Will a QLoRA fine-tune of gemma-3-4b-it fit your GPU? Verdicts from measured AMD and NVIDIA probes.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Will this LoRA config train on a card with vram_gb GB? Returns verdict fits / tight / does_not_fit, the predicted peak in MB, the margin, how each part was derived, and any measured probe with the same settings.
Card memory as sold, in GB (8, 12, 16, 24...)
windows or linux; Windows spills instead of OOM
Base weights: bitsandbytes nf4 or bf16
Longest training example, in tokens
Tokens the loss covers (the response span)
VRAM already held by other programs
Will this LoRA config train on a card with vram_gb GB? Returns verdict fits / tight / does_not_fit, the predicted peak in MB, the margin, how each part was derived, and any measured probe with the same settings.
| Argument | Type | Description |
|---|---|---|
| os | string | windows or linux; Windows spills instead of OOM |
| dora | boolean | |
| rank* | integer | |
| model | string | |
| quant | string | Base weights: bitsandbytes nf4 or bf16 |
| seq_len | integer | Longest training example, in tokens |
| vram_gb* | number | Card memory as sold, in GB (8, 12, 16, 24...) |
| lora_dropout | number | |
| response_len | integer | Tokens the loss covers (the response span) |
| other_processes_mb | number | VRAM already held by other programs |
{
"type": "object",
"required": [
"vram_gb",
"rank"
],
"properties": {
"os": {
"type": "string",
"default": "linux",
"description": "windows or linux; Windows spills instead of OOM"
},
"dora": {
"type": "boolean",
"default": false
},
"rank": {
"type": "integer",
"minimum": 1
},
"model": {
"type": "string",
"default": "unsloth/gemma-3-4b-it"
},
"quant": {
"enum": [
"4bit",
"bf16"
],
"type": "string",
"default": "4bit",
"description": "Base weights: bitsandbytes nf4 or bf16"
},
"seq_len": {
"type": "integer",
"default": 7177,
"description": "Longest training example, in tokens"
},
"vram_gb": {
"type": "number",
"description": "Card memory as sold, in GB (8, 12, 16, 24...)"
},
"lora_dropout": {
"type": "number",
"default": 0
},
"response_len": {
"type": "integer",
"default": 237,
"description": "Tokens the loss covers (the response span)"
},
"other_processes_mb": {
"type": "number",
"default": 0,
"description": "VRAM already held by other programs"
}
}
}Largest LoRA rank (multiple of 8) that fits with margin on a card with vram_gb GB, and the largest that is tight (a real run may fit, the one-step probe would not).
| Argument | Type | Description |
|---|---|---|
| dora | boolean | |
| model | string | |
| quant | string | Base weights: bitsandbytes nf4 or bf16 |
| seq_len | integer | Longest training example, in tokens |
| vram_gb* | number | Card memory as sold, in GB (8, 12, 16, 24...) |
| lora_dropout | number | |
| response_len | integer | Tokens the loss covers (the response span) |
| other_processes_mb | number | VRAM already held by other programs |
{
"type": "object",
"required": [
"vram_gb"
],
"properties": {
"dora": {
"type": "boolean",
"default": false
},
"model": {
"type": "string",
"default": "unsloth/gemma-3-4b-it"
},
"quant": {
"enum": [
"4bit",
"bf16"
],
"type": "string",
"default": "4bit",
"description": "Base weights: bitsandbytes nf4 or bf16"
},
"seq_len": {
"type": "integer",
"default": 7177,
"description": "Longest training example, in tokens"
},
"vram_gb": {
"type": "number",
"description": "Card memory as sold, in GB (8, 12, 16, 24...)"
},
"lora_dropout": {
"type": "number",
"default": 0
},
"response_len": {
"type": "integer",
"default": 237,
"description": "Tokens the loss covers (the response span)"
},
"other_processes_mb": {
"type": "number",
"default": 0,
"description": "VRAM already held by other programs"
}
}
}The raw measured probe rows (RX 7900 XTX and RTX 3060 Ti), filtered by any of the given settings. gpu matches part of the card name, e.g. '7900' or '3060'.
| Argument | Type | Description |
|---|---|---|
| gpu | string | |
| dora | boolean | |
| rank | integer | |
| quant | string | |
| seq_len | integer | |
| lora_dropout | number |
{
"type": "object",
"properties": {
"gpu": {
"type": "string"
},
"dora": {
"type": "boolean"
},
"rank": {
"type": "integer"
},
"quant": {
"type": "string"
},
"seq_len": {
"type": "integer"
},
"lora_dropout": {
"type": "number"
}
}
}