasattr(config, "partial_rotary_factor") else 1.0
+        head_dim = getattr(config, "head_dim", config.hidden_size // config.num_attention_heads)
+        dim = int(head_dim * partial_rotary_factor)
+
+    attention_factor = 1.0  # Unused in this type of RoPE
+
+    # Compute the inverse frequencies
+    inv_freq = 1.0 / (base ** (jnp.arange(0, dim, 2, dtype=jnp.int32).astype(jnp.float32) / dim))
+    return inv_freq, attention_factor
+
+def _compute_linear_scaling_rope_parameters(
+    config=None,
+    seq_len: Optional[int] = None,
+    **rope_kwargs,
+):
+    if config is not None and len(rope_kwargs) > 0:
+        raise ValueError(
+            "Unexpected arguments: `**rope_kwargs` and `config` are mutually exclusive in "
+            f"`_compute_linear_scaling_rope_parameters`, got `rope_kwargs`={rope_kwargs} and `config`={config}"
+        )
+    if len(rope_kwargs) > 0:
+        factor = rope_kwargs["factor"]
+    elif config is not None:
+        factor = config.rope_scaling["factor"]
+
+    # Gets the default RoPE parameters
+    inv_freq, attention_factor = _compute_default_rope_parameters(config, seq_len, **rope_kwargs)
+
+    # Then applies linear scaling to the frequencies.
+    # NOTE: originally, scaling was applied to the position_ids. However, we get `embs = inv_freq @ position_ids`, so
+    # applying scaling to the inverse frequencies is equivalent.
+    inv_freq /= factor
+    return inv_freq, attention_factor
+
+ROPE_INIT_FUNCTIONS = {
+    "default": _compute_default_rope_parameters,
+    "linear": _compute_linear_scaling_rope_parameters,
+}
+
+# Copied from transformers.models.llama.modeling_flax_llama.rotate_half
+def rotate_half(tensor):
+    """Rotates half the hidden dims of the input."""
+    rotate_half_tensor = jnp.concatenate(
+        (-tensor[..., tensor.shape[-1] // 2 :], tensor[..., : tensor.shape[-1] // 2]), axis=-1
+    )
+    return rotate_half_tensor
+
+
+# Adapted from transformers.models.llama.modeling_flax_llama.apply_rotary_pos_emb
+def apply_rotary_pos_emb(tensor, sin_pos, cos_pos):
+    return (tensor * cos_pos[:, :, None, :]) + (rotate_half(tensor) * sin_pos[:, :, None, :])
+
+
+class FlaxTPUGemma3RMSNorm(nn.Module):
+    config: TPUGemma3Config
+    dim_override: Optional[int] = None
+    dtype: jnp.dtype = jnp.float32
+    add_in_projection: bool = False
+    add_out_projection: bool = False
+
+    def setup(self):
+        self.epsilon = self.config.rms_norm_eps
+
+        self.weight_is_matrix = False
+
+        if self.dim_override is not None:
+            self.weight = self.param("weight", lambda _, shape: jnp.ones(shape), self.dim_override)
+        else:
+            if self.add_in_projection:
+                self.in_projection = self.param("in_projection", lambda _, shape: jnp.empty(shape), (self.config.hidden_size, self.config.previous_hidden_size))
+                self.weight = self.param("weight", lambda _, shape: jnp.ones(shape), self.config.previous_hidden_size)
+            elif self.config.project_mode == "wrap":
+                self.weight = self.param("weight", lambda _, shape: jnp.ones(shape), self.config.previous_hidden_size)
+            elif isinstance(self.config.project_mode, str) and self.config.project_mode.startswith("fuse"):
+                self.weight = self.param("weight", lambda _, shape: jnp.eye(shape), self.config.hidden_size)
+                self.weight_is_matrix = True
+            else:
+                self.weight = self.param("weight", lambda _, shape: jnp.ones(shape), self.config.hidden_size)
+
+            if self.add_out_projection:
+                self.out_projection = self.param("out_projection", lambda _, shape: jnp.empty(shape), (self.config.previous_hidden_size, self.config.hidden_size))
+
+    def __call__(self, hidden_states):
+        if self.add_in_projection:
+            hidden_states = hidden_states @ self.in_projection
+
+        variance = jnp.asarray(hidden_states, dtype=jnp.float32)
+        variance = jnp.power(variance, 2)
+        variance = variance.mean(-1, keepdims=True)
+        # use `jax.numpy.sqrt` as `jax.lax.rsqrt` does not match `torch.rsqrt`
+        hidden_states = hidden_states / jnp.sqrt(variance + self.epsilon)
+
+        if self.weight_is_matrix:
+            hidden_states = jnp.asarray(hidden_states, dtype=self.dtype) @ self.weight
+        else:
+            hidden_states = (1 + self.weight) * jnp.asarray(hidden_states, dtype=self.dtype)
+
+        if self.add_out_projection:
+            hidden_states = hidden_states @ self.out_projection
+
