Bug fixes (#3266)
* Fix mamba * Update loader.py * Update vision.py * Update loader.py * Filter vLLM standby logs (#3131) * filter vLLM standby logs * safeguard standby logger patch * Update unsloth/models/_utils.py * Update unsloth/models/_utils.py * Update unsloth/models/_utils.py --------- Co-authored-by: Daniel Han <danielhanchen@gmail.com> * Update loader.py * Add scaler * Update llama.py * Update _utils.py * Versioning * GPT OSS fix * GPT OSS fix * Update loader.py * Update vision.py * Update vision.py * Update loader.py * Update vision.py * Update vision.py * Update llama.py * Update llama.py * Update llama.py * Versioning * Update mapper.py * Update vision.py * Update vision.py * Update vision.py * Upcast norms * Update loader.py * Update vision.py * Upcast layernorms * Update llama.py * Update llama.py * Update llama.py * Update llama.py * Update llama.py * Update llama.py * Update save.py * Update rl.py * Update pyproject.toml * Update rl.py * Update rl_replacements.py * Update rl.py * Update rl.py * Update rl.py * Update _utils.py * Update __init__.py * Torch 2.8 * Update rl_replacements.py * Update loader.py * UNSLOTH_ENABLE_CCE * Fix * Update loader.py * Update loader.py * Update __init__.py * Update __init__.py * Update __init__.py * Update __init__.py * Import fixes * Update loader.py * Fix aimv2 issue * Update loader.py * Update import_fixes.py * Update import_fixes.py * Update loader.py * Update loader.py * Update loader.py * Upgrade * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update vision.py * Update vision.py * custom_datatype * recheck * Float16 * Update vision.py * Update vision.py * Update vision.py * Update vision.py * Update vision.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Bug fix * Update loader.py * Update loader.py * Update loader.py * Update loader.py * Update loader.py * torch_dtype * Update rl.py * Fix CE Loss * Versioning --------- Co-authored-by: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
This commit is contained in:
Daniel Han
GitHub
parent
490027f988
commit
5c1b0ae9dd
5 changed files with 27 additions and 14 deletions
|
|
@ -37,7 +37,7 @@ triton = [
|
|||
]
|
||||
|
||||
huggingface = [
|
||||
"unsloth_zoo>=2025.8.9",
|
||||
"unsloth_zoo>=2025.9.1",
|
||||
"packaging",
|
||||
"tyro",
|
||||
"transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1",
|
||||
|
|
@ -453,7 +453,7 @@ colab-ampere-torch220 = [
|
|||
"flash-attn>=2.6.3",
|
||||
]
|
||||
colab-new = [
|
||||
"unsloth_zoo>=2025.8.9",
|
||||
"unsloth_zoo>=2025.9.1",
|
||||
"packaging",
|
||||
"tyro",
|
||||
"transformers>=4.51.3,!=4.47.0,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1",
|
||||
|
|
|
|||
|
|
@ -214,7 +214,7 @@ elif DEVICE_TYPE == "xpu":
|
|||
# Check for unsloth_zoo
|
||||
try:
|
||||
unsloth_zoo_version = importlib_version("unsloth_zoo")
|
||||
if Version(unsloth_zoo_version) < Version("2025.8.8"):
|
||||
if Version(unsloth_zoo_version) < Version("2025.9.1"):
|
||||
print(
|
||||
"Unsloth: Please update Unsloth and Unsloth-Zoo to the latest version!\n"\
|
||||
"Do this via `pip install --upgrade --force-reinstall --no-cache-dir --no-deps unsloth unsloth_zoo`"
|
||||
|
|
|
|||
|
|
@ -1236,7 +1236,7 @@ def CausalLM_fast_forward(fast_forward_inference):
|
|||
# < 1024 Normal Unsloth uses less VRAM!
|
||||
if bsz*q_len <= 1024: RETURN_LOGITS = True
|
||||
|
||||
if not RETURN_LOGITS and HAS_CUT_CROSS_ENTROPY and labels is not None:
|
||||
if not RETURN_LOGITS and labels is not None:
|
||||
|
||||
n_items = kwargs.get("num_items_in_batch", None) or kwargs.get("n_items", None)
|
||||
|
||||
|
|
@ -1259,7 +1259,7 @@ def CausalLM_fast_forward(fast_forward_inference):
|
|||
mask = None,
|
||||
n_items = n_items,
|
||||
scaling = getattr(self, "accelerator_scaler", None),
|
||||
target_gb = 1,
|
||||
target_gb = None,
|
||||
torch_compile = True,
|
||||
logit_softcapping = logit_softcapping,
|
||||
)
|
||||
|
|
|
|||
|
|
@ -300,17 +300,30 @@ def MistralForCausalLM_fast_forward(
|
|||
# < 1024 Normal Unsloth uses less VRAM!
|
||||
if bsz * q_len <= 1024: RETURN_LOGITS = True
|
||||
|
||||
if not RETURN_LOGITS and HAS_CUT_CROSS_ENTROPY and os.environ.get("UNSLOTH_ENABLE_CCE", "1") != "0" and labels is not None:
|
||||
if not RETURN_LOGITS and labels is not None:
|
||||
n_items = kwargs.get("num_items_in_batch", None) or kwargs.get("n_items", None)
|
||||
logit_softcapping = getattr(self.config, "final_logit_softcapping", 0)
|
||||
loss = fused_linear_cross_entropy(
|
||||
hidden_states = hidden_states,
|
||||
lm_weight = lm_head,
|
||||
labels = labels,
|
||||
num_items_in_batch = n_items,
|
||||
logit_softcapping = logit_softcapping,
|
||||
)
|
||||
|
||||
# loss = fused_linear_cross_entropy(
|
||||
# hidden_states = hidden_states,
|
||||
# lm_weight = lm_head,
|
||||
# labels = labels,
|
||||
# num_items_in_batch = n_items,
|
||||
# logit_softcapping = logit_softcapping,
|
||||
# )
|
||||
loss = unsloth_fused_ce_loss(
|
||||
trainer = None,
|
||||
hidden_states = hidden_states,
|
||||
lm_head_weight = lm_head,
|
||||
lm_head_bias = None,
|
||||
labels = labels,
|
||||
mask = None,
|
||||
n_items = n_items,
|
||||
scaling = getattr(self, "accelerator_scaler", None),
|
||||
target_gb = None,
|
||||
torch_compile = True,
|
||||
logit_softcapping = logit_softcapping,
|
||||
)
|
||||
if not return_dict:
|
||||
output = (logits,) + outputs[1:]
|
||||
return (loss,) + output if loss is not None else output
|
||||
|
|
|
|||
|
|
@ -513,7 +513,7 @@ def _patch_trl_rl_trainers(trainer_file = "grpo_trainer"):
|
|||
"fp16" : False,
|
||||
"include_tokens_per_second" : False,
|
||||
"include_num_input_tokens_seen" : False,
|
||||
"auto_find_batch_size" : True, # Auto /2 batch size
|
||||
"auto_find_batch_size" : False, # Auto /2 batch size - too many people complained so removing
|
||||
"dataloader_pin_memory" : True,
|
||||
# Might fail so disable for now
|
||||
# "dataloader_persistent_workers" : True, # Keeps dataloader in RAM
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue