"```python\nfrom modal_training_gym.common.models.qwen3_1_7b import Qwen3_1_7B\n```\n\nQwen3-1.7B (1.7 billion parameters) from Alibaba.\n\nPre-configured with full `ModelArchitecture` for Megatron-based\nframeworks (slime). Downloads from `Qwen/Qwen3-1.7B` on HuggingFace.\n\n**Inherits from:** `HFModelConfiguration`, `ModelConfig`\n\n## Fields\n\n| Field | Type | Default | Description |\n|-------|------|---------|-------------|\n| `model_name` | `str` | `\"Qwen/Qwen3-1.7B\"` | |\n| `model_path` | `str \\| None` | `None` | |\n| `architecture` | `ModelArchitecture \\| None` | `ModelArchitecture(num_layers=28, hidden_size=2048, ffn_hidden_size=6144, num_attention_heads=16, group_query_attention=True, num_query_groups=8, kv_channels=128, vocab_size=151936, normalization='RMSNorm', norm_epsilon=1e-06, swiglu=True, disable_bias_linear=True, qk_layernorm=True, untie_embeddings_and_output_weights=False, num_experts=0, moe_ffn_hidden_size=0, moe_shared_expert_intermediate_size=0, moe_grouped_gemm=False, moe_shared_expert_gate=False, moe_router_topk=0, moe_router_score_function='', moe_token_drop_policy='', moe_router_dtype='', moe_permute_fusion=False, moe_aux_loss_coeff=None, megatron_spec=None, megatron_model_type='', apply_layernorm_1p=False, use_gated_attention=False, attention_output_gate=False, use_rotary_position_embeddings=True, rotary_base=1000000, rotary_percent=1.0)` | |\n| `response_parser` | `Optional[Callable[[str], ParsedResponse]]` | `<function parse_qwen3_response at 0x7f29f916fba0>` | |\n\n## Methods\n\n### `download(self) -> 'None'`\n\nDownload or materialize weights into the model volume.\n\n### `parse_response(self, text: 'str') -> 'ParsedResponse'`\n\nParse raw model output into structured content.\n\n### `response_parser(text: 'str') -> 'ParsedResponse'`\n\nParse Qwen3-family model output into structured content.\n\n**Source:** [`modal_training_gym/common/models/qwen3_1_7b.py`](https://github.com/modal-projects/training-gym/blob/main/modal_training_gym/common/models/qwen3_1_7b.py)\n"

Qwen3-1.7B

Qwen3-1.7B (1.7 billion parameters) from Alibaba.

from modal_training_gym.common.models.qwen3_1_7b import Qwen3_1_7B

Qwen3-1.7B (1.7 billion parameters) from Alibaba.

Pre-configured with full ModelArchitecture for Megatron-based frameworks (slime). Downloads from Qwen/Qwen3-1.7B on HuggingFace.

Inherits from: HFModelConfiguration, ModelConfig

Fields

Field	Type	Default
`model_name`	`str`	`"Qwen/Qwen3-1.7B"`
`model_path`	`str \| None`	`None`
`architecture`	`ModelArchitecture \| None`	ModelArchitecture(num_layers=28, hidden_size=2048, ffn_hidden_size=6144, num_attention_heads=16, group_query_attention=True, num_query_groups=8, kv_channels=128, vocab_size=151936, normalization='RMSNorm', norm_epsilon=1e-06, swiglu=True, disable_bias_linear=True, qk_layernorm=True, untie_embeddings_and_output_weights=False, num_experts=0, moe_ffn_hidden_size=0, moe_shared_expert_intermediate_size=0, moe_grouped_gemm=False, moe_shared_expert_gate=False, moe_router_topk=0, moe_router_score_function='', moe_token_drop_policy='', moe_router_dtype='', moe_permute_fusion=False, moe_aux_loss_coeff=None, megatron_spec=None, megatron_model_type='', apply_layernorm_1p=False, use_gated_attention=False, attention_output_gate=False, use_rotary_position_embeddings=True, rotary_base=1000000, rotary_percent=1.0)
`response_parser`	`Optional[Callable[[str], ParsedResponse]]`	`<function parse_qwen3_response at 0x7f29f916fba0>`

Methods

`download(self) -> 'None'`

Download or materialize weights into the model volume.

`parse_response(self, text: 'str') -> 'ParsedResponse'`

Parse raw model output into structured content.

`response_parser(text: 'str') -> 'ParsedResponse'`

Parse Qwen3-family model output into structured content.

Source: modal_training_gym/common/models/qwen3_1_7b.py