+        return hidden_states
+
+
+# Copied from transformers.models.llama.modeling_flax_llama.FlaxLlamaRotaryEmbedding with Llama->Gemma3
+class FlaxTPUGemma3RotaryEmbedding(nn.Module):
+    config: TPUGemma3Config
+    dtype: jnp.dtype = jnp.float32
+
+    def setup(self):
+        self.rope_kwargs = {}
+
+        if self.config.rope_scaling is not None:
+            self.rope_type = self.config.rope_scaling.get("rope_type", self.config.rope_scaling.get("type"))
+        else:
+            self.rope_type = "default"
+        self.max_seq_len_cached = self.config.max_position_embeddings
+        self.original_max_seq_len = self.config.max_position_embeddings
+
+        self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
+        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, **self.rope_kwargs)
+        self.inv_freq = self.original_inv_freq = inv_freq
+
+    def __call__(self, x, position_ids):
+        inv_freq_expanded = jnp.tile(
+            self.inv_freq[None, :, None].astype(jnp.float32),
+            (position_ids.shape[0], 1, 1),
+        )
+        position_ids_expanded = position_ids[:, None, :].astype(jnp.float32)
+
+        freqs = jnp.swapaxes(jnp.matmul(inv_freq_expanded, position_ids_expanded), 1, 2)
+        emb = jnp.concatenate([freqs, freqs], axis=-1)
+        cos = jnp.cos(emb)
+        sin = jnp.sin(emb)
+
+        cos = cos * self.attention_scaling
+        sin = sin * self.attention_scaling
+
+        return cos.astype(x.dtype), sin.astype(x.dtype)
+
+
+class FlaxTPUGemma3Attention(nn.Module):
+    config: TPUGemma3Config
+    layer_idx: int
+    dtype: jnp.dtype = jnp.float32
+    causal: bool = True
+    is_cross_attention: bool = False
+
+    def setup(self):
+        self.is_sliding = bool((self.layer_idx + 1) % self.config.sliding_window_pattern)
+        self.sliding_window = self.config.sliding_window if self.is_sliding else None
+
+        config = self.config
+        if self.config.project_mode == "wrap":
+            self.embed_dim = config.previous_hidden_size
+        else:
+            self.embed_dim = config.hidden_size
+
+        self.num_heads = config.num_attention_heads
+        self.head_dim = config.head_dim
+
+        # otherwise we would manually have to scale attn weights
+        assert config.query_pre_attn_scalar == config.head_dim
+
+        self.attention_softmax_in_fp32 = self.dtype is not jnp.float32
+
+        self.num_key_value_heads = config.num_key_value_heads
+        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
+
+        kernel = jax.nn.initializers.normal(self.config.initializer_range)
+        self.q_proj = nn.Dense(
+            self.num_heads * self.head_dim, use_bias=config.attention_bias, dtype=self.dtype, kernel_init=kernel
+        )
+        self.k_proj = nn.Dense(
+            self.num_key_value_heads * self.head_dim,
+            use_bias=config.attention_bias,
+            dtype=self.dtype,
+            kernel_init=kernel,
+        )
+        self.v_proj = nn.Dense(
+            self.num_key_value_heads * self.head_dim,
+            use_bias=config.attention_bias,
+            dtype=self.dtype,
+            kernel_init=kernel,
+        )
+        self.q_norm = FlaxTPUGemma3RMSNorm(self.config, dtype=self.dtype, dim_override=self.head_dim)
+        self.k_norm = FlaxTPUGemma3RMSNorm(self.config, dtype=self.dtype, dim_override=self.head_dim)
+
+        self.o_proj = nn.Dense(self.embed_dim, use_bias=config.attention_bias, dtype=self.dtype, kernel_init=kernel)
+
+        self.causal_mask = make_causal_mask(jnp.ones((1, config.max_position_embeddings), dtype="bool"), dtype="bool")
+
+    def _split_heads(self, hidden_states, num_heads):
+        return hidden_states.reshape(hidden_states.shape[:2] + (num_heads, self.head_dim))
+
+    def _merge_heads(self, hidden_states):
+        return hidden_states.reshape(hidden_states.shape[:2] + (self.num_heads * self.head_dim,))
+
+    @nn.compact
+    # Copied from transformers.models.gpt_neo.modeling_flax_gpt_neo.FlaxGPTNeoSelfAttention._concatenate_to_cache
+    def _concatenate_to_cache(self, key, value, query, attention_mask):
+        """
+        This function takes projected key, value states from a single input token and concatenates the states to cached
+        states from previous steps. This function is slighly adapted from the official Flax repository:
+        https://github.com/google/flax/blob/491ce18759622506588784b4fca0e4bf05f8c8cd/flax/linen/attention.py#L252
+        """
+        # detect if we're initializing by absence of existing cache data.
+        is_initialized = self.has_variable("cache", "cached_key")
+        cached_key = self.variable("cache", "cached_key", jnp.zeros, key.shape, key.dtype)
+        cached_value = self.variable("cache", "cached_value", jnp.zeros, value.shape, value.dtype)
+        cache_index = self.variable("cache", "cache_index", lambda: jnp.array(0, dtype=jnp.int32))
+
+        if is_initialized:
+            *batch_dims, max_length, num_heads, depth_per_head = cached_key.value.shape
+            # update key, value caches with our new 1d spatial slices
+            cur_index = cache_index.value
+            indices = (0,) * len(batch_dims) + (cur_index, 0, 0)
+            key = lax.dynamic_update_slice(cached_key.value, key, indices)
+            value = lax.dynamic_update_slice(cached_value.value, value, indices)
+            cached_key.value = key
+            cached_value.value = value
+            num_updated_cache_vectors = query.shape[1]
+            cache_index.value = cache_index.value + num_updated_cache_vectors
+            # causal mask for cached decoder self-attention: our single query position should only attend to those key positions that have already been generated and cached, not the remaining zero elements.
+            pad_mask = jnp.broadcast_to(
+                jnp.arange(max_length) < cur_index + num_updated_cache_vectors,
+                tuple(batch_dims) + (1, num_updated_cache_vectors, max_length),
+            )
+            attention_mask = combine_masks(pad_mask, attention_mask)
+        return key, value, attention_mask
+
+    def __call__(
+        self,
+        hidden_states,
+        position_embeddings,
+        attention_mask,
+        position_ids,
+        deterministic: bool = True,
+        init_cache: bool = False,
+        output_attentions: bool = False,
+    ):
+        raw_query = self.q_proj(hidden_states)
+        raw_key = self.k_proj(hidden_states)
+        raw_value = self.v_proj(hidden_states)
+
+        query = self._split_heads(raw_query, self.num_heads)
+        key = self._split_heads(raw_key, self.num_key_value_heads)
+        value = self._split_heads(raw_value, self.num_key_value_heads)
+
+        query = self.q_norm(query)
+        key = self.k_norm(key)
+
+        cos, sin = position_embeddings
+
+        key = jnp.asarray(apply_rotary_pos_emb(key, sin, cos), dtype=self.dtype)
+        query = jnp.asarray(apply_rotary_pos_emb(query, sin, cos), dtype=self.dtype)
+
+        query_length, key_length = query.shape[1], key.shape[1]
+
+        if self.has_variable("cache", "cached_key"):
+            mask_shift = self.variables["cache"]["cache_index"]
+            max_decoder_length = self.variables["cache"]["cached_key"].shape[1]
+            causal_mask = lax.dynamic_slice(
+                self.causal_mask, (0, 0, mask_shift, 0), (1, 1, query_length, max_decoder_length)
+            )
+        else:
+            causal_mask = self.causal_mask[:, :, :query_length, :key_length]
+
+        batch_size = hidden_states.shape[0]
+        causal_mask = jnp.broadcast_to(causal_mask, (batch_size,) + causal_mask.shape[1:])
+
+        attention_mask = jnp.broadcast_to(jnp.expand_dims(attention_mask, axis=(-3, -2)), causal_mask.shape)
+        attention_mask = combine_masks(attention_mask, causal_mask)
+
+        if self.sliding_window is not None:
+            min_dtype = jnp.finfo(hidden_states.dtype).min
+            sliding_window_mask = jnp.tril(
+                jnp.ones_like(attention_mask, dtype=bool), k=-self.sliding_window
+            )
+            attention_mask = jnp.where(sliding_window_mask, min_dtype, attention_mask)
+            if attention_mask.shape[-1] <= 1:  # when decoding
+                attention_mask = attention_mask[:, :, :, -self.sliding_window :]
+
+        dropout_rng = None
+        if not deterministic and self.config.attention_dropout > 0.0:
+            dropout_rng = self.make_rng("dropout")
+
+        # During fast autoregressive decoding, we feed one position at a time,
+        # and cache the keys and values step by step.
+        if self.has_variable("cache", "cached_key") or init_cache:
+            key, value, attention_mask = self._concatenate_to_cache(key, value, query, attention_mask)
+
+        # transform boolean mask into float mask
+        attention_bias = lax.select(
+            attention_mask > 0,
+            jnp.full(attention_mask.shape, 0.0).astype(self.dtype),
+            jnp.full(attention_mask.shape, jnp.finfo(self.dtype).min).astype(self.dtype),
+        )
+
+        key = jnp.repeat(key, repeats=self.num_key_value_groups, axis=2)
+        value = jnp.repeat(value, repeats=self.num_key_value_groups, axis=2)
+
+        # usual dot product attention
+        attention_dtype = jnp.float32 if self.attention_softmax_in_fp32 else self.dtype
+        attn_weights = dot_product_attention_weights(
+            query,
+            key,
+            bias=attention_bias,
+            dropout_rng=dropout_rng,
+            dropout_rate=self.config.attention_dropout,
+            deterministic=deterministic,
+            dtype=attention_dtype,
+        )
+
+        if self.config.attn_logit_softcapping is not None:
+            attn_weights = attn_weights / self.config.attn_logit_softcapping
+            attn_weights = jnp.tanh(attn_weights)
+            attn_weights = attn_weights * self.config.attn_logit_softcapping
+
+        if self.attention_softmax_in_fp32:
+            attn_weights = attn_weights.astype(self.dtype)
+
+        attn_output = jnp.einsum("...hqk,...khd->...qhd", attn_weights, value)
+        attn_output = self._merge_heads(attn_output)
+        attn_output = self.o_proj(attn_output)
+
+        outputs = (attn_output, (raw_query, raw_key, raw_value)) if output_attentions else (attn_output,)
+        return outputs
+
+
+class FlaxTPUGemma3MLP(nn.Module):
+    config: TPUGemma3Config
+    dtype: jnp.dtype = jnp.float32
+
+    def setup(self):
+        if self.config.project_mode == "wrap":
+            embed_dim = self.config.previous_hidden_size
+        else:
+            embed_dim = self.config.hidden_size
+
+        inner_dim = self.config.intermediate_size if self.config.intermediate_size is not None else 4 * embed_dim
+
+        kernel_init = jax.nn.initializers.normal(self.config.initializer_range)
+        if self.config.hidden_activation is None:
+            logger.warning_once(
+                "Gemma3's activation function should be approximate GeLU and not exact GeLU. "
+                "Changing the activation function to `gelu_pytorch_tanh`."
+                f"if you want to use the legacy `{self.config.hidden_act}`, "
+                f"edit the `model.config` to set `hidden_activation={self.config.hidden_act}` "
+                "  instead of `hidden_act`. See https://github.com/huggingface/transformers/pull/29402 for more details."
+            )
+            hidden_activation = "gelu_pytorch_tanh"
+        else:
+            hidden_activation = self.config.hidden_activation
+        self.act = ACT2FN[hidden_activation]
+
+        self.gate_proj = nn.Dense(inner_dim, use_bias=False, dtype=self.dtype, kernel_init=kernel_init)
+        self.down_proj = nn.Dense(embed_dim, use_bias=False, dtype=self.dtype, kernel_init=kernel_init)
+        self.up_proj = nn.Dense(inner_dim, use_bias=False, dtype=self.dtype, kernel_init=kernel_init)
+
+    def __call__(self, hidden_states):
+        up_proj_states = self.up_proj(hidden_states)
+        gate_states = self.act(self.gate_proj(hidden_states))
+
+        hidden_states = self.down_proj(up_proj_states * gate_states)
+        return hidden_states
+
+
+# Copied from transformers.models.llama.modeling_flax_llama.FlaxLlamaDecoderLayer with Llama->Gemma3
+class FlaxTPUGemma3DecoderLayer(nn.Module):
+    config: TPUGemma3Config
+    layer_idx: int
+    dtype: jnp.dtype = jnp.float32
+
+    def setup(self):
+        self.input_layernorm = FlaxTPUGemma3RMSNorm(self.config, dtype=self.dtype, add_in_projection=self.config.project_mode == "wrap")
+        self.self_attn = FlaxTPUGemma3Attention(self.config, self.layer_idx, dtype=self.dtype)
+        self.pre_feedforward_layernorm = FlaxTPUGemma3RMSNorm(self.config, dtype=self.dtype, add_in_projection=self.config.project_mode == "wrap")
+        self.post_feedforward_layernorm = FlaxTPUGemma3RMSNorm(self.config, dtype=self.dtype, add_out_projection=self.config.project_mode == "wrap")
+        self.post_attention_layernorm = FlaxTPUGemma3RMSNorm(self.config, dtype=self.dtype, add_out_projection=self.config.project_mode == "wrap")
+        self.mlp = FlaxTPUGemma3MLP(self.config, dtype=self.dtype)
+
+    def __call__(
+        self,
+        hidden_states,
+        position_embeddings_global,
+        position_embeddings_local,
+        attention_mask=None,
+        position_ids=None,
+        deterministic: bool = True,
+        init_cache: bool = False,
+        output_attentions: bool = False,
+    ):
+        mesh = getattr(self.config, "mesh", None)
+        if mesh is not None:
+            hidden_states = jax.lax.with_sharding_constraint(
+                hidden_states, jax.sharding.NamedSharding(mesh, P("data", None, "model"))
+            )
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # apply global RoPE to non-sliding layer only
+        if self.self_attn.is_sliding:
+            position_embeddings = position_embeddings_local
+        else:
+            position_embeddings = position_embeddings_global
+
+        outputs = self.self_attn(
+            hidden_states,
+            position_embeddings,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            deterministic=deterministic,
+            init_cache=init_cache,
+            output_attentions=output_attentions,
+        )
+        # residual connection
+        attn_output = self.post_attention_layernorm(outputs[0])
+        hidden_states = residual + attn_output
+
+        residual = hidden_states
+        hidden_states = self.pre_feedforward_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        mlp_output = self.post_feedforward_layernorm(hidden_states)
+        # residual connection
+        hidden_states = residual + mlp_output
+
+        return (hidden_states, attn_output, mlp_output)
+
+
+# Copied from transformers.models.gpt_neo.modeling_flax_gpt_neo.FlaxGPTNeoPreTrainedModel with GPTNeo->Gemma3, GPT_NEO->Gemma3, transformer->model
+class FlaxTPUGemma3PreTrainedModel(FlaxPreTrainedModel):
+    """
+    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
+    models.
+    """
+
+    config_class = TPUGemma3Config
+    base_model_prefix = "model"
+    module_class: nn.Module = None
+
+    def __init__(
+        self,
+        config: TPUGemma3Config,
+        input_shape: Tuple = (1, 1),
+        seed: int = 0,
+        dtype: jnp.dtype = jnp.float32,
+        _do_init: bool = True,
+        gradient_checkpointing: bool = False,
+        **kwargs,
+    ):
+        module = self.module_class(config=config, dtype=dtype, gradient_checkpointing=gradient_checkpointing, **kwargs)
+        super().__init__(config, module, input_shape=input_shape, seed=seed, dtype=dtype, _do_init=_do_init)
+
+    def enable_gradient_checkpointing(self):
+        self._module = self.module_class(
+            config=self.config,
+            dtype=self.dtype,
+            gradient_checkpointing=True,
+        )
+
+    @classmethod
+    def can_generate(cls) -> bool:
+        # disable generation, handled separately
+        # this is convenient since GenerationConfig.from_model_config(config) needs a pickleable config
+        return False
+
+    def init_weights(self, rng: jax.random.PRNGKey, input_shape: Tuple, params: FrozenDict = None) -> FrozenDict:
+        # init input tensors
+        input_ids = jnp.zeros(input_shape, dtype="i4")
+        attention_mask = jnp.ones_like(input_ids)
+        position_ids = jnp.broadcast_to(jnp.arange(jnp.atleast_2d(input_ids).shape[-1]), input_shape)
+        params_rng, dropout_rng = jax.random.split(rng)
+        rngs = {"params": params_rng, "dropout": dropout_rng}
+
+        random_params = self.module.init(rngs, input_ids, None, attention_mask, position_ids, return_dict=False)["params"]
+
+        if params is not None:
+            random_params = flatten_dict(unfreeze(random_params))
+            params = flatten_dict(unfreeze(params))
+            for missing_key in self._missing_keys:
+                params[missing_key] = random_params[missing_key]
+            self._missing_keys = set()
+            return freeze(unflatten_dict(params))
+        else:
+            return random_params
+
+    def init_cache(self, batch_size, max_length):
+        r"""
+        Args:
+            batch_size (`int`):
+                batch_size used for fast auto-regressive decoding. Defines the batch size of the initialized cache.
+            max_length (`int`):
+                maximum possible length for auto-regressive decoding. Defines the sequence length of the initialized
+                cache.
+        """
+        # init input variables to retrieve cache
+        input_ids = jnp.ones((batch_size, max_length))
+        attention_mask = jnp.ones_like(input_ids)
+        position_ids = jnp.broadcast_to(jnp.arange(jnp.atleast_2d(input_ids).shape[-1]), input_ids.shape)
+
+        init_variables = self.module.init(
+            jax.random.PRNGKey(0), input_ids, None, attention_mask, position_ids, return_dict=False, init_cache=True
+        )
+        return unfreeze(init_variables["cache"])
+
+    @add_start_docstrings_to_model_forward(TPU_GEMMA3_INPUTS_DOCSTRING)
+    def __call__(
+        self,
+        input_ids,
+        inputs_embeds=None,
+        attention_mask=None,
+        position_ids=None,
+        params: dict = None,
+        past_key_values: dict = None,
+        dropout_rng: jax.random.PRNGKey = None,
+        train: bool = False,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ):
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.return_dict
+
+        if input_ids is not None:
+            batch_size, sequence_length = input_ids.shape
+        else:
+            batch_size, sequence_length, _ = inputs_embeds.shape
+
+        if position_ids is None:
+            if past_key_values is not None:
+                raise ValueError("Make sure to provide `position_ids` when passing `past_key_values`.")
+
+            position_ids = jnp.broadcast_to(jnp.arange(sequence_length)[None, :], (batch_size, sequence_length))
+
+        if attention_mask is None:
+            attention_mask = jnp.ones((batch_size, sequence_length))
+
+        # Handle any PRNG if needed
+        rngs = {}
+        if dropout_rng is not None:
+            rngs["dropout"] = dropout_rng
+
+        inputs = {"params": params or self.params}
+
+        # if past_key_values are passed then cache is already initialized a private flag init_cache has to be passed down to ensure cache is used. It has to be made sure that cache is marked as mutable so that it can be changed by FlaxGemma3Attention module
+        if past_key_values:
+            inputs["cache"] = past_key_values
+            mutable = ["cache"]
+        else:
+            mutable = False
+
+        outputs = self.module.apply(
+            inputs,
+            jnp.array(input_ids, dtype="i4") if input_ids is not None else None,
+            inputs_embeds if inputs_embeds is not None else None,
+            jnp.array(attention_mask, dtype="i4"),
+            jnp.array(position_ids, dtype="i4"),
+            not train,
+            False,
+            output_attentions,
+            output_hidden_states,
+            return_dict,
+            rngs=rngs,
+            mutable=mutable,
+        )
+
+        # add updated cache to model output
+        if past_key_values is not None and return_dict:
+            outputs, past_key_values = outputs
+            outputs["past_key_values"] = unfreeze(past_key_values["cache"])
+            return outputs
+        elif past_key_values is not None and not return_dict:
+            outputs, past_key_values = outputs
+            outputs = outputs[:1] + (unfreeze(past_key_values["cache"]),) + outputs[1:]
+
+        return outputs
+
+
+# Copied from transformers.models.llama.modeling_flax_llama.FlaxLlamaLayerCollection with Llama->Gemma3
+class FlaxTPUGemma3LayerCollection(nn.Module):
+    config: TPUGemma3Config
+    dtype: jnp.dtype = jnp.float32
+    gradient_checkpointing: bool = False
+
+    def setup(self):
+        self.rotary_emb = FlaxTPUGemma3RotaryEmbedding(config=self.config)
+
+        mesh = getattr(self.config, "mesh", None)
+        del self.config.mesh
+        local_config = copy.deepcopy(self.config)
+        if mesh is not None:
+            self.config.mesh = mesh
+
+        local_config.rope_theta = self.config.rope_local_base_freq
+        local_config.rope_scaling = {"rope_type": "default"}
+        self.rotary_emb_local = FlaxTPUGemma3RotaryEmbedding(config=local_config)
+
+        if self.gradient_checkpointing:
+            FlaxTPUGemma3DecoderCheckpointLayer = remat(FlaxTPUGemma3DecoderLayer, static_argnums=(4, 5, 6))
+            self.blocks = [
+                FlaxTPUGemma3DecoderCheckpointLayer(self.config, layer_idx, dtype=self.dtype, name=str(layer_idx))
+                for layer_idx in range(self.config.num_hidden_layers)
+            ]
+        else:
+            self.blocks = [
+                FlaxTPUGemma3DecoderLayer(self.config, layer_idx, dtype=self.dtype, name=str(layer_idx))
+                for layer_idx in range(self.config.num_hidden_layers)
+            ]
+
+    def __call__(
+        self,
+        hidden_states,
+        attention_mask=None,
+        position_ids=None,
+        deterministic: bool = True,
+        init_cache: bool = False,
+        output_attentions: bool = False,
+        output_hidden_states: bool = False,
+        return_dict: bool = False,
+    ):
+        all_attentions = () if output_attentions else None
+        all_hidden_states = [(), ()] if output_hidden_states else None
+
+        position_embeddings_global = self.rotary_emb(hidden_states, position_ids)
+        position_embeddings_local = self.rotary_emb_local(hidden_states, position_ids)
+
+        if output_hidden_states:
+            all_hidden_states[0] += (hidden_states,)
+            all_hidden_states[1] += (hidden_states,)
+
+        for block_idx, block in enumerate(self.blocks):
+            layer_outputs = block(
+                hidden_states,
+                position_embeddings_global,
+                position_embeddings_local,
+                attention_mask,
+                position_ids,
+                deterministic,
+                init_cache,
+                output_attentions,
+            )
+            hidden_states = layer_outputs[0]
+
+            if output_hidden_states:
+                # last block is followed by norm - added later
+                if block_idx != len(self.blocks) - 1:
+                    all_hidden_states[0] += (hidden_states,)
+
+                all_hidden_states[1] += layer_outputs[1:]
+
+            if output_attentions:
+                raise NotImplementedError("Attention outputs are not implemented for TPUGemma3 (with projections).")
+
+        # this contains possible `None` values - `FlaxGemma3Module` will filter them out
+        outputs = (hidden_states, all_hidden_states, all_attentions)
+
+        return outputs
+
+
+# Copied from transformers.models.llama.modeling_flax_llama.FlaxLlamaModule with Llama->Gemma3
+class FlaxTPUGemma3Module(nn.Module):
+    config: TPUGemma3Config
+    dtype: jnp.dtype = jnp.float32
+    gradient_checkpointing: bool = False
+
+    def setup(self):
+        if self.config.project_mode == "wrap":
+            self.hidden_size = self.config.previous_hidden_size
+        else:
+            self.hidden_size = self.config.hidden_size
+
+        embedding_init = jax.nn.initializers.normal(stddev=self.config.initializer_range)
+
+        self.embed_tokens = nn.Embed(
+            self.config.vocab_size,
+            self.hidden_size,
+            embedding_init=embedding_init,
+            dtype=self.dtype,
+        )
+        self.layers = FlaxTPUGemma3LayerCollection(self.config, dtype=self.dtype, gradient_checkpointing=self.gradient_checkpointing)
+        self.norm = FlaxTPUGemma3RMSNorm(self.config, dtype=self.dtype, add_in_projection=self.config.project_mode == "wrap", add_out_projection=False)
+
+        if self.config.project_mode == "wrap":
+            self.embedding_projection = self.param("embedding_projection", lambda _, shape: jnp.empty(shape), (self.config.previous_hidden_size, self.config.hidden_size))
+
+    def embed(
+        self,
+        input_ids,
+    ):
+        inputs_embeds = self.embed_tokens(input_ids.astype("i4"))
+
+        if self.config.project_mode is not None:
+            scaler = self.config.previous_hidden_size ** 0.5
+        else:
+            scaler = self.config.hidden_size ** 0.5
+
+        inputs_embeds = inputs_embeds * scaler
+
+        if self.config.project_mode == "wrap":
+            inputs_embeds = inputs_embeds @ self.embedding_projection
+
+        return inputs_embeds
+
+    # Ignore copy
+    def __call__(
+        self,
+        input_ids,
+        inputs_embeds=None,
+        attention_mask=None,
+        position_ids=None,
+        deterministic=True,
+        init_cache: bool = False,
+        output_attentions: bool = False,
+        output_hidden_states: bool = False,
+        return_dict: bool = True,
+    ):
+        if inputs_embeds is None:
+            inputs_embeds = self.embed(input_ids)
+
+        outputs = self.layers(
+            inputs_embeds,
+            position_ids=position_ids,
+            attention_mask=attention_mask,
+            deterministic=deterministic,
+            init_cache=init_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = outputs[0]
+
+        if not self.config.skip_out_norm:
+            hidden_states = self.norm(hidden_states)
+
+        if output_hidden_states:
+            all_hidden_states = outputs[1]
+
+            all_hidden_states[0] += (hidden_states,)
+            outputs = (hidden_states, all_hidden_states) + outputs[2:]
+        else:
+            outputs = (hidden_states,) + outputs[1:]
+
+        if not return_dict:
+            return tuple(v for v in outputs if v is not None)
+
+        return FlaxBaseModelOutput(
+            last_hidden_state=hidden_states,
+            hidden_states=outputs[1],
+            attentions=outputs[-1],
+        )
+
+
+@add_start_docstrings(
+    "The bare Gemma3 Model transformer outputting raw hidden-states without any specific head on top.",
+    TPU_GEMMA3_START_DOCSTRING,
+)
+# Copied from transformers.models.llama.modeling_flax_llama.FlaxLlamaModel with Llama->Gemma3
+class FlaxTPUGemma3Model(FlaxTPUGemma3PreTrainedModel):
+    module_class = FlaxTPUGemma3Module
+
+
+append_call_sample_docstring(
+    FlaxTPUGemma3Model,
+    _CHECKPOINT_FOR_DOC,
+    FlaxBaseModelOutput,
+    _CONFIG_FOR_DOC,
+    real_checkpoint=_REAL_CHECKPOINT_FOR_DOC,
+)
+
+
+# Copied from transformers.models.llama.modeling_flax_llama.FlaxLlamaForCausalLMModule with Llama->Gemma3
+class FlaxTPUGemma3ForCausalLMModule(nn.Module):
+    config: TPUGemma3Config
+    dtype: jnp.dtype = jnp.float32
+    gradient_checkpointing: bool = False
+
+    def setup(self):
+        self.model = FlaxTPUGemma3Module(self.config, dtype=self.dtype, gradient_checkpointing=self.gradient_checkpointing)
+        self.lm_head = nn.Dense(
+            self.config.vocab_size,
+            use_bias=False,
+            dtype=self.dtype,
+            kernel_init=jax.nn.initializers.normal(stddev=self.config.initializer_range),
+        )
+
+    def embed(self, input_ids):
+        return self.model.embed(input_ids)
+
+    # Ignore copy
+    def __call__(
+        self,
+        input_ids,
+        inputs_embeds=None,
+        attention_mask=None,
+        position_ids=None,
+        deterministic: bool = True,
+        init_cache: bool = False,
+        output_attentions: bool = False,
+        output_hidden_states: bool = False,
+        return_dict: bool = True,
+    ):
+        outputs = self.model(
+            input_ids,
+            inputs_embeds=inputs_embeds,
+            position_ids=position_ids,
+            attention_mask=attention_mask,
+            deterministic=deterministic,
+            init_cache=init_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = outputs[0]
+        # should be skipped automatically in this case (since unused), but check if JIT actually does this
+        if not self.config.skip_out_norm:
+            if self.config.tie_word_embeddings:
+                shared_kernel = self.model.variables["params"]["embed_tokens"]["embedding"].T
+                lm_logits = self.lm_head.apply({"params": {"kernel": shared_kernel}}, hidden_states)
+            else:
+                lm_logits = self.lm_head(hidden_states)
+
+            lm_logits = jax.lax.with_sharding_constraint(
+                lm_logits,
+                jax.sharding.NamedSharding(getattr(self.config, "mesh"), P("data", None, "model")),
+            )
+
+            if self.config.final_logit_softcapping is not None:
+                lm_logits = lm_logits / self.config.final_logit_softcapping
+                lm_logits = jnp.tanh(lm_logits)
+                lm_logits = lm_logits * self.config.final_logit_softcapping
+        else:
+            lm_logits = None
+
+        if not return_dict:
+            return (lm_logits,) + outputs[1:]
+
+        return FlaxCausalLMOutput(logits=lm_logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions)
+
+
+@add_start_docstrings(
+    """
+    The Gemma3 Model transformer with a language modeling head (linear layer) on top.
+    """,
+    TPU_GEMMA3_START_DOCSTRING,
+)
+# Copied from transformers.models.gptj.modeling_flax_gptj.FlaxGPTJForCausalLM with GPTJ->Gemma3
+class FlaxTPUGemma3ForCausalLM(FlaxTPUGemma3PreTrainedModel):
+    module_class = FlaxTPUGemma3ForCausalLMModule
+
+    def prepare_inputs_for_generation(self, input_ids, max_length, attention_mask: Optional[jax.Array] = None):
+        # initializing the cache
+        batch_size, seq_length = input_ids.shape
+
+        past_key_values = self.init_cache(batch_size, max_length)
+        # Note that usually one would have to put 0's in the attention_mask for x > input_ids.shape[-1] and x < cache_length.
+        # But since Gemma3 uses a causal mask, those positions are masked anyways.
+        # Thus we can create a single static attention_mask here, which is more efficient for compilation
+        extended_attention_mask = jnp.ones((batch_size, max_length), dtype="i4")
+        if attention_mask is not None:
+            position_ids = attention_mask.cumsum(axis=-1) - 1
+            extended_attention_mask = lax.dynamic_update_slice(extended_attention_mask, attention_mask, (0, 0))
+        else:
+            position_ids = jnp.broadcast_to(jnp.arange(seq_length, dtype="i4")[None, :], (batch_size, seq_length))
+
+        return {
+            "past_key_values": past_key_values,
+            "attention_mask": extended_attention_mask,
+            "position_ids": position_ids,
+        }
+
+    def update_inputs_for_generation(self, model_outputs, model_kwargs):
+        model_kwargs["past_key_values"] = model_outputs.past_key_values
+        model_kwargs["position_ids"] = model_kwargs["position_ids"][:, -1:] + 1
+        return model_kwargs
+
+
+append_call_sample_docstring(
+    FlaxTPUGemma3ForCausalLM,
+    _CHECKPOINT_FOR_DOC,
+    FlaxCausalLMOutput,
+    _CONFIG_FOR_DOC,
+    real_checkpoint=_REAL_CHECKPOINT_FOR_DOC,
+)